initial commit

76 files changed, 28078 insertions, 0 deletions

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..bde577f
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,7 @@
+bin/
+release/
+art/
+project.4coder
+src/.gitignore
+generate-graph.sh
+download-translations.sh
\ No newline at end of file
diff --git a/COPYING b/COPYING
new file mode 100644
index 0000000..0c4ef7b
--- /dev/null
+++ b/COPYING
@@ -0,0 +1,25 @@
+BSD 2-Clause License
+
+Copyright (c) 2019, Aldrik Ramaekers
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
\ No newline at end of file
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..e08cbc4
--- /dev/null
+++ b/README.md
@@ -0,0 +1,22 @@
+# text-search
+Text-search is a GUI Program to find files and text within files for Linux. <br>
+text-search is a single and small executable.
+
+# Requirements
+
+### Linux
+- GCC
+- libglu1-mesa-dev, libgl1-mesa-dev (automatically installed with build/install script)
+- ld
+
+## Windows (not working)
+- MinGW32
+- Windres
+- ld
+
+# Build/Install
+run __build-linux.sh -r__ or __build-win.sh -r__ for building and running a debug build<br />
+run __install.sh__ as root to install to __/usr/local/bin/text-search__ or __C:\Users\\\<user>\\Desktop\\text-search.exe__ <br />
+
+# Config
+config.txt is stored at __~/.config/text-search/config.txt__ or __C:\Users\\\<user>\Local Settings\Application Data\text-search\config.txt__
diff --git a/build-linux.sh b/build-linux.sh
new file mode 100644
index 0000000..2a6d2c1
--- /dev/null
+++ b/build-linux.sh
@@ -0,0 +1,64 @@
+#!/bin/bash
+
+if [ $(dpkg-query -W -f='${Status}' libglu1-mesa-dev 2>/dev/null | grep -c "ok installed") -eq 0 ];
+then
+	if [ "$EUID" -ne 0 ]
+	then
+	  echo "Missing dependency: libglu1-mesa-dev, install this package or run this script as root"
+	else
+	  apt-get install libglu1-mesa-dev;
+	fi
+fi
+
+if [ $(dpkg-query -W -f='${Status}' libgl1-mesa-dev 2>/dev/null | grep -c "ok installed") -eq 0 ];
+then
+	if [ "$EUID" -ne 0 ]
+	then
+	  echo "Missing dependency: libgl1-mesa-dev, install this package or run this script as root"
+	else
+	  apt-get install libgl1-mesa-dev;
+	fi
+fi
+
+if [ $(dpkg-query -W -f='${Status}' libxrandr-dev 2>/dev/null | grep -c "ok installed") -eq 0 ];
+then
+	if [ "$EUID" -ne 0 ]
+	then
+	  echo "Missing dependency: libxrandr-dev, install this package or run this script as root"
+	else
+	  apt-get install libxrandr-dev;
+	fi
+fi
+
+rm -rf bin
+mkdir bin
+cd src
+
+ld -r -b binary -o ../bin/data.o \
+../data/imgs/en.png \
+../data/imgs/error.png \
+../data/imgs/folder.png \
+../data/imgs/nl.png \
+../data/imgs/search.png \
+../data/imgs/logo_64.png \
+../data/fonts/mono.ttf \
+../data/translations/en-English.mo \
+../data/translations/nl-Dutch.mo \
+
+gcc -Wall -g -m64 -DMODE_DEVELOPER -Wno-unused-label -rdynamic -Wno-unused-variable text_search.c ../bin/data.o -o ../bin/text-search -lX11 -lGL -lGLU -lXrandr -lm -lpthread -ldl
+
+rm -f ../bin/data.o
+
+if [ $? -ne 0 ]; then
+	cd ../
+	exit 1
+fi
+
+cd ../
+
+if [ "$1" == "-r" ]; then
+	cd bin
+	./text-search
+#    ./text-search --directory "/home/aldrik/Projects/text-search" --filter "*.c,*.h" --text "TODO(Aldrik)" --max-file-size 200 --locale "nl" --threads 5
+	cd ..
+fi
diff --git a/build-win.bat b/build-win.bat
new file mode 100644
index 0000000..13cbace
--- /dev/null
+++ b/build-win.bat
@@ -0,0 +1,22 @@
+@echo off
+
+windres misc/icon.rc -O coff -o misc/icon.res
+
+DEL /S /Q bin
+cd src
+
+ld -r -b binary -o ../bin/data.o ../data/imgs/en.png ../data/imgs/error.png ../data/imgs/folder.png ../data/imgs/nl.png ../data/imgs/search.png ../data/imgs/logo_64.png ../data/fonts/mono.ttf ../data/translations/en-English.mo ../data/translations/nl-Dutch.mo ../data/imgs/list.png ../data/imgs/delete.png ../data/imgs/exclaim.png
+
+if "%1"=="-w" (SET defs=-DMODE_DEVELOPER -DMODE_GDBDEBUG) else (SET defs=-DMODE_DEVELOPER)
+
+x86_64-w64-mingw32-gcc -m64 -Wall -g %defs% -Wno-unused-label -Wno-unused-variable mo_edit.c ../bin/data.o -o ../bin/moedit.exe ../misc/icon.res -lopengl32 -lkernel32 -lglu32 -lgdi32 -lcomdlg32 -lgdiplus -lole32 -lshlwapi
+
+DEL /Q "../bin/data.o"
+
+FOR %%A IN ("../bin/moedit.exe") DO set size=%%~zA
+echo size = %size%
+
+cd ../
+
+if "%1"=="-r" start bin/moedit.exe
+if "%1"=="-w" start gdb -ex run bin/moedit.exe
\ No newline at end of file
diff --git a/data/config.txt b/data/config.txt
new file mode 100644
index 0000000..c0d0264
--- /dev/null
+++ b/data/config.txt
@@ -0,0 +1,10 @@
+SEARCH_DIRECTORY = "/opt/textsearch/"
+SEARCH_DIRECTORIES = "1"
+SEARCH_TEXT = "*hello world*"
+FILE_FILTER = "*.txt,*.c"
+MAX_THEAD_COUNT = "20"
+MAX_FILE_SIZE = "200"
+WINDOW_WIDTH = "800"
+WINDOW_HEIGHT = "600"
+PARALLELIZE_SEARCH = "1"
+LOCALE = "en"
diff --git a/data/fonts/mono.ttf b/data/fonts/mono.ttf
new file mode 100644
index 0000000..1a39bc7
--- /dev/null
+++ b/data/fonts/mono.ttf
diff --git a/data/imgs/delete.png b/data/imgs/delete.png
new file mode 100644
index 0000000..fa0ab77
--- /dev/null
+++ b/data/imgs/delete.png
diff --git a/data/imgs/en.png b/data/imgs/en.png
new file mode 100644
index 0000000..02fa2a1
--- /dev/null
+++ b/data/imgs/en.png
diff --git a/data/imgs/error.png b/data/imgs/error.png
new file mode 100644
index 0000000..aacd682
--- /dev/null
+++ b/data/imgs/error.png
diff --git a/data/imgs/exclaim.png b/data/imgs/exclaim.png
new file mode 100644
index 0000000..40cc4b6
--- /dev/null
+++ b/data/imgs/exclaim.png
diff --git a/data/imgs/folder.png b/data/imgs/folder.png
new file mode 100644
index 0000000..9c0db10
--- /dev/null
+++ b/data/imgs/folder.png
diff --git a/data/imgs/list.png b/data/imgs/list.png
new file mode 100644
index 0000000..14c389d
--- /dev/null
+++ b/data/imgs/list.png
diff --git a/data/imgs/logo_64.png b/data/imgs/logo_64.png
new file mode 100644
index 0000000..78916bf
--- /dev/null
+++ b/data/imgs/logo_64.png
diff --git a/data/imgs/nl.png b/data/imgs/nl.png
new file mode 100644
index 0000000..505b9ce
--- /dev/null
+++ b/data/imgs/nl.png
diff --git a/data/imgs/search.png b/data/imgs/search.png
new file mode 100644
index 0000000..0a696b2
--- /dev/null
+++ b/data/imgs/search.png
diff --git a/data/marketing/cover_image.png b/data/marketing/cover_image.png
new file mode 100644
index 0000000..88e1baf
--- /dev/null
+++ b/data/marketing/cover_image.png
diff --git a/data/marketing/logo_32.png b/data/marketing/logo_32.png
new file mode 100644
index 0000000..89fa0dc
--- /dev/null
+++ b/data/marketing/logo_32.png
diff --git a/data/marketing/logo_512.ico b/data/marketing/logo_512.ico
new file mode 100644
index 0000000..c014524
--- /dev/null
+++ b/data/marketing/logo_512.ico
diff --git a/data/marketing/logo_512.png b/data/marketing/logo_512.png
new file mode 100644
index 0000000..2de1f36
--- /dev/null
+++ b/data/marketing/logo_512.png
diff --git a/data/marketing/logo_64.png b/data/marketing/logo_64.png
new file mode 100644
index 0000000..78916bf
--- /dev/null
+++ b/data/marketing/logo_64.png
diff --git a/data/marketing/main.png b/data/marketing/main.png
new file mode 100644
index 0000000..b7628e6
--- /dev/null
+++ b/data/marketing/main.png
diff --git a/data/marketing/main2.png b/data/marketing/main2.png
new file mode 100644
index 0000000..b5fd1b0
--- /dev/null
+++ b/data/marketing/main2.png
diff --git a/data/marketing/settings1.png b/data/marketing/settings1.png
new file mode 100644
index 0000000..92398e5
--- /dev/null
+++ b/data/marketing/settings1.png
diff --git a/data/marketing/settings2.png b/data/marketing/settings2.png
new file mode 100644
index 0000000..ecfb674
--- /dev/null
+++ b/data/marketing/settings2.png
diff --git a/data/marketing/text-search_banner.png b/data/marketing/text-search_banner.png
new file mode 100644
index 0000000..b78de48
--- /dev/null
+++ b/data/marketing/text-search_banner.png
diff --git a/data/translations/en-English.mo b/data/translations/en-English.mo
new file mode 100644
index 0000000..8f5ea46
--- /dev/null
+++ b/data/translations/en-English.mo
diff --git a/data/translations/nl-Dutch.mo b/data/translations/nl-Dutch.mo
new file mode 100644
index 0000000..005e775
--- /dev/null
+++ b/data/translations/nl-Dutch.mo
diff --git a/install.sh b/install.sh
new file mode 100644
index 0000000..d615539
--- /dev/null
+++ b/install.sh
@@ -0,0 +1,87 @@
+#!/bin/bash
+
+########################################################################
+########################################################################
+if [ "$(expr substr $(uname -s) 1 5)" == "Linux" ]; then
+
+if [ "$EUID" -ne 0 ]
+  then echo "Please run this script as root."
+  exit
+fi
+
+echo "Checking if dependencies are installed.."
+if [ $(dpkg-query -W -f='${Status}' libglu1-mesa-dev 2>/dev/null | grep -c "ok installed") -eq 0 ];
+then
+  apt-get install libglu1-mesa-dev;
+fi
+
+
+if [ $(dpkg-query -W -f='${Status}' libgl1-mesa-dev 2>/dev/null | grep -c "ok installed") -eq 0 ];
+then
+  apt-get install libgl1-mesa-dev;
+fi
+echo "Dependencies are installed"
+
+echo "Compiling program.."
+cd src
+
+ld -r -b binary -o ../bin/data.o \
+../data/imgs/en.png \
+../data/imgs/error.png \
+../data/imgs/folder.png \
+../data/imgs/nl.png \
+../data/imgs/search.png \
+../data/imgs/logo_64.png \
+../data/fonts/mono.ttf \
+../data/translations/en-English.mo \
+../data/translations/nl-Dutch.mo \
+
+gcc -Wall -O3 -m64 -Wno-unused-label -Wno-unused-variable text_search.c ../bin/data.o -o ../bin/text-search -lX11 -lGL -lGLU -lXrandr -lm -lpthread -ldl
+
+rm -f ../bin/data.o
+
+echo "Done compiling program"
+
+cp --remove-destination ../bin/text-search /usr/local/bin/text-search
+
+cd ../
+
+########################################################################
+########################################################################
+elif [ "$(expr substr $(uname -s) 1 10)" == "MINGW64_NT" ]; then
+
+windres misc/icon.rc -O coff -o misc/icon.res
+
+echo "Compiling program.."
+cd src
+
+ld -r -b binary -o ../bin/data.o \
+../data/imgs/en.png \
+../data/imgs/error.png \
+../data/imgs/folder.png \
+../data/imgs/nl.png \
+../data/imgs/search.png \
+../data/imgs/logo_64.png \
+../data/fonts/mono.ttf \
+../data/translations/en-English.mo \
+../data/translations/nl-Dutch.mo \
+
+x86_64-w64-mingw32-gcc -Wall -m64 -O3 -Wno-unused-label -Wno-unused-variable text_search.c ../bin/data.o -o ../bin/text-search.exe ../misc/icon.res -lopengl32 -lkernel32 -lglu32 -lgdi32 -lcomdlg32 -lgdiplus -lole32 -lshlwapi
+
+rm -f ../bin/data.o
+
+echo "Done compiling program, text-search.exe is located in 'C:\Manually installed'"
+
+cp --remove-destination "../bin/text-search.exe" "C:\Manually installed programs\text-search.exe"
+
+cd ../
+
+########################################################################
+########################################################################
+elif [ "$(uname)" == "Darwin" ]; then
+    echo "OSX Platform not supported"
+########################################################################
+########################################################################
+elif [ "$(expr substr $(uname -s) 1 10)" == "MINGW32_NT" ]; then
+	echo "32bit Windows versions not supported"
+fi
\ No newline at end of file
diff --git a/misc/icon.rc b/misc/icon.rc
new file mode 100644
index 0000000..85b2f1d
--- /dev/null
+++ b/misc/icon.rc
@@ -0,0 +1,28 @@
+#include <windows.h>
+#include <wingdi.h>
+#include <winuser.h>
+
+RC_LOGO ICON "logo_512.ico"
+1 VERSIONINFO
+FILEVERSION     0,1,0,0
+PRODUCTVERSION  0,1,0,0
+BEGIN
+  BLOCK "StringFileInfo"
+  BEGIN
+    BLOCK "080904E4"
+    BEGIN
+      VALUE "CompanyName", "Aldrik Ramaekers"
+      VALUE "FileDescription", "Text-search"
+      VALUE "FileVersion", "1.0"
+      VALUE "InternalName", "Text-search"
+      VALUE "LegalCopyright", "Aldrik Ramaekers"
+      VALUE "OriginalFilename", "Text-search.exe"
+      VALUE "ProductName", "Text-search"
+      VALUE "ProductVersion", "1.0"
+    END
+  END
+  BLOCK "VarFileInfo"
+  BEGIN
+    VALUE "Translation", 0x809, 1252
+  END
+END
diff --git a/misc/icon.res b/misc/icon.res
new file mode 100644
index 0000000..ef35332
--- /dev/null
+++ b/misc/icon.res
diff --git a/misc/logo_512.ico b/misc/logo_512.ico
new file mode 100644
index 0000000..c014524
--- /dev/null
+++ b/misc/logo_512.ico
diff --git a/src/array.c b/src/array.c
new file mode 100644
index 0000000..2c8127a
--- /dev/null
+++ b/src/array.c
@@ -0,0 +1,204 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+array array_create(u64 entry_size)
+{
+	array new_array;
+	new_array.length = 0;
+	new_array.reserved_length = 0;
+	new_array.entry_size = entry_size;
+	new_array.data = 0;
+	new_array.reserve_jump = 1;
+	new_array.mutex = mutex_create_recursive();
+	
+	return new_array;
+}
+
+int array_push(array *array, void *data)
+{
+	assert(array);
+	assert(data);
+	assert(array->reserve_jump >= 1);
+	
+	mutex_lock(&array->mutex);
+	array->length++;
+	
+	if (!array->data)
+	{
+		array->data = mem_alloc(array->entry_size * array->reserve_jump);
+		array->reserved_length = array->reserve_jump;
+	}
+	
+	if (array->reserved_length < array->length)
+	{
+		array->reserved_length += array->reserve_jump;
+		array->data = mem_realloc(array->data, (array->reserved_length*array->entry_size));
+	}
+	
+	memcpy(array->data + ((array->length-1) * array->entry_size),
+		   data, array->entry_size);
+	
+	s32 result = array->length -1;
+	mutex_unlock(&array->mutex);
+	return result;
+}
+
+int array_push_size(array *array, void *data, s32 data_size)
+{
+	assert(array);
+	assert(data);
+	assert(array->reserve_jump >= 1);
+	
+	mutex_lock(&array->mutex);
+	array->length++;
+	
+	if (!array->data)
+	{
+		array->data = mem_alloc(array->entry_size * array->reserve_jump);
+		array->reserved_length = array->reserve_jump;
+	}
+	
+	if (array->reserved_length < array->length)
+	{
+		array->reserved_length += array->reserve_jump;
+		array->data = mem_realloc(array->data, (array->reserved_length*array->entry_size));
+	}
+	
+	memcpy(array->data + ((array->length-1) * array->entry_size),
+		   data, data_size);
+	
+	// fill remaining space with 0
+	if (array->entry_size > data_size)
+	{
+		s32 remaining = array->entry_size - data_size;
+		memset(array->data + ((array->length-1) * array->entry_size) + data_size,
+			   0, remaining);
+	}
+	
+	s32 result = array->length -1;
+	mutex_unlock(&array->mutex);
+	return result;
+}
+
+void array_reserve(array *array, u32 reserve_count)
+{
+	assert(array);
+	
+	mutex_lock(&array->mutex);
+	u32 reserved_count = array->reserved_length - array->length;
+	reserve_count -= reserved_count;
+	
+	if (reserve_count > 0)
+	{
+		array->reserved_length += reserve_count;
+		
+		if (array->data)
+			array->data = mem_realloc(array->data, (array->reserved_length*array->entry_size));
+		else
+			array->data = mem_alloc(array->reserved_length*array->entry_size);
+	}
+	mutex_unlock(&array->mutex);
+}
+
+void array_remove_at(array *array, u32 at)
+{
+	assert(array);
+	assert(at >= 0);
+	assert(at < array->length);
+	
+	mutex_lock(&array->mutex);
+	if (array->length > 1)
+	{
+		int offset = at * array->entry_size;
+		int size = (array->length - at - 1) * array->entry_size;
+		memcpy(array->data + offset,
+			   array->data + offset + array->entry_size,
+			   size);
+		
+		//array->data = realloc(array->data, array->length * array->entry_size);
+	}
+	
+	array->length--;
+	mutex_unlock(&array->mutex);
+}
+
+void array_remove(array *array, void *ptr)
+{
+	mutex_lock(&array->mutex);
+	int offset = ptr - array->data;
+	int at = offset / array->entry_size;
+	array_remove_at(array, at);
+	mutex_unlock(&array->mutex);
+}
+
+void array_remove_by(array *array, void *data)
+{
+	assert(array);
+	
+	mutex_lock(&array->mutex);
+	for (int i = 0; i < array->length; i++)
+	{
+		void *d = array_at(array, i);
+		if (memcmp(d, data, array->entry_size) == 0)
+		{
+			array_remove_at(array, i);
+			return;
+		}
+	}
+	mutex_unlock(&array->mutex);
+}
+
+void *array_at(array *array, u32 at)
+{
+	mutex_lock(&array->mutex);
+	assert(array);
+	assert(at >= 0);
+	assert(at < array->length);
+	
+	void *result =  array->data + (at * array->entry_size);
+	mutex_unlock(&array->mutex);
+	return result;
+}
+
+void array_destroy(array *array)
+{
+	assert(array);
+	mem_free(array->data);
+	mutex_destroy(&array->mutex);
+}
+
+void array_swap(array *array, u32 swap1, u32 swap2)
+{
+	assert(array);
+	assert(swap2 >= 0);
+	assert(swap2 < array->length);
+	if (swap1 == swap2) return;
+	
+	void *swap1_at = array_at(array, swap1);
+	void *swap2_at = array_at(array, swap2);
+	
+	mutex_lock(&array->mutex);
+	char swap1_buffer[array->entry_size];
+	memcpy(swap1_buffer, swap1_at, array->entry_size);
+	memcpy(swap1_at, swap2_at, array->entry_size);
+	memcpy(swap2_at, swap1_buffer, array->entry_size);
+	mutex_unlock(&array->mutex);
+}
+
+array array_copy(array *arr)
+{
+	array new_array;
+	new_array.length = arr->length;
+	new_array.reserved_length = arr->reserved_length;
+	new_array.entry_size = arr->entry_size;
+	new_array.data = mem_alloc(new_array.entry_size*new_array.reserved_length);
+	new_array.mutex = mutex_create();
+	
+	mutex_lock(&arr->mutex);
+	memcpy(new_array.data, arr->data, new_array.entry_size*new_array.reserved_length);
+	mutex_unlock(&arr->mutex);
+	return new_array;
+}
\ No newline at end of file
diff --git a/src/array.h b/src/array.h
new file mode 100644
index 0000000..f18f780
--- /dev/null
+++ b/src/array.h
@@ -0,0 +1,32 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_ARRAY
+#define INCLUDE_ARRAY
+
+typedef struct t_array
+{
+	u32 length;
+	u32 reserved_length;
+	u64 entry_size;
+	u16 reserve_jump;
+	void *data;
+	mutex mutex;
+} array;
+
+array array_create(u64 entry_size);
+int array_push(array *array, void *data);
+int array_push_size(array *array, void *data, s32 data_size);
+void array_remove_at(array *array, u32 at);
+void array_remove(array *array, void *ptr);
+void array_remove_by(array *array, void *data);
+void *array_at(array *array, u32 at);
+void array_destroy(array *array);
+void array_swap(array *array, u32 swap1, u32 swap2);
+void array_reserve(array *array, u32 reserve_count);
+array array_copy(array *array);
+
+#endif
\ No newline at end of file
diff --git a/src/assets.c b/src/assets.c
new file mode 100644
index 0000000..37334dd
--- /dev/null
+++ b/src/assets.c
@@ -0,0 +1,313 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+void assets_create()
+{
+	assets asset_collection;
+	asset_collection.images = array_create(sizeof(image));
+	asset_collection.fonts = array_create(sizeof(font));
+	
+	array_reserve(&asset_collection.images, ASSET_IMAGE_COUNT);
+	array_reserve(&asset_collection.fonts, ASSET_FONT_COUNT);
+	
+	asset_collection.queue.queue = array_create(sizeof(asset_task));
+	asset_collection.post_process_queue = array_create(sizeof(asset_task));
+	
+	array_reserve(&asset_collection.queue.queue, ASSET_QUEUE_COUNT);
+	array_reserve(&asset_collection.post_process_queue, ASSET_QUEUE_COUNT);
+	
+	asset_mutex = mutex_create();
+	asset_collection.valid = true;
+	asset_collection.done_loading_assets = false;
+	
+	global_asset_collection = asset_collection;
+}
+
+inline static bool is_big_endian()
+{
+	volatile uint32_t i=0x01234567;
+    // return 1 for big endian, 0 for little endian.
+    return !((*((uint8_t*)(&i))) == 0x67);
+}
+
+void assets_do_post_process()
+{
+	mutex_lock(&asset_mutex);
+	
+	for (int i = 0; i < global_asset_collection.post_process_queue.length; i++)
+	{
+		asset_task *task = array_at(&global_asset_collection.post_process_queue, i);
+		
+		if (task->type == ASSET_IMAGE)
+		{
+			if (task->image->data && task->valid)
+			{
+				glGenTextures(1, &task->image->textureID);
+				glBindTexture(GL_TEXTURE_2D, task->image->textureID);
+				
+				s32 flag = is_big_endian() ? GL_UNSIGNED_INT_8_8_8_8 : 
+				GL_UNSIGNED_INT_8_8_8_8_REV;
+				
+				glTexImage2D(GL_TEXTURE_2D, 0,GL_RGBA8, task->image->width, 
+							 task->image->height, 0,  GL_RGBA, flag, task->image->data);
+				glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+				glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+				task->image->loaded = true;
+				
+				if (!task->image->keep_in_memory)
+					stbi_image_free(task->image->data);
+			}
+		}
+		else if (task->type == ASSET_FONT)
+		{
+			if (task->valid)
+			{
+				for (s32 i = TEXT_CHARSET_START; i < TEXT_CHARSET_END; i++)
+				{
+					glyph *g = &task->font->glyphs[i];
+					
+					glGenTextures(1, &g->textureID);
+					glBindTexture(GL_TEXTURE_2D, g->textureID);
+					
+					glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST );
+					glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST );
+                    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE );
+                    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE );
+					glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+					glTexImage2D( GL_TEXTURE_2D, 0, GL_ALPHA, g->width,g->height,
+								 0, GL_ALPHA, GL_UNSIGNED_BYTE, g->bitmap );
+					
+					mem_free(g->bitmap);
+				}
+				
+				task->font->loaded = true;
+			}
+		}
+		
+		array_remove_at(&global_asset_collection.post_process_queue, i);
+	}
+	
+	mutex_unlock(&asset_mutex);
+}
+
+bool assets_queue_worker_load_image(image *image)
+{
+	set_active_directory(binary_path);
+	
+	image->data = stbi_load_from_memory(image->start_addr,
+										image->end_addr - image->start_addr,
+										&image->width,
+										&image->height,
+										&image->channels,
+										STBI_rgb_alpha);
+	
+	return !(image->data == 0);
+}
+
+bool assets_queue_worker_load_font(font *font)
+{
+	unsigned char *ttf_buffer = (unsigned char*)font->start_addr;
+	
+    stbtt_fontinfo info;
+    if (!stbtt_InitFont(&info, ttf_buffer, stbtt_GetFontOffsetForIndex(ttf_buffer,0)))
+    {
+		return false;
+	}
+	float scale = stbtt_ScaleForPixelHeight(&info, font->size);
+	
+	for (s32 i = TEXT_CHARSET_START; i < TEXT_CHARSET_END; i++)
+	{
+		s32 w, h, xoff, yoff;
+		
+		glyph new_glyph;
+		new_glyph.bitmap = stbtt_GetCodepointBitmap(&info, 0, scale, i, &w, &h, &xoff, &yoff);
+		new_glyph.width = w;
+		new_glyph.height = h;
+		new_glyph.xoff = xoff;
+		new_glyph.yoff = yoff;
+		
+		if (i == 'M') font->px_h = -yoff;
+		if (i == ' ') new_glyph.xoff = font->size/3;
+		
+		font->glyphs[i-TEXT_CHARSET_START] = new_glyph;
+	}
+	
+	font->info = info;
+	font->scale = scale;
+	
+	return true;
+}
+
+void *assets_queue_worker()
+{
+	while (global_asset_collection.valid && !global_asset_collection.done_loading_assets)
+	{
+		if (mutex_trylock(&asset_mutex))
+		{
+			int queue_length = global_asset_collection.queue.queue.length;
+			if (!queue_length) 
+			{
+				mutex_unlock(&asset_mutex);
+				continue;
+			}
+			
+			asset_task *task = array_at(&global_asset_collection.queue.queue, 0);
+			asset_task buf = *task;
+			array_remove_at(&global_asset_collection.queue.queue, 0);
+			mutex_unlock(&asset_mutex);
+			
+			// load here
+			if (buf.type == ASSET_IMAGE)
+			{
+				bool result = assets_queue_worker_load_image(buf.image);
+				buf.valid = result;
+			}
+			else if (buf.type == ASSET_FONT)
+			{
+				bool result = assets_queue_worker_load_font(buf.font);
+				buf.valid = result;
+			}
+			
+			mutex_lock(&asset_mutex);
+			
+			assert(global_asset_collection.post_process_queue.reserved_length > 
+				   global_asset_collection.post_process_queue.length);
+			
+			array_push(&global_asset_collection.post_process_queue, &buf);
+			mutex_unlock(&asset_mutex);
+		}
+		
+		// 3 ms
+		thread_sleep(3000);
+	}
+	
+	return 0;
+}
+
+image *assets_load_image(u8 *start_addr, u8 *end_addr, bool keep_in_memory)
+{
+	// check if image is already loaded or loading
+	for (int i = 0; i < global_asset_collection.images.length; i++)
+	{
+		image *img_at = array_at(&global_asset_collection.images, i);
+		
+		if (start_addr == img_at->start_addr)
+		{
+			// image is already loaded/loading
+			img_at->references++;
+			return img_at;
+		}
+	}
+	
+	image new_image;
+	new_image.loaded = false;
+	new_image.start_addr = start_addr;
+	new_image.end_addr = end_addr;
+	new_image.references = 1;
+	new_image.keep_in_memory = keep_in_memory;
+	
+	// NOTE(Aldrik): we should never realloc the image array because pointers will be 
+	// invalidated.
+	assert(global_asset_collection.images.reserved_length > global_asset_collection.images.length);
+	
+	int index = array_push(&global_asset_collection.images, &new_image);
+	
+	asset_task task;
+	task.type = ASSET_IMAGE;
+	task.image = array_at(&global_asset_collection.images, index);
+	
+	mutex_lock(&asset_mutex);
+	array_push(&global_asset_collection.queue.queue, &task);
+	mutex_unlock(&asset_mutex);
+	
+	return task.image;
+}
+
+void assets_destroy_image(image *image_to_destroy)
+{
+	if (image_to_destroy->references == 1)
+	{
+		glBindTexture(GL_TEXTURE_2D, 0);
+		glDeleteTextures(1, &image_to_destroy->textureID);
+		
+		if (image_to_destroy->keep_in_memory)
+			stbi_image_free(image_to_destroy->data);
+		
+		//array_remove(&global_asset_collection.images, image_at);
+	}
+	else
+	{
+		image_to_destroy->references--;
+	}
+}
+
+font *assets_load_font(u8 *start_addr, u8 *end_addr, s16 size)
+{
+	//assert(!(size % 4));
+	for (int i = 0; i < global_asset_collection.fonts.length; i++)
+	{
+		font *font_at = array_at(&global_asset_collection.fonts, i);
+		
+		if (start_addr == font_at->start_addr && font_at->size == size)
+		{
+			// font is already loaded/loading
+			font_at->references++;
+			return font_at;
+		}
+	}
+	
+	font new_font;
+	new_font.loaded = false;
+	new_font.start_addr = start_addr;
+	new_font.end_addr = end_addr;
+	new_font.size = size;
+	new_font.references = 1;
+	
+	// NOTE(Aldrik): we should never realloc the font array because pointers will be 
+	// invalidated.
+	assert(global_asset_collection.fonts.reserved_length > global_asset_collection.fonts.length);
+	
+	int index = array_push(&global_asset_collection.fonts, &new_font);
+	
+	asset_task task;
+	task.type = ASSET_FONT;
+	task.font = array_at(&global_asset_collection.fonts, index);
+	
+	mutex_lock(&asset_mutex);
+	array_push(&global_asset_collection.queue.queue, &task);
+	mutex_unlock(&asset_mutex);
+	
+	return task.font;
+}
+
+void assets_destroy_font(font *font_to_destroy)
+{
+	if (font_to_destroy->references == 1)
+	{
+		//glBindTexture(GL_TEXTURE_2D, 0);
+		//glDeleteTextures(1, font_to_destroy->textureIDs);
+	}
+	else
+	{
+		font_to_destroy->references--;
+	}
+}
+
+void assets_destroy()
+{
+	global_asset_collection.valid = false;
+	global_asset_collection.done_loading_assets = false;
+	
+	array_destroy(&global_asset_collection.images);
+	array_destroy(&global_asset_collection.fonts);
+	
+	array_destroy(&global_asset_collection.queue.queue);
+	array_destroy(&global_asset_collection.post_process_queue);
+	
+	mem_free(binary_path);
+	
+	mutex_destroy(&asset_mutex);
+}
diff --git a/src/assets.h b/src/assets.h
new file mode 100644
index 0000000..3cfac52
--- /dev/null
+++ b/src/assets.h
@@ -0,0 +1,149 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_ASSETS
+#define INCLUDE_ASSETS
+
+#ifndef ASSET_IMAGE_COUNT
+#define ASSET_IMAGE_COUNT 10
+#endif
+
+#ifndef ASSET_FONT_COUNT
+#define ASSET_FONT_COUNT 10
+#endif
+
+#ifndef ASSET_QUEUE_COUNT
+#define ASSET_QUEUE_COUNT 10
+#endif
+
+// binary blobs
+extern u8 _binary____data_imgs_en_png_start[];
+extern u8 _binary____data_imgs_en_png_end[];
+
+extern u8 _binary____data_imgs_error_png_start[];
+extern u8 _binary____data_imgs_error_png_end[];
+
+extern u8 _binary____data_imgs_folder_png_start[];
+extern u8 _binary____data_imgs_folder_png_end[];
+
+extern u8 _binary____data_imgs_nl_png_start[];
+extern u8 _binary____data_imgs_nl_png_end[];
+
+extern u8 _binary____data_imgs_search_png_start[];
+extern u8 _binary____data_imgs_search_png_end[];
+
+extern u8 _binary____data_imgs_logo_64_png_start[];
+extern u8 _binary____data_imgs_logo_64_png_end[];
+
+extern u8 _binary____data_imgs_logo_512_png_start[];
+extern u8 _binary____data_imgs_logo_512_png_end[];
+
+extern u8 _binary____data_fonts_mono_ttf_start[];
+extern u8 _binary____data_fonts_mono_ttf_end[];
+
+extern u8 _binary____data_translations_en_English_mo_start[];
+extern u8 _binary____data_translations_en_English_mo_end[];
+
+extern u8 _binary____data_translations_nl_Dutch_mo_start[];
+extern u8 _binary____data_translations_nl_Dutch_mo_end[];
+
+extern u8 _binary____data_imgs_list_png_start[];
+extern u8 _binary____data_imgs_list_png_end[];
+
+extern u8 _binary____data_imgs_delete_png_start[];
+extern u8 _binary____data_imgs_delete_png_end[];
+
+extern u8 _binary____data_imgs_exclaim_png_start[];
+extern u8 _binary____data_imgs_exclaim_png_end[];
+
+
+typedef struct t_image {
+	u8 *start_addr;
+	u8 *end_addr;
+	bool loaded;
+	bool keep_in_memory;
+	s32 width;
+	s32 height;
+	s32 channels;
+	void *data;
+	s16 references;
+	GLuint textureID;
+} image;
+
+#define TEXT_CHARSET_START 0
+#define TEXT_CHARSET_END 2000
+#define TOTAL_GLYPHS TEXT_CHARSET_END-TEXT_CHARSET_START
+
+typedef struct t_glyph
+{
+	s32 width;
+	s32 height;
+	s32 xoff;
+	s32 yoff;
+	void *bitmap;
+	GLuint textureID;
+} glyph;
+
+typedef struct t_font
+{
+	u8 *start_addr;
+	u8 *end_addr;
+	bool loaded;
+	s16 references;
+	s16 size;
+	s32 px_h;
+	float32 scale;
+	stbtt_fontinfo info;
+	glyph glyphs[TOTAL_GLYPHS];
+} font;
+
+typedef enum t_asset_task_type
+{
+	ASSET_IMAGE,
+	ASSET_FONT,
+} asset_task_type;
+
+typedef struct t_asset_task
+{
+	s8 type;
+	bool valid;
+	union {
+		image *image;
+		font *font;
+	};
+} asset_task;
+
+typedef struct t_asset_queue {
+	array queue;
+} asset_queue;
+
+typedef struct t_assets {
+	array images;
+	array fonts;
+	asset_queue queue;
+	array post_process_queue;
+	bool valid;
+	bool done_loading_assets;
+} assets;
+
+char *binary_path;
+
+mutex asset_mutex;
+assets global_asset_collection;
+
+void assets_create();
+void assets_destroy();
+
+void assets_do_post_process();
+void *assets_queue_worker();
+
+image *assets_load_image(u8 *start_addr, u8 *end_addr, bool keep_in_memory);
+void assets_destroy_image(image *image);
+
+font *assets_load_font(u8 *start_addr, u8 *end_addr, s16 size);
+void assets_destroy_font(font *font);
+
+#endif
\ No newline at end of file
diff --git a/src/camera.c b/src/camera.c
new file mode 100644
index 0000000..5bde9f9
--- /dev/null
+++ b/src/camera.c
@@ -0,0 +1,22 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+void camera_apply_transformations(platform_window *window, camera *camera)
+{
+	s32 x = (window->width/2)+(camera->x);
+	s32 y =  (window->height/2)+(camera->y);
+	glTranslatef(x, y, 0.0f);
+	glRotatef(camera->rotation, 0.0f, 0.0f, 1.0f);
+	glTranslatef(-x, -y, 0.0f);
+	
+	glMatrixMode(GL_PROJECTION);
+	glLoadIdentity();
+    
+    glOrtho(camera->x, window->width+camera->x, 
+            window->height+camera->y, camera->y, -100, 100);
+    
+	glMatrixMode(GL_MODELVIEW);
+}
\ No newline at end of file
diff --git a/src/camera.h b/src/camera.h
new file mode 100644
index 0000000..c6e4093
--- /dev/null
+++ b/src/camera.h
@@ -0,0 +1,19 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_CAMERA
+#define INCLUDE_CAMERA
+
+typedef struct t_camera
+{
+	float32 x;
+	float32 y;
+	float32 rotation;
+} camera;
+
+void camera_apply_transformations(platform_window *window, camera *camera);
+
+#endif
\ No newline at end of file
diff --git a/src/command_line.c b/src/command_line.c
new file mode 100644
index 0000000..b1718f9
--- /dev/null
+++ b/src/command_line.c
@@ -0,0 +1,265 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#if 0
+void find_text_in_files(search_result *search_result);
+s32 prepare_search_directory_path(char *path, s32 len);
+search_result *create_empty_search_result();
+void do_search();
+bool export_results(search_result *result);
+
+static void print_license_message()
+{
+	time_t t = time(0);
+	struct tm *now = localtime(&t);
+	
+	printf("BSD 2-Clause License\n"
+		   "\n"
+		   "Copyright (c) %d, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com\n"
+		   "All rights reserved.\n"
+		   "\n"
+		   "Redistribution and use in source and binary forms, with or without\n"
+		   "modification, are permitted provided that the following conditions are met:\n"
+		   "\n"
+		   "1. Redistributions of source code must retain the above copyright notice, this\n"
+		   "list of conditions and the following disclaimer.\n"
+		   "\n"
+		   "2. Redistributions in binary form must reproduce the above copyright notice,\n"
+		   "this list of conditions and the following disclaimer in the documentation\n"
+		   "and/or other materials provided with the distribution.\n"
+		   "\n"
+		   "THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n"
+		   "AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n"
+		   "IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n"
+		   "DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE\n"
+		   "FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\n"
+		   "DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR\n"
+		   "SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER\n"
+		   "CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,\n"
+		   "OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n"
+		   "OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n\n", now->tm_year+1900);
+	
+	printf("The following parts of the Software are separately licensed:\n"
+		   "- The Liberation font family is licensed under the SIL Open Font License" "(version 2 onwards)\n\n");
+}
+
+static void print_help_message()
+{
+#define explain_required_argument(c,e) printf("   %-18s%-18s%s\n", c, "REQUIRED", e);
+#define explain_other_argument(c,e) printf("   %-18s%-18s%s\n", c, " ", e);
+#define explain_optional_argument(c,d,e) printf("   %-18s%-18s%s\n", c, d, e);
+#define DEFAULT(d) "default=\""d"\""
+	
+	printf("Usage: text-search [OPTION] ... [OPTION] ...\n");
+	printf("Example: text-search "
+		   "--directory \"/home/john/Documents\" --text \"homework\" --recursive 0\n");
+	printf("Text-search, search for files and text within files.\n");
+	printf("Matches will be printed to console in format: [path]:[line]:[filter]\n\n");
+	
+	printf("Available arguments:\n");
+	explain_required_argument("--directory", "The directory to search");
+	explain_optional_argument("--text", DEFAULT("*"), "The text to search for within files, supports wildcards '*' and '?'");
+	explain_optional_argument("--filter", DEFAULT("*"), "Used to filter on specific files,  supports wildcards '*' and '?'");
+	explain_optional_argument("--recursive", DEFAULT("1"), "Recursively search through directories");
+	explain_optional_argument("--max-file-size", DEFAULT("0"), "The maximum size, in kb, a file will be searched through for matching text. 0 for no limit");
+	explain_optional_argument("--threads", DEFAULT("10"), "The number of threads used for searching, minimum of 1 thread.");
+	explain_optional_argument("--locale", DEFAULT("en"), "The language errors will be reported in. Available locales are: 'en', 'nl'");
+	explain_optional_argument("--export", DEFAULT(""), "Export the results to a file in json format");
+	
+	printf("\nOther arguments:\n");
+	explain_other_argument("--help", "Display this help message");
+	explain_other_argument("--license", "Display the license");
+}
+
+static bool is_valid_argument(char *arg)
+{
+	if (string_equals(arg, "--directory")) return true;
+	if (string_equals(arg, "--text")) return true;
+	if (string_equals(arg, "--filter")) return true;
+	if (string_equals(arg, "--recursive")) return true;
+	if (string_equals(arg, "--max-file-size")) return true;
+	if (string_equals(arg, "--threads")) return true;
+	if (string_equals(arg, "--locale")) return true;
+	if (string_equals(arg, "--export")) return true;
+	
+	return false;
+}
+#endif
+void handle_command_line_arguments(int argc, char **argv)
+{
+#if 0
+	load_available_localizations();
+	set_locale("en");
+	
+	s32 current_arg_index = 1;
+	bool is_help_request = string_equals(argv[current_arg_index], "--help");
+	bool is_license_request = string_equals(argv[current_arg_index], "--license");
+	
+	if (is_help_request)
+	{
+		print_help_message();
+		return;
+	}
+	else if (is_license_request)
+	{
+		print_license_message();
+		return;
+	}
+	
+	char directory[MAX_INPUT_LENGTH];
+	string_copyn(directory, "", MAX_INPUT_LENGTH);
+	
+	char text[MAX_INPUT_LENGTH];
+	string_copyn(text, "*", MAX_INPUT_LENGTH);
+	
+	char filter[MAX_INPUT_LENGTH];
+	string_copyn(filter, "*", MAX_INPUT_LENGTH);
+	
+	bool recursive = true;
+	s32 max_file_size = 0;
+	s32 threads = 10;
+	
+	char locale[MAX_INPUT_LENGTH];
+	string_copyn(locale, "en", MAX_INPUT_LENGTH);
+	
+	char export_path[MAX_INPUT_LENGTH];
+	string_copyn(export_path, "", MAX_INPUT_LENGTH);
+	
+	bool expect_argument_name = true;
+	for (s32 i = current_arg_index; i < argc; i++)
+	{
+		if (expect_argument_name && !is_valid_argument(argv[i]))
+		{
+			printf("%s: %s\n", localize("invalid_argument"), argv[i]);
+		}
+		
+		if (!expect_argument_name)
+		{
+			if (string_equals(argv[i-1], "--directory"))
+			{
+				string_copyn(directory, argv[i], MAX_INPUT_LENGTH);
+			}
+			if (string_equals(argv[i-1], "--text"))
+			{
+				string_copyn(text, argv[i], MAX_INPUT_LENGTH);
+			}
+			if (string_equals(argv[i-1], "--filter"))
+			{
+				string_copyn(filter, argv[i], MAX_INPUT_LENGTH);
+			}
+			if (string_equals(argv[i-1], "--recursive"))
+			{
+				recursive = string_to_u32(argv[i]);
+			}
+			if (string_equals(argv[i-1], "--max-file-size"))
+			{
+				max_file_size = string_to_u32(argv[i]);
+			}
+			if (string_equals(argv[i-1], "--threads"))
+			{
+				threads = string_to_u32(argv[i]);
+			}
+			if (string_equals(argv[i-1], "--locale"))
+			{
+				string_copyn(locale, argv[i], MAX_INPUT_LENGTH);
+			}
+			if (string_equals(argv[i-1], "--export"))
+			{
+				string_copyn(export_path, argv[i], MAX_INPUT_LENGTH);
+			}
+		}
+		
+		expect_argument_name = !expect_argument_name;
+	}
+	
+	// input validation
+	if (!set_locale(locale))
+	{
+		printf(localize("warning_locale_not_available"), locale);
+		printf("\n");
+	}
+	
+	if (string_equals(directory, ""))
+	{
+		printf("%s", localize("error_directory_not_specified"));
+		printf("\n");
+		return;
+	}
+	
+	if (!platform_directory_exists(directory))
+	{
+		printf(localize("error_directory_not_found"), directory);
+		printf("\n");
+		return;
+	}
+	
+	if (string_equals(text, ""))
+	{
+		printf("%s", localize("error_text_argument_empty"));
+		printf("\n");
+		return;
+	}
+	
+	if (string_equals(filter, ""))
+	{
+		printf("%s", localize("error_filter_argument_empty"));
+		printf("\n");
+		return;
+	}
+	
+	if (threads < 1)
+	{
+		printf("%s", localize("error_threads_too_low"));
+		printf("\n");
+		return;
+	}
+	
+	if (!string_equals(export_path, ""))
+	{
+		char dir_buffer[MAX_INPUT_LENGTH];
+		get_directory_from_path(dir_buffer, export_path);
+		
+		if (!platform_directory_exists(dir_buffer))
+		{
+			printf(localize("error_invalid_export_path"), dir_buffer);
+			printf("\n");
+			return;
+		}
+	}
+	
+	search_result *result = create_empty_search_result();
+	string_copyn(result->directory_to_search, directory, MAX_INPUT_LENGTH);
+	string_copyn(result->file_filter, filter, MAX_INPUT_LENGTH);
+	string_copyn(result->text_to_find, text, MAX_INPUT_LENGTH);
+	string_copyn(result->export_path, export_path, MAX_INPUT_LENGTH);
+	result->max_thread_count = threads;
+	result->max_file_size = max_file_size;
+	result->is_recursive = recursive;
+	result->is_command_line_search = true;
+	
+	
+	// begin search (code below is equal to code in text_search.c)
+	result->walking_file_system = true;
+	result->done_finding_matches = false;
+	
+	result->search_result_source_dir_len = strlen(result->directory_to_search);
+	result->search_result_source_dir_len = prepare_search_directory_path(result->directory_to_search, 
+																		 result->search_result_source_dir_len);
+	result->start_time = platform_get_time(TIME_FULL, TIME_US);
+	
+	platform_list_files(&result->files, result->directory_to_search, result->file_filter, result->is_recursive, &result->mem_bucket,
+						&result->cancel_search,
+						&result->done_finding_files);
+	find_text_in_files(result);
+	
+	while(!result->threads_closed) { thread_sleep(1000); }
+	
+	if (!string_equals(export_path, ""))
+	{
+		export_results(result);
+	}
+#endif
+}
\ No newline at end of file
diff --git a/src/command_line.h b/src/command_line.h
new file mode 100644
index 0000000..74daa8d
--- /dev/null
+++ b/src/command_line.h
@@ -0,0 +1,12 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_COMMAND_LINE
+#define INCLUDE_COMMAND_LINE
+
+void handle_command_line_arguments(int argc, char **argv);
+
+#endif
\ No newline at end of file
diff --git a/src/config.h b/src/config.h
new file mode 100644
index 0000000..a1a7611
--- /dev/null
+++ b/src/config.h
@@ -0,0 +1,18 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_CONFIG
+#define INCLUDE_CONFIG
+
+#define TARGET_FRAMERATE 1000/24.0
+
+#define SCROLL_SPEED 50
+#define FILE_RESERVE_COUNT 500
+#define ERROR_RESERVE_COUNT 10
+#define MAX_ERROR_MESSAGE_LENGTH 120
+#define MAX_STATUS_TEXT_LENGTH 120
+
+#endif
\ No newline at end of file
diff --git a/src/external/LooplessSizeMove.c b/src/external/LooplessSizeMove.c
new file mode 100644
index 0000000..1843cb9
--- /dev/null
+++ b/src/external/LooplessSizeMove.c
@@ -0,0 +1,864 @@
+/*
+    LooplessSizeMove.c
+    Implements functions for modal-less window resizing and movement in Windows
+	
+    Author: Nathaniel J Fries
+	
+    The author asserts no copyright, this work is released into the public domain.
+*/
+
+
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+
+#ifdef TIME_LOOP
+#include <stdio.h>
+#endif /* TIME_LOOP */
+
+/*	fills the MINMAXINFO structure pointed to by the second argument with
+ default MIMAX values, then sends WM_GETMINMAXINFO to allow the
+ user's Window Procedure to modify it.
+*/
+void GetMinMaxInfo(HWND hwnd, PMINMAXINFO info);
+/* 	begins the loopless resize/move process */
+LRESULT PrepareSizeMove(HWND hwnd, WPARAM action, DWORD dwPos);
+/* 	stops the loopless resize/move process.
+ if cancel is TRUE, restores window size and position to
+ what they were before resizing/moving.
+*/
+void StopSizing(BOOL cancel);
+/* 	see LooplessSizeMove.h */
+LRESULT CALLBACK LSMProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam);
+BOOLEAN SizingCheck(const MSG *lpmsg);
+
+#define GetWindowLongAW(hwnd, lp)\
+(IsWindowUnicode(hwnd)) ? \
+(GetWindowLongW(hwnd, lp)) : \
+(GetWindowLongA(hwnd, lp))
+#define SendMessageAW(hwnd, msg, wParam, lParam)\
+(IsWindowUnicode(hwnd)) ? \
+(SendMessageW(hwnd, msg, wParam, lParam)) : \
+(SendMessageA(hwnd, msg, wParam, lParam))
+#define PostMessageAW(hwnd, msg, wParam, lParam)\
+(IsWindowUnicode(hwnd)) ? \
+(PostMessageW(hwnd, msg, wParam, lParam)) : \
+(PostMessageA(hwnd, msg, wParam, lParam))
+#define DefWindowProcAW(hwnd, msg, wParam, lParam)\
+(IsWindowUnicode(hwnd)) ? \
+(DefWindowProcW(hwnd, msg, wParam, lParam)) : \
+(DefWindowProcA(hwnd, msg, wParam, lParam))
+
+#define RECTWIDTH(r)	((r).right - (r).left)
+#define RECTHEIGHT(r)	((r).bottom - (r).top)
+
+#define LSM_LEFT 	0x01
+#define LSM_TOP 	0x02
+#define LSM_RIGHT   0x04
+#define LSM_BOTTOM  0x08
+#define LSM_CAPTION 0x00
+#define LSM_NOGRAB  0xF0
+
+#define LSM_SHAKE_MINTIME 20 /* minimum time between bothering */
+#define LSM_SHAKE_MAXTIME 1000 /* maximum time between movements for "shake" effect */
+#define LSM_SHAKE_STATE_MAXIMIZED 1
+#define LSM_SNAP_HELPER_CLASS "LSM_SNAP_HELPER"
+/*
+    information required to reverse a "shake" action.
+*/
+typedef struct _SHAKERESTORENODE
+{
+    HWND hwnd;
+    WINDOWPLACEMENT place;
+    struct _SHAKERESTORENODE *next;
+} SHAKERESTORENODE, *PSHAKERESTORENODE;
+typedef struct
+{
+    DWORD dwPrevTime;
+    WORD wDir;
+    WORD wCount;
+    SHAKERESTORENODE *restoreList;
+} SHAKEDATA;
+typedef BOOL (*SHAKEFOREACHFN)(PSHAKERESTORENODE, LPARAM);
+
+typedef struct
+{
+    HWND helperWindow;
+    WORD isSnapped;
+    WORD snapType; /* LSM_TOP, LSM_LEFT, LSM_RIGHT */
+    RECT rcWork;
+    RECT rcRestore;
+} SNAPDATA;
+
+/* holds all data that needs to be held on to for resizing */
+typedef struct
+{
+    HWND hwnd;
+    MINMAXINFO minmax;
+    RECT rcWin;
+    RECT rcOrig;
+    POINT ptCapture;
+    LONG grab;
+    SHAKEDATA shake;
+    SNAPDATA snap;
+} SIZEMOVEDATA;
+
+DWORD dwSizeMoveTlsIndex = 0;
+#define LSMTlsCheck() if(dwSizeMoveTlsIndex == 0){ dwSizeMoveTlsIndex = TlsAlloc(); }
+#define LSMSet(data) TlsSetValue(dwSizeMoveTlsIndex, data)
+#define LSMGet() TlsGetValue(dwSizeMoveTlsIndex)
+
+void GetMinMaxInfo(HWND hwnd, PMINMAXINFO info)
+{
+    RECT rc;
+    LONG style = GetWindowLongAW(hwnd, GWL_STYLE);
+    LONG altStyle = ((style & WS_CAPTION) == WS_CAPTION)?
+        (style & ~WS_BORDER):(style);
+	
+    /* calculate the default values in case WindowProc does not respond */
+    GetClientRect(GetParent(hwnd), &rc);
+    AdjustWindowRectEx(&rc, altStyle, ((style & WS_POPUP) && GetMenu(hwnd)),
+					   GetWindowLongAW(hwnd, GWL_EXSTYLE));
+    info->ptMaxPosition.x = rc.left;
+    info->ptMaxPosition.y = rc.top;
+    info->ptMaxSize.x = rc.right - rc.left;
+    info->ptMaxSize.y = rc.bottom - rc.top;
+    if(style & WS_CAPTION)
+    {
+        info->ptMinTrackSize.x = GetSystemMetrics(SM_CXMINTRACK);
+        info->ptMaxTrackSize.y = GetSystemMetrics(SM_CYMINTRACK);
+    }
+    else
+    {
+        /* why not zero? this is what ReactOS and presumably Wine do,
+        and they're the experts at replicating Windows UI behavior */
+        info->ptMinTrackSize.x = info->ptMaxPosition.x * -2;
+        info->ptMaxTrackSize.y = info->ptMaxPosition.y * -2;
+    }
+    info->ptMaxTrackSize.x = GetSystemMetrics(SM_CXMAXTRACK);
+    info->ptMaxTrackSize.y = GetSystemMetrics(SM_CYMAXTRACK);
+	
+    /* ask Window proc to make any changes */
+    SendMessageAW(hwnd, WM_GETMINMAXINFO, 0, (LPARAM)info);
+}
+
+BOOL ForEachShakeNodeFree(PSHAKERESTORENODE node, LPARAM ignore)
+{
+    LocalFree(node);
+    return TRUE;
+}
+BOOL ForEachShakeNode(SHAKEDATA *pShake, SHAKEFOREACHFN fn, LPARAM lParam)
+{
+    SHAKERESTORENODE *next, *curr;
+    BOOL res = TRUE;
+    if(pShake->restoreList)
+    {
+        curr = pShake->restoreList;
+        while(res && curr)
+        {
+            next = curr->next;
+            fn(curr, lParam);
+            curr = next;
+        }
+    }
+    return res;
+}
+
+void SnapCleanup(SIZEMOVEDATA *sizemove)
+{
+    ShowWindow(sizemove->snap.helperWindow, SW_HIDE);
+    ZeroMemory(&sizemove->snap.rcWork, sizeof(RECT));
+}
+
+LRESULT WINAPI SnapHelperWinProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam)
+{
+    switch(msg)
+    {
+		case WM_PAINT:
+        {
+            SIZEMOVEDATA *sizemove = LSMGet();
+            HDC hDC = GetDC(hwnd);
+            HPEN hPen = CreatePen(PS_INSIDEFRAME, 1, GetSysColor(COLOR_HIGHLIGHT));
+            LOGBRUSH blog = { BS_HOLLOW, 0, 0 };
+            HBRUSH hBrush = CreateBrushIndirect(&blog);
+            if(!sizemove || !hDC || !hPen || !hBrush)
+                break;
+			
+            SelectObject(hDC, hPen);
+            SelectObject(hDC, hBrush);
+			
+            Rectangle(hDC, 0, 0, RECTWIDTH(sizemove->snap.rcWork), RECTHEIGHT(sizemove->snap.rcWork));
+			
+            DeleteObject(hBrush);
+            DeleteObject(hPen);
+            break;
+        }
+		default:
+        {
+            break;
+        }
+    }
+    return DefWindowProcA(hwnd, msg, wParam, lParam);
+}
+
+LRESULT PrepareSizeMove(HWND hwnd, WPARAM action, DWORD dwPos)
+{
+    WINDOWINFO winfo;
+    SIZEMOVEDATA *sm;
+    RECT rcClipCursor;
+	
+    winfo.cbSize = sizeof(WINDOWINFO);
+    /* most likely not a valid window */
+    if(GetWindowInfo(hwnd, &winfo) == FALSE)
+        return 0;
+    /* can't move or resize an invisible window */
+    if(!IsWindowVisible(hwnd))
+        return 0;
+    /* can't resize a window without the resizing border */
+    if((action & 0xfff0) == SC_MOVE && !(winfo.dwStyle & WS_SIZEBOX))
+        return 0;
+	
+    /*
+      if another window on this thread has capture,
+      it might be using this too...
+      tell it to clean up before setting the tls value
+    */
+    ReleaseCapture();
+	
+    if(!(sm = LSMGet()))
+    {
+        WNDCLASSA wndcls;
+        HINSTANCE hInstance = GetModuleHandleA(NULL);
+        sm = LocalAlloc(0, sizeof(SIZEMOVEDATA));
+        if(!sm)
+        {
+            /* error */
+            return 1;
+        }
+        if(!LSMSet(sm))
+        {
+            LocalFree(sm);
+            return 2;
+        }
+        /* prevent potential crashes and bad initialization bugs */
+        ZeroMemory(sm, sizeof(SIZEMOVEDATA));
+		
+        if(!GetClassInfoA(hInstance, LSM_SNAP_HELPER_CLASS, &wndcls))
+        {
+            ZeroMemory(&wndcls, sizeof(WNDCLASSA));
+            wndcls.style = CS_SAVEBITS;
+            wndcls.hbrBackground = GetSysColorBrush(COLOR_HOTLIGHT);
+            wndcls.lpszClassName = LSM_SNAP_HELPER_CLASS;
+            wndcls.lpfnWndProc = SnapHelperWinProc;
+            wndcls.hInstance = hInstance;
+            RegisterClassA(&wndcls);
+        }
+        /*  WS_EX_LAYERED: window is not fully opaque
+            WS_EX_TRANSPARENT: mouse events pass through layered window
+            ES_EX_NOACTIVATE: window cannot be activated by user (no loss of mouse capture or keyboard focus)
+        */
+        sm->snap.helperWindow = CreateWindowExA(WS_EX_TRANSPARENT | WS_EX_LAYERED | WS_EX_NOACTIVATE,
+												LSM_SNAP_HELPER_CLASS, "", WS_POPUP /* no borders */, 0, 0, 100, 100, NULL, NULL, hInstance, NULL);
+        if(sm->snap.helperWindow)
+        {
+            SetLayeredWindowAttributes(sm->snap.helperWindow, 0, 75, LWA_ALPHA);
+        }
+    }
+    else if(sm->hwnd != hwnd)
+    {
+        /* forget shake data, consistent with Windows 7 behavior */
+        ForEachShakeNode(&sm->shake, ForEachShakeNodeFree, 0);
+        SnapCleanup(sm);
+        sm->shake.restoreList = NULL;
+    }
+    sm->grab = action & 0x000f;
+    sm->hwnd = hwnd;
+    GetMinMaxInfo(hwnd, &sm->minmax);
+    sm->rcWin = winfo.rcWindow;
+    if(winfo.dwStyle & WS_CHILD)
+    {
+        /* map points into the parent's coordinate space */
+        HWND parent = GetParent(hwnd);
+        MapWindowPoints(0, parent, (LPPOINT)&sm->rcWin, 2);
+        GetWindowRect(parent, &rcClipCursor);
+        MapWindowPoints(parent, HWND_DESKTOP, (LPPOINT)&rcClipCursor, 2);
+    }
+    else if(!(winfo.dwExStyle & WS_EX_TOPMOST))
+    {
+        SystemParametersInfoW(SPI_GETWORKAREA, 0, &rcClipCursor, 0);
+    }
+    else
+    {
+        rcClipCursor.left = rcClipCursor.top = 0;
+        rcClipCursor.right = GetSystemMetrics(SM_CXSCREEN);
+        rcClipCursor.bottom = GetSystemMetrics(SM_CYSCREEN);
+    }
+    sm->rcOrig = sm->rcWin;
+	
+    sm->ptCapture.x = (short)LOWORD(dwPos);
+    sm->ptCapture.y = (short)HIWORD(dwPos);
+    //ClipCursor(&rcClipCursor);
+	
+	
+    /* notify WinProc we're beginning, but return instead of looping */
+    SendMessageAW(hwnd, WM_ENTERSIZEMOVE, 0, 0);
+    if(GetCapture() != hwnd)
+    {
+        SetCapture(hwnd);
+    }
+    return 0;
+}
+
+LRESULT CALLBACK LSMProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam)
+{
+    LSMTlsCheck();
+    switch(msg)
+    {
+        case WM_NCLBUTTONDOWN:
+        {
+            switch(wParam)
+            {
+                case HTLEFT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_LEFT, lParam);
+                case HTTOPLEFT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_LEFT | LSM_TOP, lParam);
+                case HTBOTTOMLEFT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_LEFT | LSM_BOTTOM, lParam);
+                case HTRIGHT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_RIGHT, lParam);
+                case HTTOPRIGHT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_RIGHT | LSM_TOP, lParam);
+                case HTBOTTOMRIGHT:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_RIGHT | LSM_BOTTOM, lParam);
+                case HTTOP:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_TOP, lParam);
+                case HTBOTTOM:
+				return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_SIZE | LSM_BOTTOM, lParam);
+                case HTCAPTION:
+				//return SendMessageAW(hwnd, WM_SYSCOMMAND, SC_MOVE | LSM_CAPTION, lParam);
+                default:
+				break;
+            }
+        }
+        case WM_SYSCOMMAND:
+        {
+            switch(wParam & 0xfff0)
+            {
+                case SC_SIZE:
+				/* begin resize 'loop' */
+				return PrepareSizeMove(hwnd, wParam, lParam);
+                default:
+				break;
+            }
+            break;
+        }
+        case WM_CAPTURECHANGED:
+        {
+            /* nothing we can do; stop resizing & do clean-up */
+            SIZEMOVEDATA *sizemove = LSMGet();
+            if(sizemove && (HWND)lParam != sizemove->hwnd)
+            {
+                SendMessageAW(sizemove->hwnd, WM_EXITSIZEMOVE, 0, 0);
+                ClipCursor(NULL);
+                /* no longer unset LSM data:
+                    1) never really needed to
+                    2) necessary to hold onto shake data
+                */
+                if(GetForegroundWindow() != sizemove->hwnd)
+                {
+                    /* forget shake data, consistent with Windows 7 behavior */
+                    ForEachShakeNode(&sizemove->shake, ForEachShakeNodeFree, 0);
+                    sizemove->shake.restoreList = NULL;
+                }
+                sizemove->shake.wCount = 0;
+                sizemove->shake.wDir = 0;
+                sizemove->shake.dwPrevTime = 0;
+                SnapCleanup(sizemove);
+				
+                sizemove->grab = LSM_NOGRAB;
+            }
+        }
+        default:
+		break;
+    }
+    return DefWindowProcAW(hwnd, msg, wParam, lParam);
+}
+
+BOOL CALLBACK EnumWindowsShakeMinimize(HWND hwnd, LPARAM lParam)
+{
+    SIZEMOVEDATA *sizemove = (SIZEMOVEDATA *)(lParam);
+    if(hwnd != sizemove->hwnd && IsWindowVisible(hwnd) && !IsIconic(hwnd))
+    {
+        SHAKERESTORENODE *pNode = 0;
+        WINDOWPLACEMENT tPlace;
+        /* note: there are a few shell-created windows that glitch up graphically if you try to minimize them */
+        DWORD dwProcID, dwShellPID;
+        GetWindowThreadProcessId(hwnd, &dwProcID);
+        GetWindowThreadProcessId(GetShellWindow(), &dwShellPID);
+        if(dwProcID == dwShellPID)
+        {
+            return TRUE;
+        }
+        /* note: not supposed to minimize "parent application" windows. I assume this means windows of
+            the window's application?
+            See: http://social.technet.microsoft.com/Forums/windows/en-US/b047378c-2a5a-4661-a871-cec459cb9bdc/aeroshake-not-working?forum=w7itproui
+        */
+        GetWindowThreadProcessId(sizemove->hwnd, &dwShellPID);
+        if(dwProcID == dwShellPID)
+        {
+            return TRUE;
+        }
+		
+		
+        pNode = LocalAlloc(0, sizeof(SHAKERESTORENODE));
+        if(!pNode)
+            return FALSE;
+		
+        pNode->hwnd = hwnd;
+		
+        pNode->place.length = sizeof(WINDOWPLACEMENT);
+        pNode->place.flags = WPF_ASYNCWINDOWPLACEMENT;
+        GetWindowPlacement(hwnd, &pNode->place);
+        tPlace = pNode->place;
+        tPlace.showCmd = SW_SHOWMINNOACTIVE;
+        SetWindowPlacement(hwnd, &tPlace);
+		
+        pNode->next = sizemove->shake.restoreList;
+        sizemove->shake.restoreList = pNode;
+    }
+    return TRUE;
+}
+
+BOOL ForEachShakeNodeRestore(PSHAKERESTORENODE node, LPARAM ignore)
+{
+    if(node)
+    {
+        SetWindowPlacement(node->hwnd, &node->place);
+        LocalFree(node);
+        return TRUE;
+    }
+    return FALSE;
+}
+
+void ShakeCheck(const MSG *lpmsg, SIZEMOVEDATA *sizemove)
+{
+    static DWORD dwShakeDisabled = 2;
+    WORD dir = 0;
+    int dx, dy;
+    if(lpmsg->message != WM_MOUSEMOVE && lpmsg->message != WM_NCMOUSEMOVE)
+        return;
+    if(dwShakeDisabled == 2)
+    {
+        HKEY hKey;
+        DWORD dwSizeDW = 4;
+        DWORD dwType = REG_DWORD;
+        if(/* major version number */LOBYTE(LOWORD(GetVersion())) < 6)
+        {
+            dwShakeDisabled = 1;
+        }
+        else
+        {
+            dwShakeDisabled = 0;
+            if(RegOpenKeyExA(HKEY_CURRENT_USER, "Software\\Policies\\Microsoft\\Windows\\Explorer", 0, KEY_READ, &hKey)
+			   == ERROR_SUCCESS)
+            {
+                RegQueryValueExA(hKey, "NoWindowMinimizingShortcuts", 0, &dwType, (LPBYTE)&dwShakeDisabled, &dwSizeDW);
+                RegCloseKey(hKey);
+            }
+        }
+    }
+    if(dwShakeDisabled)
+        return;
+    if(lpmsg->time - sizemove->shake.dwPrevTime < LSM_SHAKE_MINTIME)
+        return;
+	
+    dx = lpmsg->pt.x - sizemove->ptCapture.x;
+    dy = lpmsg->pt.y - sizemove->ptCapture.y;
+    if(dx > 8)
+        dir |= LSM_LEFT;
+    else if(dx < -8)
+        dir |= LSM_RIGHT;
+    if(dy > 8)
+        dir |= LSM_TOP;
+    else if(dy < -8)
+        dir |= LSM_BOTTOM;
+	
+    if(dir && ((lpmsg->time - sizemove->shake.dwPrevTime) < LSM_SHAKE_MAXTIME) && (dir != sizemove->shake.wDir))
+    {
+        /* Do Shake */
+        sizemove->shake.wCount++;
+        if(sizemove->shake.wCount >= 3)
+        {
+            if(sizemove->shake.restoreList)
+            {
+                ForEachShakeNode(&sizemove->shake, ForEachShakeNodeRestore, (LPARAM)sizemove);
+                sizemove->shake.restoreList = 0;
+            }
+            else
+            {
+                EnumWindows(EnumWindowsShakeMinimize, (LPARAM)(sizemove));
+            }
+            RedrawWindow(GetDesktopWindow(), NULL, NULL, RDW_ERASE | RDW_FRAME | RDW_INTERNALPAINT | RDW_INVALIDATE
+						 | RDW_ALLCHILDREN | RDW_UPDATENOW);
+            sizemove->shake.wCount = 0;
+        }
+    }
+    else
+    {
+        sizemove->shake.wCount = 0;
+    }
+    sizemove->shake.dwPrevTime = lpmsg->time;
+    sizemove->shake.wDir = dir;
+}
+
+void SnapCheck(POINT pt, SIZEMOVEDATA *sizemove)
+{
+    static DWORD dwSnapActive = 2;
+    HMONITOR monitor = MonitorFromPoint(pt, MONITOR_DEFAULTTONEAREST);
+    MONITORINFO minfo;
+	
+    if(dwSnapActive == 2)
+    {
+        HKEY hKey;
+        DWORD dwSizeDW = 4;
+        DWORD dwType = REG_DWORD;
+        dwSnapActive = 0;
+        if(/* major version number */LOBYTE(LOWORD(GetVersion())) >= 6)
+        {
+            if(RegOpenKeyExA(HKEY_CURRENT_USER, "Control Panel\\Desktop", 0, KEY_READ, &hKey)
+			   == ERROR_SUCCESS)
+            {
+                RegQueryValueExA(hKey, "WindowArrangementActive", 0, &dwType, (LPBYTE)&dwSnapActive, &dwSizeDW);
+                RegCloseKey(hKey);
+            }
+        }
+    }
+    if(!dwSnapActive)
+        return;
+	
+    minfo.cbSize = sizeof(MONITORINFO);
+    if(monitor == NULL)
+        return;
+    if(!GetMonitorInfo(monitor, (LPMONITORINFO)&minfo))
+        return;
+    /* unsnap maximized windows */
+    if(IsZoomed(sizemove->hwnd))
+    {
+        WINDOWPLACEMENT place;
+        int dx, w;
+        place.length = sizeof(WINDOWPLACEMENT);
+        GetWindowPlacement(sizemove->hwnd, &place);
+        place.showCmd = SW_RESTORE;
+        OffsetRect(&place.rcNormalPosition, (sizemove->rcWin.left - place.rcNormalPosition.left),
+                   (sizemove->rcWin.top - place.rcNormalPosition.top));
+        w = RECTWIDTH(place.rcNormalPosition);
+        if((dx = (place.rcNormalPosition.left - sizemove->ptCapture.x)) > 0)
+        {
+            place.rcNormalPosition.left += dx << 1;
+            place.rcNormalPosition.right = place.rcNormalPosition.left + w;
+        }
+        if((dx = (sizemove->ptCapture.x - place.rcNormalPosition.right)) > 0)
+        {
+            place.rcNormalPosition.left += dx << 1;
+            place.rcNormalPosition.right = place.rcNormalPosition.left + w;
+        }
+        sizemove->rcWin = place.rcNormalPosition;
+        SetWindowPlacement(sizemove->hwnd, &place);
+        /* don't return here, it's possible to unsnap without leaving the top of the screen */
+    }
+    else if(sizemove->snap.isSnapped)
+    {
+        /* TODO: Aero snap lets you drag half-screen snapped windows across the top of the screen */
+		
+        sizemove->rcWin.right = sizemove->snap.rcRestore.right;
+        sizemove->rcWin.bottom = sizemove->snap.rcRestore.bottom;
+        sizemove->snap.isSnapped = 0;
+        return;
+    }
+    sizemove->snap.rcWork = minfo.rcWork;
+    if(pt.x <= minfo.rcWork.left)
+    {
+        /* left side stretch */
+        sizemove->snap.rcWork.right = (RECTWIDTH(minfo.rcWork) - GetSystemMetrics(SM_CXFRAME)) >> 1;
+        sizemove->snap.snapType = LSM_LEFT;
+    }
+    else if(pt.x+1 >= minfo.rcWork.right)
+    {
+        /* right side stretch */
+        sizemove->snap.rcWork.left = (minfo.rcWork.left + RECTWIDTH(minfo.rcWork) + GetSystemMetrics(SM_CXFRAME)) >> 1;
+        sizemove->snap.snapType = LSM_RIGHT;
+    }
+    else if(pt.y <= minfo.rcWork.top)
+    {
+        /* full stretch */
+        sizemove->snap.snapType = LSM_TOP;
+    }
+    else
+    {
+        /* not snapping, clean up snap state */
+        SnapCleanup(sizemove);
+        return;
+    }
+    SetWindowPos(sizemove->snap.helperWindow, HWND_TOPMOST,
+                 sizemove->snap.rcWork.left, sizemove->snap.rcWork.top,
+                 RECTWIDTH(sizemove->snap.rcWork), RECTHEIGHT(sizemove->snap.rcWork),
+                 SWP_NOACTIVATE);
+    ShowWindow(sizemove->snap.helperWindow, SW_SHOWNA);
+}
+
+void SnapFinalize(SIZEMOVEDATA *sizemove)
+{
+    if(sizemove->snap.rcWork.left != sizemove->snap.rcWork.right
+       && sizemove->snap.rcWork.top != sizemove->snap.rcWork.bottom)
+    {
+        if(sizemove->snap.snapType == LSM_TOP)
+        {
+            /* Aero appears to animate this... so we will too */
+            ShowWindow(sizemove->hwnd, SW_MAXIMIZE);
+        }
+        else
+        {
+            SetWindowPos(sizemove->hwnd, 0, sizemove->snap.rcWork.left, sizemove->snap.rcWork.top,
+						 RECTWIDTH(sizemove->snap.rcWork), RECTHEIGHT(sizemove->snap.rcWork), 0);
+            ZeroMemory(&sizemove->snap.rcRestore, sizeof(RECT));
+            sizemove->snap.rcRestore.right = RECTWIDTH(sizemove->rcWin);
+            sizemove->snap.rcRestore.bottom = RECTWIDTH(sizemove->rcWin);
+            sizemove->snap.isSnapped = 1;
+        }
+    }
+    SnapCleanup(sizemove);
+}
+
+BOOLEAN SizingCheck(const MSG *lpmsg)
+{
+    SIZEMOVEDATA *sizemove = LSMGet();
+    POINT pt = lpmsg->pt;
+    int dx = 0, dy = 0;
+	/*
+        Discussion of rev3 changes.
+        There was a bug in previous revisions that would cause
+            the resize state to continue even if the Window lost
+            mouse capture. Windows provides notification of losing
+            mouse capture, but it crashes the program to take capture
+            back while processing that message.
+            So, we choose to yield to this other program and stop resizing.
+            This is probably user32 behavior anyway.
+            Windows also sends this notification in response to calling ReleaseCapture,
+                so all clean-up code has been moved to the handler.
+            This also allowed us to eliminate the function StopSizing.
+ */
+    if(!sizemove) /* not sizing */
+        return 0;
+    if(sizemove->grab == LSM_NOGRAB) /* not sizing */
+        return 0;
+    if(lpmsg->hwnd != sizemove->hwnd) /* wrong window */
+        return 0;
+    if(lpmsg->message == WM_NCLBUTTONUP || lpmsg->message == WM_LBUTTONUP)
+    {
+        SnapFinalize(sizemove);
+        ReleaseCapture();
+        return 1;
+    }
+    if(lpmsg->message == WM_KEYDOWN)
+    {
+        switch(lpmsg->wParam)
+        {
+            case VK_RETURN:
+			ReleaseCapture();
+			return 1;
+            case VK_ESCAPE:
+            {
+                SetWindowPos(sizemove->hwnd, 0, sizemove->rcOrig.left, sizemove->rcOrig.top,
+							 RECTWIDTH(sizemove->rcOrig), RECTHEIGHT(sizemove->rcOrig), 0);
+                ReleaseCapture();
+                return 1;
+            }
+            case VK_UP:
+			pt.y-=8;
+			break;
+            case VK_DOWN:
+			pt.y+=8;
+			break;
+            case VK_LEFT:
+			pt.x-=8;
+			break;
+            case VK_RIGHT:
+			pt.x+=8;
+			break;
+            default:
+			break;
+        }
+    }
+	
+    /* used to handle WM_MOUSEMOVE. This was unnecessary code  */
+	
+    dx = pt.x - sizemove->ptCapture.x;
+    dy = pt.y - sizemove->ptCapture.y;
+    if(dx || dy)
+    {
+        BOOL changeCursor = (lpmsg->message == WM_KEYDOWN);
+        WPARAM wpHit = 0;
+#ifdef TIME_LOOP
+        LARGE_INTEGER pfBegin;
+        LARGE_INTEGER pfDraw;
+        LARGE_INTEGER pfFinal;
+        LARGE_INTEGER pfFreq;
+        QueryPerformanceFrequency(&pfFreq);
+        QueryPerformanceCounter(&pfBegin);
+#endif
+		
+        if(sizemove->grab == LSM_CAPTION)
+        {
+            ShakeCheck(lpmsg, sizemove);
+            SnapCheck(pt, sizemove);
+            OffsetRect(&sizemove->rcWin, dx, dy);
+        }
+		else
+        {
+            /* note on minmax correction
+                if you do not correct the capture pos (set later from `pt`),
+                window will expand massively if user pulls back mouse
+                after failing to shrink when resizing from the
+                bottom or the right borders.
+            */
+            /* when resizing using keys, Windows also moves the cursor */
+            if(sizemove->grab & LSM_LEFT)
+            {
+                int lmax = sizemove->rcWin.right - sizemove->minmax.ptMaxTrackSize.x;
+                int lmin = sizemove->rcWin.right - sizemove->minmax.ptMinTrackSize.x;
+                if(sizemove->rcWin.left + dx < lmax)
+                {
+                    sizemove->rcWin.left = lmax;
+                }
+                else if(sizemove->rcWin.left + dx > lmin)
+                {
+                    sizemove->rcWin.left = lmin;
+                }
+                else
+                {
+                    sizemove->rcWin.left += dx;
+                }
+                pt.x = sizemove->rcWin.left;
+                wpHit = WMSZ_LEFT;
+            }
+            else if(sizemove->grab & LSM_RIGHT)
+            {
+                int rmax = sizemove->rcWin.left + sizemove->minmax.ptMaxTrackSize.x;
+                int rmin = sizemove->rcWin.left + sizemove->minmax.ptMinTrackSize.x;
+                if(sizemove->rcWin.right + dx > rmax)
+                {
+                    sizemove->rcWin.right = rmax;
+                }
+                else if(sizemove->rcWin.right + dx < rmin)
+                {
+                    sizemove->rcWin.right = rmin;
+                }
+                else
+                {
+                    sizemove->rcWin.right += dx;
+                }
+                pt.x = sizemove->rcWin.right;
+                wpHit = WMSZ_RIGHT;
+            }
+            if(sizemove->grab & LSM_TOP)
+            {
+                int tmax = sizemove->rcWin.bottom - sizemove->minmax.ptMaxTrackSize.y;
+                int tmin = sizemove->rcWin.bottom - sizemove->minmax.ptMinTrackSize.y;
+                if(sizemove->rcWin.top + dy < tmax)
+                {
+                    sizemove->rcWin.top = tmax;
+                }
+                else if(sizemove->rcWin.top + dy > tmin)
+                {
+                    sizemove->rcWin.top = tmin;
+                }
+                else
+                {
+                    sizemove->rcWin.top += dy;
+                }
+                pt.y = sizemove->rcWin.top;
+                if(wpHit == WMSZ_LEFT)
+                {
+                    wpHit = WMSZ_TOPLEFT;
+                }
+                else if(wpHit == WMSZ_RIGHT)
+                {
+                    wpHit = WMSZ_TOPRIGHT;
+                }
+                else
+                {
+                    wpHit = WMSZ_TOP;
+                }
+            }
+            else if(sizemove->grab & LSM_BOTTOM)
+            {
+                int bmax = sizemove->rcWin.top + sizemove->minmax.ptMaxTrackSize.y;
+                int bmin = sizemove->rcWin.top + sizemove->minmax.ptMinTrackSize.y;
+                if(sizemove->rcWin.bottom + dy > bmax)
+                {
+                    sizemove->rcWin.bottom = bmax;
+                }
+                else if(sizemove->rcWin.bottom + dy < bmin)
+                {
+                    sizemove->rcWin.bottom = bmin;
+                }
+                else
+                {
+                    sizemove->rcWin.bottom += dy;
+                }
+                pt.y = sizemove->rcWin.bottom;
+                if(wpHit == WMSZ_LEFT)
+                {
+                    wpHit = WMSZ_BOTTOMLEFT;
+                }
+                else if(wpHit == WMSZ_RIGHT)
+                {
+                    wpHit = WMSZ_BOTTOMRIGHT;
+                }
+                else
+                {
+                    wpHit = WMSZ_BOTTOM;
+                }
+            }
+        }
+#ifdef TIME_LOOP
+        QueryPerformanceCounter(&pfDraw);
+#endif
+        SendMessageAW(sizemove->hwnd, WM_SIZING, wpHit, (LPARAM)&sizemove->rcWin);
+        SetWindowPos(sizemove->hwnd, 0, sizemove->rcWin.left, sizemove->rcWin.top,
+					 RECTWIDTH(sizemove->rcWin), RECTHEIGHT(sizemove->rcWin), 0);
+		
+        sizemove->ptCapture = pt;
+        /* when resizing using keys, Windows also moves the cursor */
+        if(changeCursor)
+            SetCursorPos(pt.x, pt.y);
+#ifdef TIME_LOOP
+        QueryPerformanceCounter(&pfFinal);
+        if(1)
+        {
+            double msTotal, msDraw;
+            msTotal = (double)(*(long long*)(&pfFinal) - *(long long *)(&pfBegin)) / (*(long long *)(&pfFreq) / 1000);
+            msDraw = (double)(*(long long*)(&pfFinal) - *(long long *)(&pfDraw)) / (*(long long *)(&pfFreq) / 1000);
+            printf("draw time: %.03fms\ntotal:     %.03fms\n", msDraw, msTotal);
+        }
+#endif
+	}
+	return 1;
+}
+
+void LSMCleanup()
+{
+    SIZEMOVEDATA *sm = LSMGet();
+    if(sm)
+    {
+        if(sm->shake.restoreList)
+        {
+            ForEachShakeNode(&sm->shake, ForEachShakeNodeFree, 0);
+        }
+        if(sm->snap.helperWindow)
+        {
+            DestroyWindow(sm->snap.helperWindow);
+        }
+        LocalFree(sm);
+        LSMSet(NULL);
+    }
+}
diff --git a/src/external/cJSON.c b/src/external/cJSON.c
new file mode 100644
index 0000000..ded7006
--- /dev/null
+++ b/src/external/cJSON.c
@@ -0,0 +1,2980 @@
+/*
+  Copyright (c) 2009-2017 Dave Gamble and cJSON contributors
+
+  Permission is hereby granted, free of charge, to any person obtaining a copy
+  of this software and associated documentation files (the "Software"), to deal
+  in the Software without restriction, including without limitation the rights
+  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+  copies of the Software, and to permit persons to whom the Software is
+  furnished to do so, subject to the following conditions:
+
+  The above copyright notice and this permission notice shall be included in
+  all copies or substantial portions of the Software.
+
+  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+  THE SOFTWARE.
+*/
+
+/* cJSON */
+/* JSON parser in C. */
+
+/* disable warnings about old C89 functions in MSVC */
+#if !defined(_CRT_SECURE_NO_DEPRECATE) && defined(_MSC_VER)
+#define _CRT_SECURE_NO_DEPRECATE
+#endif
+
+#ifdef __GNUC__
+#pragma GCC visibility push(default)
+#endif
+#if defined(_MSC_VER)
+#pragma warning (push)
+/* disable warning about single line comments in system headers */
+#pragma warning (disable : 4001)
+#endif
+
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+#include <stdlib.h>
+#include <limits.h>
+#include <ctype.h>
+
+#ifdef ENABLE_LOCALES
+#include <locale.h>
+#endif
+
+#if defined(_MSC_VER)
+#pragma warning (pop)
+#endif
+#ifdef __GNUC__
+#pragma GCC visibility pop
+#endif
+
+#include "cJSON.h"
+
+/* define our own boolean type */
+#ifdef true
+#undef true
+#endif
+#define true ((cJSON_bool)1)
+
+#ifdef false
+#undef false
+#endif
+#define false ((cJSON_bool)0)
+
+typedef struct {
+    const unsigned char *json;
+    size_t position;
+} error;
+static error global_error = { NULL, 0 };
+
+CJSON_PUBLIC(const char *) cJSON_GetErrorPtr(void)
+{
+    return (const char*) (global_error.json + global_error.position);
+}
+
+CJSON_PUBLIC(char *) cJSON_GetStringValue(const cJSON * const item) {
+    if (!cJSON_IsString(item)) {
+        return NULL;
+    }
+
+    return item->valuestring;
+}
+
+/* This is a safeguard to prevent copy-pasters from using incompatible C and header files */
+#if (CJSON_VERSION_MAJOR != 1) || (CJSON_VERSION_MINOR != 7) || (CJSON_VERSION_PATCH != 12)
+    #error cJSON.h and cJSON.c have different versions. Make sure that both have the same.
+#endif
+
+CJSON_PUBLIC(const char*) cJSON_Version(void)
+{
+    static char version[15];
+    sprintf(version, "%i.%i.%i", CJSON_VERSION_MAJOR, CJSON_VERSION_MINOR, CJSON_VERSION_PATCH);
+
+    return version;
+}
+
+/* Case insensitive string comparison, doesn't consider two NULL pointers equal though */
+static int case_insensitive_strcmp(const unsigned char *string1, const unsigned char *string2)
+{
+    if ((string1 == NULL) || (string2 == NULL))
+    {
+        return 1;
+    }
+
+    if (string1 == string2)
+    {
+        return 0;
+    }
+
+    for(; tolower(*string1) == tolower(*string2); (void)string1++, string2++)
+    {
+        if (*string1 == '\0')
+        {
+            return 0;
+        }
+    }
+
+    return tolower(*string1) - tolower(*string2);
+}
+
+typedef struct internal_hooks
+{
+    void *(CJSON_CDECL *allocate)(size_t size);
+    void (CJSON_CDECL *deallocate)(void *pointer);
+    void *(CJSON_CDECL *reallocate)(void *pointer, size_t size);
+} internal_hooks;
+
+#if defined(_MSC_VER)
+/* work around MSVC error C2322: '...' address of dllimport '...' is not static */
+static void * CJSON_CDECL internal_malloc(size_t size)
+{
+    return malloc(size);
+}
+static void CJSON_CDECL internal_free(void *pointer)
+{
+    free(pointer);
+}
+static void * CJSON_CDECL internal_realloc(void *pointer, size_t size)
+{
+    return realloc(pointer, size);
+}
+#else
+#define internal_malloc malloc
+#define internal_free free
+#define internal_realloc realloc
+#endif
+
+/* strlen of character literals resolved at compile time */
+#define static_strlen(string_literal) (sizeof(string_literal) - sizeof(""))
+
+static internal_hooks global_hooks = { internal_malloc, internal_free, internal_realloc };
+
+static unsigned char* cJSON_strdup(const unsigned char* string, const internal_hooks * const hooks)
+{
+    size_t length = 0;
+    unsigned char *copy = NULL;
+
+    if (string == NULL)
+    {
+        return NULL;
+    }
+
+    length = strlen((const char*)string) + sizeof("");
+    copy = (unsigned char*)hooks->allocate(length);
+    if (copy == NULL)
+    {
+        return NULL;
+    }
+    memcpy(copy, string, length);
+
+    return copy;
+}
+
+CJSON_PUBLIC(void) cJSON_InitHooks(cJSON_Hooks* hooks)
+{
+    if (hooks == NULL)
+    {
+        /* Reset hooks */
+        global_hooks.allocate = malloc;
+        global_hooks.deallocate = free;
+        global_hooks.reallocate = realloc;
+        return;
+    }
+
+    global_hooks.allocate = malloc;
+    if (hooks->malloc_fn != NULL)
+    {
+        global_hooks.allocate = hooks->malloc_fn;
+    }
+
+    global_hooks.deallocate = free;
+    if (hooks->free_fn != NULL)
+    {
+        global_hooks.deallocate = hooks->free_fn;
+    }
+
+    /* use realloc only if both free and malloc are used */
+    global_hooks.reallocate = NULL;
+    if ((global_hooks.allocate == malloc) && (global_hooks.deallocate == free))
+    {
+        global_hooks.reallocate = realloc;
+    }
+}
+
+/* Internal constructor. */
+static cJSON *cJSON_New_Item(const internal_hooks * const hooks)
+{
+    cJSON* node = (cJSON*)hooks->allocate(sizeof(cJSON));
+    if (node)
+    {
+        memset(node, '\0', sizeof(cJSON));
+    }
+
+    return node;
+}
+
+/* Delete a cJSON structure. */
+CJSON_PUBLIC(void) cJSON_Delete(cJSON *item)
+{
+    cJSON *next = NULL;
+    while (item != NULL)
+    {
+        next = item->next;
+        if (!(item->type & cJSON_IsReference) && (item->child != NULL))
+        {
+            cJSON_Delete(item->child);
+        }
+        if (!(item->type & cJSON_IsReference) && (item->valuestring != NULL))
+        {
+            global_hooks.deallocate(item->valuestring);
+        }
+        if (!(item->type & cJSON_StringIsConst) && (item->string != NULL))
+        {
+            global_hooks.deallocate(item->string);
+        }
+        global_hooks.deallocate(item);
+        item = next;
+    }
+}
+
+/* get the decimal point character of the current locale */
+static unsigned char get_decimal_point(void)
+{
+#ifdef ENABLE_LOCALES
+    struct lconv *lconv = localeconv();
+    return (unsigned char) lconv->decimal_point[0];
+#else
+    return '.';
+#endif
+}
+
+typedef struct
+{
+    const unsigned char *content;
+    size_t length;
+    size_t offset;
+    size_t depth; /* How deeply nested (in arrays/objects) is the input at the current offset. */
+    internal_hooks hooks;
+} parse_buffer;
+
+/* check if the given size is left to read in a given parse buffer (starting with 1) */
+#define can_read(buffer, size) ((buffer != NULL) && (((buffer)->offset + size) <= (buffer)->length))
+/* check if the buffer can be accessed at the given index (starting with 0) */
+#define can_access_at_index(buffer, index) ((buffer != NULL) && (((buffer)->offset + index) < (buffer)->length))
+#define cannot_access_at_index(buffer, index) (!can_access_at_index(buffer, index))
+/* get a pointer to the buffer at the position */
+#define buffer_at_offset(buffer) ((buffer)->content + (buffer)->offset)
+
+/* Parse the input text to generate a number, and populate the result into item. */
+static cJSON_bool parse_number(cJSON * const item, parse_buffer * const input_buffer)
+{
+    double number = 0;
+    unsigned char *after_end = NULL;
+    unsigned char number_c_string[64];
+    unsigned char decimal_point = get_decimal_point();
+    size_t i = 0;
+
+    if ((input_buffer == NULL) || (input_buffer->content == NULL))
+    {
+        return false;
+    }
+
+    /* copy the number into a temporary buffer and replace '.' with the decimal point
+     * of the current locale (for strtod)
+     * This also takes care of '\0' not necessarily being available for marking the end of the input */
+    for (i = 0; (i < (sizeof(number_c_string) - 1)) && can_access_at_index(input_buffer, i); i++)
+    {
+        switch (buffer_at_offset(input_buffer)[i])
+        {
+            case '0':
+            case '1':
+            case '2':
+            case '3':
+            case '4':
+            case '5':
+            case '6':
+            case '7':
+            case '8':
+            case '9':
+            case '+':
+            case '-':
+            case 'e':
+            case 'E':
+                number_c_string[i] = buffer_at_offset(input_buffer)[i];
+                break;
+
+            case '.':
+                number_c_string[i] = decimal_point;
+                break;
+
+            default:
+                goto loop_end;
+        }
+    }
+loop_end:
+    number_c_string[i] = '\0';
+
+    number = strtod((const char*)number_c_string, (char**)&after_end);
+    if (number_c_string == after_end)
+    {
+        return false; /* parse_error */
+    }
+
+    item->valuedouble = number;
+
+    /* use saturation in case of overflow */
+    if (number >= INT_MAX)
+    {
+        item->valueint = INT_MAX;
+    }
+    else if (number <= (double)INT_MIN)
+    {
+        item->valueint = INT_MIN;
+    }
+    else
+    {
+        item->valueint = (int)number;
+    }
+
+    item->type = cJSON_Number;
+
+    input_buffer->offset += (size_t)(after_end - number_c_string);
+    return true;
+}
+
+/* don't ask me, but the original cJSON_SetNumberValue returns an integer or double */
+CJSON_PUBLIC(double) cJSON_SetNumberHelper(cJSON *object, double number)
+{
+    if (number >= INT_MAX)
+    {
+        object->valueint = INT_MAX;
+    }
+    else if (number <= (double)INT_MIN)
+    {
+        object->valueint = INT_MIN;
+    }
+    else
+    {
+        object->valueint = (int)number;
+    }
+
+    return object->valuedouble = number;
+}
+
+typedef struct
+{
+    unsigned char *buffer;
+    size_t length;
+    size_t offset;
+    size_t depth; /* current nesting depth (for formatted printing) */
+    cJSON_bool noalloc;
+    cJSON_bool format; /* is this print a formatted print */
+    internal_hooks hooks;
+} printbuffer;
+
+/* realloc printbuffer if necessary to have at least "needed" bytes more */
+static unsigned char* ensure(printbuffer * const p, size_t needed)
+{
+    unsigned char *newbuffer = NULL;
+    size_t newsize = 0;
+
+    if ((p == NULL) || (p->buffer == NULL))
+    {
+        return NULL;
+    }
+
+    if ((p->length > 0) && (p->offset >= p->length))
+    {
+        /* make sure that offset is valid */
+        return NULL;
+    }
+
+    if (needed > INT_MAX)
+    {
+        /* sizes bigger than INT_MAX are currently not supported */
+        return NULL;
+    }
+
+    needed += p->offset + 1;
+    if (needed <= p->length)
+    {
+        return p->buffer + p->offset;
+    }
+
+    if (p->noalloc) {
+        return NULL;
+    }
+
+    /* calculate new buffer size */
+    if (needed > (INT_MAX / 2))
+    {
+        /* overflow of int, use INT_MAX if possible */
+        if (needed <= INT_MAX)
+        {
+            newsize = INT_MAX;
+        }
+        else
+        {
+            return NULL;
+        }
+    }
+    else
+    {
+        newsize = needed * 2;
+    }
+
+    if (p->hooks.reallocate != NULL)
+    {
+        /* reallocate with realloc if available */
+        newbuffer = (unsigned char*)p->hooks.reallocate(p->buffer, newsize);
+        if (newbuffer == NULL)
+        {
+            p->hooks.deallocate(p->buffer);
+            p->length = 0;
+            p->buffer = NULL;
+
+            return NULL;
+        }
+    }
+    else
+    {
+        /* otherwise reallocate manually */
+        newbuffer = (unsigned char*)p->hooks.allocate(newsize);
+        if (!newbuffer)
+        {
+            p->hooks.deallocate(p->buffer);
+            p->length = 0;
+            p->buffer = NULL;
+
+            return NULL;
+        }
+        if (newbuffer)
+        {
+            memcpy(newbuffer, p->buffer, p->offset + 1);
+        }
+        p->hooks.deallocate(p->buffer);
+    }
+    p->length = newsize;
+    p->buffer = newbuffer;
+
+    return newbuffer + p->offset;
+}
+
+/* calculate the new length of the string in a printbuffer and update the offset */
+static void update_offset(printbuffer * const buffer)
+{
+    const unsigned char *buffer_pointer = NULL;
+    if ((buffer == NULL) || (buffer->buffer == NULL))
+    {
+        return;
+    }
+    buffer_pointer = buffer->buffer + buffer->offset;
+
+    buffer->offset += strlen((const char*)buffer_pointer);
+}
+
+/* securely comparison of floating-point variables */
+static cJSON_bool compare_double(double a, double b)
+{
+    return (fabs(a - b) <= CJSON_DOUBLE_PRECISION);
+}
+
+/* Render the number nicely from the given item into a string. */
+static cJSON_bool print_number(const cJSON * const item, printbuffer * const output_buffer)
+{
+    unsigned char *output_pointer = NULL;
+    double d = item->valuedouble;
+    int length = 0;
+    size_t i = 0;
+    unsigned char number_buffer[26] = {0}; /* temporary buffer to print the number into */
+    unsigned char decimal_point = get_decimal_point();
+    double test = 0.0;
+
+    if (output_buffer == NULL)
+    {
+        return false;
+    }
+
+    /* This checks for NaN and Infinity */
+    if (!compare_double(d * 0, 0))
+    {
+        length = sprintf((char*)number_buffer, "null");
+    }
+    else
+    {
+        /* Try 15 decimal places of precision to avoid nonsignificant nonzero digits */
+        length = sprintf((char*)number_buffer, "%1.15g", d);
+
+        /* Check whether the original double can be recovered */
+        if ((sscanf((char*)number_buffer, "%lg", &test) != 1) || !compare_double((double)test, d))
+        {
+            /* If not, print with 17 decimal places of precision */
+            length = sprintf((char*)number_buffer, "%1.17g", d);
+        }
+    }
+
+    /* sprintf failed or buffer overrun occurred */
+    if ((length < 0) || (length > (int)(sizeof(number_buffer) - 1)))
+    {
+        return false;
+    }
+
+    /* reserve appropriate space in the output */
+    output_pointer = ensure(output_buffer, (size_t)length + sizeof(""));
+    if (output_pointer == NULL)
+    {
+        return false;
+    }
+
+    /* copy the printed number to the output and replace locale
+     * dependent decimal point with '.' */
+    for (i = 0; i < ((size_t)length); i++)
+    {
+        if (number_buffer[i] == decimal_point)
+        {
+            output_pointer[i] = '.';
+            continue;
+        }
+
+        output_pointer[i] = number_buffer[i];
+    }
+    output_pointer[i] = '\0';
+
+    output_buffer->offset += (size_t)length;
+
+    return true;
+}
+
+/* parse 4 digit hexadecimal number */
+static unsigned parse_hex4(const unsigned char * const input)
+{
+    unsigned int h = 0;
+    size_t i = 0;
+
+    for (i = 0; i < 4; i++)
+    {
+        /* parse digit */
+        if ((input[i] >= '0') && (input[i] <= '9'))
+        {
+            h += (unsigned int) input[i] - '0';
+        }
+        else if ((input[i] >= 'A') && (input[i] <= 'F'))
+        {
+            h += (unsigned int) 10 + input[i] - 'A';
+        }
+        else if ((input[i] >= 'a') && (input[i] <= 'f'))
+        {
+            h += (unsigned int) 10 + input[i] - 'a';
+        }
+        else /* invalid */
+        {
+            return 0;
+        }
+
+        if (i < 3)
+        {
+            /* shift left to make place for the next nibble */
+            h = h << 4;
+        }
+    }
+
+    return h;
+}
+
+/* converts a UTF-16 literal to UTF-8
+ * A literal can be one or two sequences of the form \uXXXX */
+static unsigned char utf16_literal_to_utf8(const unsigned char * const input_pointer, const unsigned char * const input_end, unsigned char **output_pointer)
+{
+    long unsigned int codepoint = 0;
+    unsigned int first_code = 0;
+    const unsigned char *first_sequence = input_pointer;
+    unsigned char utf8_length = 0;
+    unsigned char utf8_position = 0;
+    unsigned char sequence_length = 0;
+    unsigned char first_byte_mark = 0;
+
+    if ((input_end - first_sequence) < 6)
+    {
+        /* input ends unexpectedly */
+        goto fail;
+    }
+
+    /* get the first utf16 sequence */
+    first_code = parse_hex4(first_sequence + 2);
+
+    /* check that the code is valid */
+    if (((first_code >= 0xDC00) && (first_code <= 0xDFFF)))
+    {
+        goto fail;
+    }
+
+    /* UTF16 surrogate pair */
+    if ((first_code >= 0xD800) && (first_code <= 0xDBFF))
+    {
+        const unsigned char *second_sequence = first_sequence + 6;
+        unsigned int second_code = 0;
+        sequence_length = 12; /* \uXXXX\uXXXX */
+
+        if ((input_end - second_sequence) < 6)
+        {
+            /* input ends unexpectedly */
+            goto fail;
+        }
+
+        if ((second_sequence[0] != '\\') || (second_sequence[1] != 'u'))
+        {
+            /* missing second half of the surrogate pair */
+            goto fail;
+        }
+
+        /* get the second utf16 sequence */
+        second_code = parse_hex4(second_sequence + 2);
+        /* check that the code is valid */
+        if ((second_code < 0xDC00) || (second_code > 0xDFFF))
+        {
+            /* invalid second half of the surrogate pair */
+            goto fail;
+        }
+
+
+        /* calculate the unicode codepoint from the surrogate pair */
+        codepoint = 0x10000 + (((first_code & 0x3FF) << 10) | (second_code & 0x3FF));
+    }
+    else
+    {
+        sequence_length = 6; /* \uXXXX */
+        codepoint = first_code;
+    }
+
+    /* encode as UTF-8
+     * takes at maximum 4 bytes to encode:
+     * 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */
+    if (codepoint < 0x80)
+    {
+        /* normal ascii, encoding 0xxxxxxx */
+        utf8_length = 1;
+    }
+    else if (codepoint < 0x800)
+    {
+        /* two bytes, encoding 110xxxxx 10xxxxxx */
+        utf8_length = 2;
+        first_byte_mark = 0xC0; /* 11000000 */
+    }
+    else if (codepoint < 0x10000)
+    {
+        /* three bytes, encoding 1110xxxx 10xxxxxx 10xxxxxx */
+        utf8_length = 3;
+        first_byte_mark = 0xE0; /* 11100000 */
+    }
+    else if (codepoint <= 0x10FFFF)
+    {
+        /* four bytes, encoding 1110xxxx 10xxxxxx 10xxxxxx 10xxxxxx */
+        utf8_length = 4;
+        first_byte_mark = 0xF0; /* 11110000 */
+    }
+    else
+    {
+        /* invalid unicode codepoint */
+        goto fail;
+    }
+
+    /* encode as utf8 */
+    for (utf8_position = (unsigned char)(utf8_length - 1); utf8_position > 0; utf8_position--)
+    {
+        /* 10xxxxxx */
+        (*output_pointer)[utf8_position] = (unsigned char)((codepoint | 0x80) & 0xBF);
+        codepoint >>= 6;
+    }
+    /* encode first byte */
+    if (utf8_length > 1)
+    {
+        (*output_pointer)[0] = (unsigned char)((codepoint | first_byte_mark) & 0xFF);
+    }
+    else
+    {
+        (*output_pointer)[0] = (unsigned char)(codepoint & 0x7F);
+    }
+
+    *output_pointer += utf8_length;
+
+    return sequence_length;
+
+fail:
+    return 0;
+}
+
+/* Parse the input text into an unescaped cinput, and populate item. */
+static cJSON_bool parse_string(cJSON * const item, parse_buffer * const input_buffer)
+{
+    const unsigned char *input_pointer = buffer_at_offset(input_buffer) + 1;
+    const unsigned char *input_end = buffer_at_offset(input_buffer) + 1;
+    unsigned char *output_pointer = NULL;
+    unsigned char *output = NULL;
+
+    /* not a string */
+    if (buffer_at_offset(input_buffer)[0] != '\"')
+    {
+        goto fail;
+    }
+
+    {
+        /* calculate approximate size of the output (overestimate) */
+        size_t allocation_length = 0;
+        size_t skipped_bytes = 0;
+        while (((size_t)(input_end - input_buffer->content) < input_buffer->length) && (*input_end != '\"'))
+        {
+            /* is escape sequence */
+            if (input_end[0] == '\\')
+            {
+                if ((size_t)(input_end + 1 - input_buffer->content) >= input_buffer->length)
+                {
+                    /* prevent buffer overflow when last input character is a backslash */
+                    goto fail;
+                }
+                skipped_bytes++;
+                input_end++;
+            }
+            input_end++;
+        }
+        if (((size_t)(input_end - input_buffer->content) >= input_buffer->length) || (*input_end != '\"'))
+        {
+            goto fail; /* string ended unexpectedly */
+        }
+
+        /* This is at most how much we need for the output */
+        allocation_length = (size_t) (input_end - buffer_at_offset(input_buffer)) - skipped_bytes;
+        output = (unsigned char*)input_buffer->hooks.allocate(allocation_length + sizeof(""));
+        if (output == NULL)
+        {
+            goto fail; /* allocation failure */
+        }
+    }
+
+    output_pointer = output;
+    /* loop through the string literal */
+    while (input_pointer < input_end)
+    {
+        if (*input_pointer != '\\')
+        {
+            *output_pointer++ = *input_pointer++;
+        }
+        /* escape sequence */
+        else
+        {
+            unsigned char sequence_length = 2;
+            if ((input_end - input_pointer) < 1)
+            {
+                goto fail;
+            }
+
+            switch (input_pointer[1])
+            {
+                case 'b':
+                    *output_pointer++ = '\b';
+                    break;
+                case 'f':
+                    *output_pointer++ = '\f';
+                    break;
+                case 'n':
+                    *output_pointer++ = '\n';
+                    break;
+                case 'r':
+                    *output_pointer++ = '\r';
+                    break;
+                case 't':
+                    *output_pointer++ = '\t';
+                    break;
+                case '\"':
+                case '\\':
+                case '/':
+                    *output_pointer++ = input_pointer[1];
+                    break;
+
+                /* UTF-16 literal */
+                case 'u':
+                    sequence_length = utf16_literal_to_utf8(input_pointer, input_end, &output_pointer);
+                    if (sequence_length == 0)
+                    {
+                        /* failed to convert UTF16-literal to UTF-8 */
+                        goto fail;
+                    }
+                    break;
+
+                default:
+                    goto fail;
+            }
+            input_pointer += sequence_length;
+        }
+    }
+
+    /* zero terminate the output */
+    *output_pointer = '\0';
+
+    item->type = cJSON_String;
+    item->valuestring = (char*)output;
+
+    input_buffer->offset = (size_t) (input_end - input_buffer->content);
+    input_buffer->offset++;
+
+    return true;
+
+fail:
+    if (output != NULL)
+    {
+        input_buffer->hooks.deallocate(output);
+    }
+
+    if (input_pointer != NULL)
+    {
+        input_buffer->offset = (size_t)(input_pointer - input_buffer->content);
+    }
+
+    return false;
+}
+
+/* Render the cstring provided to an escaped version that can be printed. */
+static cJSON_bool print_string_ptr(const unsigned char * const input, printbuffer * const output_buffer)
+{
+    const unsigned char *input_pointer = NULL;
+    unsigned char *output = NULL;
+    unsigned char *output_pointer = NULL;
+    size_t output_length = 0;
+    /* numbers of additional characters needed for escaping */
+    size_t escape_characters = 0;
+
+    if (output_buffer == NULL)
+    {
+        return false;
+    }
+
+    /* empty string */
+    if (input == NULL)
+    {
+        output = ensure(output_buffer, sizeof("\"\""));
+        if (output == NULL)
+        {
+            return false;
+        }
+        strcpy((char*)output, "\"\"");
+
+        return true;
+    }
+
+    /* set "flag" to 1 if something needs to be escaped */
+    for (input_pointer = input; *input_pointer; input_pointer++)
+    {
+        switch (*input_pointer)
+        {
+            case '\"':
+            case '\\':
+            case '\b':
+            case '\f':
+            case '\n':
+            case '\r':
+            case '\t':
+                /* one character escape sequence */
+                escape_characters++;
+                break;
+            default:
+                if (*input_pointer < 32)
+                {
+                    /* UTF-16 escape sequence uXXXX */
+                    escape_characters += 5;
+                }
+                break;
+        }
+    }
+    output_length = (size_t)(input_pointer - input) + escape_characters;
+
+    output = ensure(output_buffer, output_length + sizeof("\"\""));
+    if (output == NULL)
+    {
+        return false;
+    }
+
+    /* no characters have to be escaped */
+    if (escape_characters == 0)
+    {
+        output[0] = '\"';
+        memcpy(output + 1, input, output_length);
+        output[output_length + 1] = '\"';
+        output[output_length + 2] = '\0';
+
+        return true;
+    }
+
+    output[0] = '\"';
+    output_pointer = output + 1;
+    /* copy the string */
+    for (input_pointer = input; *input_pointer != '\0'; (void)input_pointer++, output_pointer++)
+    {
+        if ((*input_pointer > 31) && (*input_pointer != '\"') && (*input_pointer != '\\'))
+        {
+            /* normal character, copy */
+            *output_pointer = *input_pointer;
+        }
+        else
+        {
+            /* character needs to be escaped */
+            *output_pointer++ = '\\';
+            switch (*input_pointer)
+            {
+                case '\\':
+                    *output_pointer = '\\';
+                    break;
+                case '\"':
+                    *output_pointer = '\"';
+                    break;
+                case '\b':
+                    *output_pointer = 'b';
+                    break;
+                case '\f':
+                    *output_pointer = 'f';
+                    break;
+                case '\n':
+                    *output_pointer = 'n';
+                    break;
+                case '\r':
+                    *output_pointer = 'r';
+                    break;
+                case '\t':
+                    *output_pointer = 't';
+                    break;
+                default:
+                    /* escape and print as unicode codepoint */
+                    sprintf((char*)output_pointer, "u%04x", *input_pointer);
+                    output_pointer += 4;
+                    break;
+            }
+        }
+    }
+    output[output_length + 1] = '\"';
+    output[output_length + 2] = '\0';
+
+    return true;
+}
+
+/* Invoke print_string_ptr (which is useful) on an item. */
+static cJSON_bool print_string(const cJSON * const item, printbuffer * const p)
+{
+    return print_string_ptr((unsigned char*)item->valuestring, p);
+}
+
+/* Predeclare these prototypes. */
+static cJSON_bool parse_value(cJSON * const item, parse_buffer * const input_buffer);
+static cJSON_bool print_value(const cJSON * const item, printbuffer * const output_buffer);
+static cJSON_bool parse_array(cJSON * const item, parse_buffer * const input_buffer);
+static cJSON_bool print_array(const cJSON * const item, printbuffer * const output_buffer);
+static cJSON_bool parse_object(cJSON * const item, parse_buffer * const input_buffer);
+static cJSON_bool print_object(const cJSON * const item, printbuffer * const output_buffer);
+
+/* Utility to jump whitespace and cr/lf */
+static parse_buffer *buffer_skip_whitespace(parse_buffer * const buffer)
+{
+    if ((buffer == NULL) || (buffer->content == NULL))
+    {
+        return NULL;
+    }
+
+    while (can_access_at_index(buffer, 0) && (buffer_at_offset(buffer)[0] <= 32))
+    {
+       buffer->offset++;
+    }
+
+    if (buffer->offset == buffer->length)
+    {
+        buffer->offset--;
+    }
+
+    return buffer;
+}
+
+/* skip the UTF-8 BOM (byte order mark) if it is at the beginning of a buffer */
+static parse_buffer *skip_utf8_bom(parse_buffer * const buffer)
+{
+    if ((buffer == NULL) || (buffer->content == NULL) || (buffer->offset != 0))
+    {
+        return NULL;
+    }
+
+    if (can_access_at_index(buffer, 4) && (strncmp((const char*)buffer_at_offset(buffer), "\xEF\xBB\xBF", 3) == 0))
+    {
+        buffer->offset += 3;
+    }
+
+    return buffer;
+}
+
+/* Parse an object - create a new root, and populate. */
+CJSON_PUBLIC(cJSON *) cJSON_ParseWithOpts(const char *value, const char **return_parse_end, cJSON_bool require_null_terminated)
+{
+    parse_buffer buffer = { 0, 0, 0, 0, { 0, 0, 0 } };
+    cJSON *item = NULL;
+
+    /* reset error position */
+    global_error.json = NULL;
+    global_error.position = 0;
+
+    if (value == NULL)
+    {
+        goto fail;
+    }
+
+    buffer.content = (const unsigned char*)value;
+    buffer.length = strlen((const char*)value) + sizeof("");
+    buffer.offset = 0;
+    buffer.hooks = global_hooks;
+
+    item = cJSON_New_Item(&global_hooks);
+    if (item == NULL) /* memory fail */
+    {
+        goto fail;
+    }
+
+    if (!parse_value(item, buffer_skip_whitespace(skip_utf8_bom(&buffer))))
+    {
+        /* parse failure. ep is set. */
+        goto fail;
+    }
+
+    /* if we require null-terminated JSON without appended garbage, skip and then check for a null terminator */
+    if (require_null_terminated)
+    {
+        buffer_skip_whitespace(&buffer);
+        if ((buffer.offset >= buffer.length) || buffer_at_offset(&buffer)[0] != '\0')
+        {
+            goto fail;
+        }
+    }
+    if (return_parse_end)
+    {
+        *return_parse_end = (const char*)buffer_at_offset(&buffer);
+    }
+
+    return item;
+
+fail:
+    if (item != NULL)
+    {
+        cJSON_Delete(item);
+    }
+
+    if (value != NULL)
+    {
+        error local_error;
+        local_error.json = (const unsigned char*)value;
+        local_error.position = 0;
+
+        if (buffer.offset < buffer.length)
+        {
+            local_error.position = buffer.offset;
+        }
+        else if (buffer.length > 0)
+        {
+            local_error.position = buffer.length - 1;
+        }
+
+        if (return_parse_end != NULL)
+        {
+            *return_parse_end = (const char*)local_error.json + local_error.position;
+        }
+
+        global_error = local_error;
+    }
+
+    return NULL;
+}
+
+/* Default options for cJSON_Parse */
+CJSON_PUBLIC(cJSON *) cJSON_Parse(const char *value)
+{
+    return cJSON_ParseWithOpts(value, 0, 0);
+}
+
+#define cjson_min(a, b) ((a < b) ? a : b)
+
+static unsigned char *print(const cJSON * const item, cJSON_bool format, const internal_hooks * const hooks)
+{
+    static const size_t default_buffer_size = 256;
+    printbuffer buffer[1];
+    unsigned char *printed = NULL;
+
+    memset(buffer, 0, sizeof(buffer));
+
+    /* create buffer */
+    buffer->buffer = (unsigned char*) hooks->allocate(default_buffer_size);
+    buffer->length = default_buffer_size;
+    buffer->format = format;
+    buffer->hooks = *hooks;
+    if (buffer->buffer == NULL)
+    {
+        goto fail;
+    }
+
+    /* print the value */
+    if (!print_value(item, buffer))
+    {
+        goto fail;
+    }
+    update_offset(buffer);
+
+    /* check if reallocate is available */
+    if (hooks->reallocate != NULL)
+    {
+        printed = (unsigned char*) hooks->reallocate(buffer->buffer, buffer->offset + 1);
+        if (printed == NULL) {
+            goto fail;
+        }
+        buffer->buffer = NULL;
+    }
+    else /* otherwise copy the JSON over to a new buffer */
+    {
+        printed = (unsigned char*) hooks->allocate(buffer->offset + 1);
+        if (printed == NULL)
+        {
+            goto fail;
+        }
+        memcpy(printed, buffer->buffer, cjson_min(buffer->length, buffer->offset + 1));
+        printed[buffer->offset] = '\0'; /* just to be sure */
+
+        /* free the buffer */
+        hooks->deallocate(buffer->buffer);
+    }
+
+    return printed;
+
+fail:
+    if (buffer->buffer != NULL)
+    {
+        hooks->deallocate(buffer->buffer);
+    }
+
+    if (printed != NULL)
+    {
+        hooks->deallocate(printed);
+    }
+
+    return NULL;
+}
+
+/* Render a cJSON item/entity/structure to text. */
+CJSON_PUBLIC(char *) cJSON_Print(const cJSON *item)
+{
+    return (char*)print(item, true, &global_hooks);
+}
+
+CJSON_PUBLIC(char *) cJSON_PrintUnformatted(const cJSON *item)
+{
+    return (char*)print(item, false, &global_hooks);
+}
+
+CJSON_PUBLIC(char *) cJSON_PrintBuffered(const cJSON *item, int prebuffer, cJSON_bool fmt)
+{
+    printbuffer p = { 0, 0, 0, 0, 0, 0, { 0, 0, 0 } };
+
+    if (prebuffer < 0)
+    {
+        return NULL;
+    }
+
+    p.buffer = (unsigned char*)global_hooks.allocate((size_t)prebuffer);
+    if (!p.buffer)
+    {
+        return NULL;
+    }
+
+    p.length = (size_t)prebuffer;
+    p.offset = 0;
+    p.noalloc = false;
+    p.format = fmt;
+    p.hooks = global_hooks;
+
+    if (!print_value(item, &p))
+    {
+        global_hooks.deallocate(p.buffer);
+        return NULL;
+    }
+
+    return (char*)p.buffer;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_PrintPreallocated(cJSON *item, char *buffer, const int length, const cJSON_bool format)
+{
+    printbuffer p = { 0, 0, 0, 0, 0, 0, { 0, 0, 0 } };
+
+    if ((length < 0) || (buffer == NULL))
+    {
+        return false;
+    }
+
+    p.buffer = (unsigned char*)buffer;
+    p.length = (size_t)length;
+    p.offset = 0;
+    p.noalloc = true;
+    p.format = format;
+    p.hooks = global_hooks;
+
+    return print_value(item, &p);
+}
+
+/* Parser core - when encountering text, process appropriately. */
+static cJSON_bool parse_value(cJSON * const item, parse_buffer * const input_buffer)
+{
+    if ((input_buffer == NULL) || (input_buffer->content == NULL))
+    {
+        return false; /* no input */
+    }
+
+    /* parse the different types of values */
+    /* null */
+    if (can_read(input_buffer, 4) && (strncmp((const char*)buffer_at_offset(input_buffer), "null", 4) == 0))
+    {
+        item->type = cJSON_NULL;
+        input_buffer->offset += 4;
+        return true;
+    }
+    /* false */
+    if (can_read(input_buffer, 5) && (strncmp((const char*)buffer_at_offset(input_buffer), "false", 5) == 0))
+    {
+        item->type = cJSON_False;
+        input_buffer->offset += 5;
+        return true;
+    }
+    /* true */
+    if (can_read(input_buffer, 4) && (strncmp((const char*)buffer_at_offset(input_buffer), "true", 4) == 0))
+    {
+        item->type = cJSON_True;
+        item->valueint = 1;
+        input_buffer->offset += 4;
+        return true;
+    }
+    /* string */
+    if (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == '\"'))
+    {
+        return parse_string(item, input_buffer);
+    }
+    /* number */
+    if (can_access_at_index(input_buffer, 0) && ((buffer_at_offset(input_buffer)[0] == '-') || ((buffer_at_offset(input_buffer)[0] >= '0') && (buffer_at_offset(input_buffer)[0] <= '9'))))
+    {
+        return parse_number(item, input_buffer);
+    }
+    /* array */
+    if (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == '['))
+    {
+        return parse_array(item, input_buffer);
+    }
+    /* object */
+    if (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == '{'))
+    {
+        return parse_object(item, input_buffer);
+    }
+
+    return false;
+}
+
+/* Render a value to text. */
+static cJSON_bool print_value(const cJSON * const item, printbuffer * const output_buffer)
+{
+    unsigned char *output = NULL;
+
+    if ((item == NULL) || (output_buffer == NULL))
+    {
+        return false;
+    }
+
+    switch ((item->type) & 0xFF)
+    {
+        case cJSON_NULL:
+            output = ensure(output_buffer, 5);
+            if (output == NULL)
+            {
+                return false;
+            }
+            strcpy((char*)output, "null");
+            return true;
+
+        case cJSON_False:
+            output = ensure(output_buffer, 6);
+            if (output == NULL)
+            {
+                return false;
+            }
+            strcpy((char*)output, "false");
+            return true;
+
+        case cJSON_True:
+            output = ensure(output_buffer, 5);
+            if (output == NULL)
+            {
+                return false;
+            }
+            strcpy((char*)output, "true");
+            return true;
+
+        case cJSON_Number:
+            return print_number(item, output_buffer);
+
+        case cJSON_Raw:
+        {
+            size_t raw_length = 0;
+            if (item->valuestring == NULL)
+            {
+                return false;
+            }
+
+            raw_length = strlen(item->valuestring) + sizeof("");
+            output = ensure(output_buffer, raw_length);
+            if (output == NULL)
+            {
+                return false;
+            }
+            memcpy(output, item->valuestring, raw_length);
+            return true;
+        }
+
+        case cJSON_String:
+            return print_string(item, output_buffer);
+
+        case cJSON_Array:
+            return print_array(item, output_buffer);
+
+        case cJSON_Object:
+            return print_object(item, output_buffer);
+
+        default:
+            return false;
+    }
+}
+
+/* Build an array from input text. */
+static cJSON_bool parse_array(cJSON * const item, parse_buffer * const input_buffer)
+{
+    cJSON *head = NULL; /* head of the linked list */
+    cJSON *current_item = NULL;
+
+    if (input_buffer->depth >= CJSON_NESTING_LIMIT)
+    {
+        return false; /* to deeply nested */
+    }
+    input_buffer->depth++;
+
+    if (buffer_at_offset(input_buffer)[0] != '[')
+    {
+        /* not an array */
+        goto fail;
+    }
+
+    input_buffer->offset++;
+    buffer_skip_whitespace(input_buffer);
+    if (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == ']'))
+    {
+        /* empty array */
+        goto success;
+    }
+
+    /* check if we skipped to the end of the buffer */
+    if (cannot_access_at_index(input_buffer, 0))
+    {
+        input_buffer->offset--;
+        goto fail;
+    }
+
+    /* step back to character in front of the first element */
+    input_buffer->offset--;
+    /* loop through the comma separated array elements */
+    do
+    {
+        /* allocate next item */
+        cJSON *new_item = cJSON_New_Item(&(input_buffer->hooks));
+        if (new_item == NULL)
+        {
+            goto fail; /* allocation failure */
+        }
+
+        /* attach next item to list */
+        if (head == NULL)
+        {
+            /* start the linked list */
+            current_item = head = new_item;
+        }
+        else
+        {
+            /* add to the end and advance */
+            current_item->next = new_item;
+            new_item->prev = current_item;
+            current_item = new_item;
+        }
+
+        /* parse next value */
+        input_buffer->offset++;
+        buffer_skip_whitespace(input_buffer);
+        if (!parse_value(current_item, input_buffer))
+        {
+            goto fail; /* failed to parse value */
+        }
+        buffer_skip_whitespace(input_buffer);
+    }
+    while (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == ','));
+
+    if (cannot_access_at_index(input_buffer, 0) || buffer_at_offset(input_buffer)[0] != ']')
+    {
+        goto fail; /* expected end of array */
+    }
+
+success:
+    input_buffer->depth--;
+
+    item->type = cJSON_Array;
+    item->child = head;
+
+    input_buffer->offset++;
+
+    return true;
+
+fail:
+    if (head != NULL)
+    {
+        cJSON_Delete(head);
+    }
+
+    return false;
+}
+
+/* Render an array to text */
+static cJSON_bool print_array(const cJSON * const item, printbuffer * const output_buffer)
+{
+    unsigned char *output_pointer = NULL;
+    size_t length = 0;
+    cJSON *current_element = item->child;
+
+    if (output_buffer == NULL)
+    {
+        return false;
+    }
+
+    /* Compose the output array. */
+    /* opening square bracket */
+    output_pointer = ensure(output_buffer, 1);
+    if (output_pointer == NULL)
+    {
+        return false;
+    }
+
+    *output_pointer = '[';
+    output_buffer->offset++;
+    output_buffer->depth++;
+
+    while (current_element != NULL)
+    {
+        if (!print_value(current_element, output_buffer))
+        {
+            return false;
+        }
+        update_offset(output_buffer);
+        if (current_element->next)
+        {
+            length = (size_t) (output_buffer->format ? 2 : 1);
+            output_pointer = ensure(output_buffer, length + 1);
+            if (output_pointer == NULL)
+            {
+                return false;
+            }
+            *output_pointer++ = ',';
+            if(output_buffer->format)
+            {
+                *output_pointer++ = ' ';
+            }
+            *output_pointer = '\0';
+            output_buffer->offset += length;
+        }
+        current_element = current_element->next;
+    }
+
+    output_pointer = ensure(output_buffer, 2);
+    if (output_pointer == NULL)
+    {
+        return false;
+    }
+    *output_pointer++ = ']';
+    *output_pointer = '\0';
+    output_buffer->depth--;
+
+    return true;
+}
+
+/* Build an object from the text. */
+static cJSON_bool parse_object(cJSON * const item, parse_buffer * const input_buffer)
+{
+    cJSON *head = NULL; /* linked list head */
+    cJSON *current_item = NULL;
+
+    if (input_buffer->depth >= CJSON_NESTING_LIMIT)
+    {
+        return false; /* to deeply nested */
+    }
+    input_buffer->depth++;
+
+    if (cannot_access_at_index(input_buffer, 0) || (buffer_at_offset(input_buffer)[0] != '{'))
+    {
+        goto fail; /* not an object */
+    }
+
+    input_buffer->offset++;
+    buffer_skip_whitespace(input_buffer);
+    if (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == '}'))
+    {
+        goto success; /* empty object */
+    }
+
+    /* check if we skipped to the end of the buffer */
+    if (cannot_access_at_index(input_buffer, 0))
+    {
+        input_buffer->offset--;
+        goto fail;
+    }
+
+    /* step back to character in front of the first element */
+    input_buffer->offset--;
+    /* loop through the comma separated array elements */
+    do
+    {
+        /* allocate next item */
+        cJSON *new_item = cJSON_New_Item(&(input_buffer->hooks));
+        if (new_item == NULL)
+        {
+            goto fail; /* allocation failure */
+        }
+
+        /* attach next item to list */
+        if (head == NULL)
+        {
+            /* start the linked list */
+            current_item = head = new_item;
+        }
+        else
+        {
+            /* add to the end and advance */
+            current_item->next = new_item;
+            new_item->prev = current_item;
+            current_item = new_item;
+        }
+
+        /* parse the name of the child */
+        input_buffer->offset++;
+        buffer_skip_whitespace(input_buffer);
+        if (!parse_string(current_item, input_buffer))
+        {
+            goto fail; /* failed to parse name */
+        }
+        buffer_skip_whitespace(input_buffer);
+
+        /* swap valuestring and string, because we parsed the name */
+        current_item->string = current_item->valuestring;
+        current_item->valuestring = NULL;
+
+        if (cannot_access_at_index(input_buffer, 0) || (buffer_at_offset(input_buffer)[0] != ':'))
+        {
+            goto fail; /* invalid object */
+        }
+
+        /* parse the value */
+        input_buffer->offset++;
+        buffer_skip_whitespace(input_buffer);
+        if (!parse_value(current_item, input_buffer))
+        {
+            goto fail; /* failed to parse value */
+        }
+        buffer_skip_whitespace(input_buffer);
+    }
+    while (can_access_at_index(input_buffer, 0) && (buffer_at_offset(input_buffer)[0] == ','));
+
+    if (cannot_access_at_index(input_buffer, 0) || (buffer_at_offset(input_buffer)[0] != '}'))
+    {
+        goto fail; /* expected end of object */
+    }
+
+success:
+    input_buffer->depth--;
+
+    item->type = cJSON_Object;
+    item->child = head;
+
+    input_buffer->offset++;
+    return true;
+
+fail:
+    if (head != NULL)
+    {
+        cJSON_Delete(head);
+    }
+
+    return false;
+}
+
+/* Render an object to text. */
+static cJSON_bool print_object(const cJSON * const item, printbuffer * const output_buffer)
+{
+    unsigned char *output_pointer = NULL;
+    size_t length = 0;
+    cJSON *current_item = item->child;
+
+    if (output_buffer == NULL)
+    {
+        return false;
+    }
+
+    /* Compose the output: */
+    length = (size_t) (output_buffer->format ? 2 : 1); /* fmt: {\n */
+    output_pointer = ensure(output_buffer, length + 1);
+    if (output_pointer == NULL)
+    {
+        return false;
+    }
+
+    *output_pointer++ = '{';
+    output_buffer->depth++;
+    if (output_buffer->format)
+    {
+        *output_pointer++ = '\n';
+    }
+    output_buffer->offset += length;
+
+    while (current_item)
+    {
+        if (output_buffer->format)
+        {
+            size_t i;
+            output_pointer = ensure(output_buffer, output_buffer->depth);
+            if (output_pointer == NULL)
+            {
+                return false;
+            }
+            for (i = 0; i < output_buffer->depth; i++)
+            {
+                *output_pointer++ = '\t';
+            }
+            output_buffer->offset += output_buffer->depth;
+        }
+
+        /* print key */
+        if (!print_string_ptr((unsigned char*)current_item->string, output_buffer))
+        {
+            return false;
+        }
+        update_offset(output_buffer);
+
+        length = (size_t) (output_buffer->format ? 2 : 1);
+        output_pointer = ensure(output_buffer, length);
+        if (output_pointer == NULL)
+        {
+            return false;
+        }
+        *output_pointer++ = ':';
+        if (output_buffer->format)
+        {
+            *output_pointer++ = '\t';
+        }
+        output_buffer->offset += length;
+
+        /* print value */
+        if (!print_value(current_item, output_buffer))
+        {
+            return false;
+        }
+        update_offset(output_buffer);
+
+        /* print comma if not last */
+        length = ((size_t)(output_buffer->format ? 1 : 0) + (size_t)(current_item->next ? 1 : 0));
+        output_pointer = ensure(output_buffer, length + 1);
+        if (output_pointer == NULL)
+        {
+            return false;
+        }
+        if (current_item->next)
+        {
+            *output_pointer++ = ',';
+        }
+
+        if (output_buffer->format)
+        {
+            *output_pointer++ = '\n';
+        }
+        *output_pointer = '\0';
+        output_buffer->offset += length;
+
+        current_item = current_item->next;
+    }
+
+    output_pointer = ensure(output_buffer, output_buffer->format ? (output_buffer->depth + 1) : 2);
+    if (output_pointer == NULL)
+    {
+        return false;
+    }
+    if (output_buffer->format)
+    {
+        size_t i;
+        for (i = 0; i < (output_buffer->depth - 1); i++)
+        {
+            *output_pointer++ = '\t';
+        }
+    }
+    *output_pointer++ = '}';
+    *output_pointer = '\0';
+    output_buffer->depth--;
+
+    return true;
+}
+
+/* Get Array size/item / object item. */
+CJSON_PUBLIC(int) cJSON_GetArraySize(const cJSON *array)
+{
+    cJSON *child = NULL;
+    size_t size = 0;
+
+    if (array == NULL)
+    {
+        return 0;
+    }
+
+    child = array->child;
+
+    while(child != NULL)
+    {
+        size++;
+        child = child->next;
+    }
+
+    /* FIXME: Can overflow here. Cannot be fixed without breaking the API */
+
+    return (int)size;
+}
+
+static cJSON* get_array_item(const cJSON *array, size_t index)
+{
+    cJSON *current_child = NULL;
+
+    if (array == NULL)
+    {
+        return NULL;
+    }
+
+    current_child = array->child;
+    while ((current_child != NULL) && (index > 0))
+    {
+        index--;
+        current_child = current_child->next;
+    }
+
+    return current_child;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_GetArrayItem(const cJSON *array, int index)
+{
+    if (index < 0)
+    {
+        return NULL;
+    }
+
+    return get_array_item(array, (size_t)index);
+}
+
+static cJSON *get_object_item(const cJSON * const object, const char * const name, const cJSON_bool case_sensitive)
+{
+    cJSON *current_element = NULL;
+
+    if ((object == NULL) || (name == NULL))
+    {
+        return NULL;
+    }
+
+    current_element = object->child;
+    if (case_sensitive)
+    {
+        while ((current_element != NULL) && (current_element->string != NULL) && (strcmp(name, current_element->string) != 0))
+        {
+            current_element = current_element->next;
+        }
+    }
+    else
+    {
+        while ((current_element != NULL) && (case_insensitive_strcmp((const unsigned char*)name, (const unsigned char*)(current_element->string)) != 0))
+        {
+            current_element = current_element->next;
+        }
+    }
+
+    if ((current_element == NULL) || (current_element->string == NULL)) {
+        return NULL;
+    }
+
+    return current_element;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_GetObjectItem(const cJSON * const object, const char * const string)
+{
+    return get_object_item(object, string, false);
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_GetObjectItemCaseSensitive(const cJSON * const object, const char * const string)
+{
+    return get_object_item(object, string, true);
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_HasObjectItem(const cJSON *object, const char *string)
+{
+    return cJSON_GetObjectItem(object, string) ? 1 : 0;
+}
+
+/* Utility for array list handling. */
+static void suffix_object(cJSON *prev, cJSON *item)
+{
+    prev->next = item;
+    item->prev = prev;
+}
+
+/* Utility for handling references. */
+static cJSON *create_reference(const cJSON *item, const internal_hooks * const hooks)
+{
+    cJSON *reference = NULL;
+    if (item == NULL)
+    {
+        return NULL;
+    }
+
+    reference = cJSON_New_Item(hooks);
+    if (reference == NULL)
+    {
+        return NULL;
+    }
+
+    memcpy(reference, item, sizeof(cJSON));
+    reference->string = NULL;
+    reference->type |= cJSON_IsReference;
+    reference->next = reference->prev = NULL;
+    return reference;
+}
+
+static cJSON_bool add_item_to_array(cJSON *array, cJSON *item)
+{
+    cJSON *child = NULL;
+
+    if ((item == NULL) || (array == NULL))
+    {
+        return false;
+    }
+
+    child = array->child;
+
+    if (child == NULL)
+    {
+        /* list is empty, start new one */
+        array->child = item;
+    }
+    else
+    {
+        /* append to the end */
+        while (child->next)
+        {
+            child = child->next;
+        }
+        suffix_object(child, item);
+    }
+
+    return true;
+}
+
+/* Add item to array/object. */
+CJSON_PUBLIC(void) cJSON_AddItemToArray(cJSON *array, cJSON *item)
+{
+    add_item_to_array(array, item);
+}
+
+#if defined(__clang__) || (defined(__GNUC__)  && ((__GNUC__ > 4) || ((__GNUC__ == 4) && (__GNUC_MINOR__ > 5))))
+    #pragma GCC diagnostic push
+#endif
+#ifdef __GNUC__
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+/* helper function to cast away const */
+static void* cast_away_const(const void* string)
+{
+    return (void*)string;
+}
+#if defined(__clang__) || (defined(__GNUC__)  && ((__GNUC__ > 4) || ((__GNUC__ == 4) && (__GNUC_MINOR__ > 5))))
+    #pragma GCC diagnostic pop
+#endif
+
+
+static cJSON_bool add_item_to_object(cJSON * const object, const char * const string, cJSON * const item, const internal_hooks * const hooks, const cJSON_bool constant_key)
+{
+    char *new_key = NULL;
+    int new_type = cJSON_Invalid;
+
+    if ((object == NULL) || (string == NULL) || (item == NULL))
+    {
+        return false;
+    }
+
+    if (constant_key)
+    {
+        new_key = (char*)cast_away_const(string);
+        new_type = item->type | cJSON_StringIsConst;
+    }
+    else
+    {
+        new_key = (char*)cJSON_strdup((const unsigned char*)string, hooks);
+        if (new_key == NULL)
+        {
+            return false;
+        }
+
+        new_type = item->type & ~cJSON_StringIsConst;
+    }
+
+    if (!(item->type & cJSON_StringIsConst) && (item->string != NULL))
+    {
+        hooks->deallocate(item->string);
+    }
+
+    item->string = new_key;
+    item->type = new_type;
+
+    return add_item_to_array(object, item);
+}
+
+CJSON_PUBLIC(void) cJSON_AddItemToObject(cJSON *object, const char *string, cJSON *item)
+{
+    add_item_to_object(object, string, item, &global_hooks, false);
+}
+
+/* Add an item to an object with constant string as key */
+CJSON_PUBLIC(void) cJSON_AddItemToObjectCS(cJSON *object, const char *string, cJSON *item)
+{
+    add_item_to_object(object, string, item, &global_hooks, true);
+}
+
+CJSON_PUBLIC(void) cJSON_AddItemReferenceToArray(cJSON *array, cJSON *item)
+{
+    if (array == NULL)
+    {
+        return;
+    }
+
+    add_item_to_array(array, create_reference(item, &global_hooks));
+}
+
+CJSON_PUBLIC(void) cJSON_AddItemReferenceToObject(cJSON *object, const char *string, cJSON *item)
+{
+    if ((object == NULL) || (string == NULL))
+    {
+        return;
+    }
+
+    add_item_to_object(object, string, create_reference(item, &global_hooks), &global_hooks, false);
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddNullToObject(cJSON * const object, const char * const name)
+{
+    cJSON *null = cJSON_CreateNull();
+    if (add_item_to_object(object, name, null, &global_hooks, false))
+    {
+        return null;
+    }
+
+    cJSON_Delete(null);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddTrueToObject(cJSON * const object, const char * const name)
+{
+    cJSON *true_item = cJSON_CreateTrue();
+    if (add_item_to_object(object, name, true_item, &global_hooks, false))
+    {
+        return true_item;
+    }
+
+    cJSON_Delete(true_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddFalseToObject(cJSON * const object, const char * const name)
+{
+    cJSON *false_item = cJSON_CreateFalse();
+    if (add_item_to_object(object, name, false_item, &global_hooks, false))
+    {
+        return false_item;
+    }
+
+    cJSON_Delete(false_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddBoolToObject(cJSON * const object, const char * const name, const cJSON_bool boolean)
+{
+    cJSON *bool_item = cJSON_CreateBool(boolean);
+    if (add_item_to_object(object, name, bool_item, &global_hooks, false))
+    {
+        return bool_item;
+    }
+
+    cJSON_Delete(bool_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddNumberToObject(cJSON * const object, const char * const name, const double number)
+{
+    cJSON *number_item = cJSON_CreateNumber(number);
+    if (add_item_to_object(object, name, number_item, &global_hooks, false))
+    {
+        return number_item;
+    }
+
+    cJSON_Delete(number_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddStringToObject(cJSON * const object, const char * const name, const char * const string)
+{
+    cJSON *string_item = cJSON_CreateString(string);
+    if (add_item_to_object(object, name, string_item, &global_hooks, false))
+    {
+        return string_item;
+    }
+
+    cJSON_Delete(string_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddRawToObject(cJSON * const object, const char * const name, const char * const raw)
+{
+    cJSON *raw_item = cJSON_CreateRaw(raw);
+    if (add_item_to_object(object, name, raw_item, &global_hooks, false))
+    {
+        return raw_item;
+    }
+
+    cJSON_Delete(raw_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddObjectToObject(cJSON * const object, const char * const name)
+{
+    cJSON *object_item = cJSON_CreateObject();
+    if (add_item_to_object(object, name, object_item, &global_hooks, false))
+    {
+        return object_item;
+    }
+
+    cJSON_Delete(object_item);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON*) cJSON_AddArrayToObject(cJSON * const object, const char * const name)
+{
+    cJSON *array = cJSON_CreateArray();
+    if (add_item_to_object(object, name, array, &global_hooks, false))
+    {
+        return array;
+    }
+
+    cJSON_Delete(array);
+    return NULL;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemViaPointer(cJSON *parent, cJSON * const item)
+{
+    if ((parent == NULL) || (item == NULL))
+    {
+        return NULL;
+    }
+
+    if (item->prev != NULL)
+    {
+        /* not the first element */
+        item->prev->next = item->next;
+    }
+    if (item->next != NULL)
+    {
+        /* not the last element */
+        item->next->prev = item->prev;
+    }
+
+    if (item == parent->child)
+    {
+        /* first element */
+        parent->child = item->next;
+    }
+    /* make sure the detached item doesn't point anywhere anymore */
+    item->prev = NULL;
+    item->next = NULL;
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromArray(cJSON *array, int which)
+{
+    if (which < 0)
+    {
+        return NULL;
+    }
+
+    return cJSON_DetachItemViaPointer(array, get_array_item(array, (size_t)which));
+}
+
+CJSON_PUBLIC(void) cJSON_DeleteItemFromArray(cJSON *array, int which)
+{
+    cJSON_Delete(cJSON_DetachItemFromArray(array, which));
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromObject(cJSON *object, const char *string)
+{
+    cJSON *to_detach = cJSON_GetObjectItem(object, string);
+
+    return cJSON_DetachItemViaPointer(object, to_detach);
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromObjectCaseSensitive(cJSON *object, const char *string)
+{
+    cJSON *to_detach = cJSON_GetObjectItemCaseSensitive(object, string);
+
+    return cJSON_DetachItemViaPointer(object, to_detach);
+}
+
+CJSON_PUBLIC(void) cJSON_DeleteItemFromObject(cJSON *object, const char *string)
+{
+    cJSON_Delete(cJSON_DetachItemFromObject(object, string));
+}
+
+CJSON_PUBLIC(void) cJSON_DeleteItemFromObjectCaseSensitive(cJSON *object, const char *string)
+{
+    cJSON_Delete(cJSON_DetachItemFromObjectCaseSensitive(object, string));
+}
+
+/* Replace array/object items with new ones. */
+CJSON_PUBLIC(void) cJSON_InsertItemInArray(cJSON *array, int which, cJSON *newitem)
+{
+    cJSON *after_inserted = NULL;
+
+    if (which < 0)
+    {
+        return;
+    }
+
+    after_inserted = get_array_item(array, (size_t)which);
+    if (after_inserted == NULL)
+    {
+        add_item_to_array(array, newitem);
+        return;
+    }
+
+    newitem->next = after_inserted;
+    newitem->prev = after_inserted->prev;
+    after_inserted->prev = newitem;
+    if (after_inserted == array->child)
+    {
+        array->child = newitem;
+    }
+    else
+    {
+        newitem->prev->next = newitem;
+    }
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_ReplaceItemViaPointer(cJSON * const parent, cJSON * const item, cJSON * replacement)
+{
+    if ((parent == NULL) || (replacement == NULL) || (item == NULL))
+    {
+        return false;
+    }
+
+    if (replacement == item)
+    {
+        return true;
+    }
+
+    replacement->next = item->next;
+    replacement->prev = item->prev;
+
+    if (replacement->next != NULL)
+    {
+        replacement->next->prev = replacement;
+    }
+    if (replacement->prev != NULL)
+    {
+        replacement->prev->next = replacement;
+    }
+    if (parent->child == item)
+    {
+        parent->child = replacement;
+    }
+
+    item->next = NULL;
+    item->prev = NULL;
+    cJSON_Delete(item);
+
+    return true;
+}
+
+CJSON_PUBLIC(void) cJSON_ReplaceItemInArray(cJSON *array, int which, cJSON *newitem)
+{
+    if (which < 0)
+    {
+        return;
+    }
+
+    cJSON_ReplaceItemViaPointer(array, get_array_item(array, (size_t)which), newitem);
+}
+
+static cJSON_bool replace_item_in_object(cJSON *object, const char *string, cJSON *replacement, cJSON_bool case_sensitive)
+{
+    if ((replacement == NULL) || (string == NULL))
+    {
+        return false;
+    }
+
+    /* replace the name in the replacement */
+    if (!(replacement->type & cJSON_StringIsConst) && (replacement->string != NULL))
+    {
+        cJSON_free(replacement->string);
+    }
+    replacement->string = (char*)cJSON_strdup((const unsigned char*)string, &global_hooks);
+    replacement->type &= ~cJSON_StringIsConst;
+
+    cJSON_ReplaceItemViaPointer(object, get_object_item(object, string, case_sensitive), replacement);
+
+    return true;
+}
+
+CJSON_PUBLIC(void) cJSON_ReplaceItemInObject(cJSON *object, const char *string, cJSON *newitem)
+{
+    replace_item_in_object(object, string, newitem, false);
+}
+
+CJSON_PUBLIC(void) cJSON_ReplaceItemInObjectCaseSensitive(cJSON *object, const char *string, cJSON *newitem)
+{
+    replace_item_in_object(object, string, newitem, true);
+}
+
+/* Create basic types: */
+CJSON_PUBLIC(cJSON *) cJSON_CreateNull(void)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_NULL;
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateTrue(void)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_True;
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateFalse(void)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_False;
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateBool(cJSON_bool boolean)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = boolean ? cJSON_True : cJSON_False;
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateNumber(double num)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_Number;
+        item->valuedouble = num;
+
+        /* use saturation in case of overflow */
+        if (num >= INT_MAX)
+        {
+            item->valueint = INT_MAX;
+        }
+        else if (num <= (double)INT_MIN)
+        {
+            item->valueint = INT_MIN;
+        }
+        else
+        {
+            item->valueint = (int)num;
+        }
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateString(const char *string)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_String;
+        item->valuestring = (char*)cJSON_strdup((const unsigned char*)string, &global_hooks);
+        if(!item->valuestring)
+        {
+            cJSON_Delete(item);
+            return NULL;
+        }
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateStringReference(const char *string)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if (item != NULL)
+    {
+        item->type = cJSON_String | cJSON_IsReference;
+        item->valuestring = (char*)cast_away_const(string);
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateObjectReference(const cJSON *child)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if (item != NULL) {
+        item->type = cJSON_Object | cJSON_IsReference;
+        item->child = (cJSON*)cast_away_const(child);
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateArrayReference(const cJSON *child) {
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if (item != NULL) {
+        item->type = cJSON_Array | cJSON_IsReference;
+        item->child = (cJSON*)cast_away_const(child);
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateRaw(const char *raw)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type = cJSON_Raw;
+        item->valuestring = (char*)cJSON_strdup((const unsigned char*)raw, &global_hooks);
+        if(!item->valuestring)
+        {
+            cJSON_Delete(item);
+            return NULL;
+        }
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateArray(void)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if(item)
+    {
+        item->type=cJSON_Array;
+    }
+
+    return item;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateObject(void)
+{
+    cJSON *item = cJSON_New_Item(&global_hooks);
+    if (item)
+    {
+        item->type = cJSON_Object;
+    }
+
+    return item;
+}
+
+/* Create Arrays: */
+CJSON_PUBLIC(cJSON *) cJSON_CreateIntArray(const int *numbers, int count)
+{
+    size_t i = 0;
+    cJSON *n = NULL;
+    cJSON *p = NULL;
+    cJSON *a = NULL;
+
+    if ((count < 0) || (numbers == NULL))
+    {
+        return NULL;
+    }
+
+    a = cJSON_CreateArray();
+    for(i = 0; a && (i < (size_t)count); i++)
+    {
+        n = cJSON_CreateNumber(numbers[i]);
+        if (!n)
+        {
+            cJSON_Delete(a);
+            return NULL;
+        }
+        if(!i)
+        {
+            a->child = n;
+        }
+        else
+        {
+            suffix_object(p, n);
+        }
+        p = n;
+    }
+
+    return a;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateFloatArray(const float *numbers, int count)
+{
+    size_t i = 0;
+    cJSON *n = NULL;
+    cJSON *p = NULL;
+    cJSON *a = NULL;
+
+    if ((count < 0) || (numbers == NULL))
+    {
+        return NULL;
+    }
+
+    a = cJSON_CreateArray();
+
+    for(i = 0; a && (i < (size_t)count); i++)
+    {
+        n = cJSON_CreateNumber((double)numbers[i]);
+        if(!n)
+        {
+            cJSON_Delete(a);
+            return NULL;
+        }
+        if(!i)
+        {
+            a->child = n;
+        }
+        else
+        {
+            suffix_object(p, n);
+        }
+        p = n;
+    }
+
+    return a;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateDoubleArray(const double *numbers, int count)
+{
+    size_t i = 0;
+    cJSON *n = NULL;
+    cJSON *p = NULL;
+    cJSON *a = NULL;
+
+    if ((count < 0) || (numbers == NULL))
+    {
+        return NULL;
+    }
+
+    a = cJSON_CreateArray();
+
+    for(i = 0;a && (i < (size_t)count); i++)
+    {
+        n = cJSON_CreateNumber(numbers[i]);
+        if(!n)
+        {
+            cJSON_Delete(a);
+            return NULL;
+        }
+        if(!i)
+        {
+            a->child = n;
+        }
+        else
+        {
+            suffix_object(p, n);
+        }
+        p = n;
+    }
+
+    return a;
+}
+
+CJSON_PUBLIC(cJSON *) cJSON_CreateStringArray(const char *const *strings, int count)
+{
+    size_t i = 0;
+    cJSON *n = NULL;
+    cJSON *p = NULL;
+    cJSON *a = NULL;
+
+    if ((count < 0) || (strings == NULL))
+    {
+        return NULL;
+    }
+
+    a = cJSON_CreateArray();
+
+    for (i = 0; a && (i < (size_t)count); i++)
+    {
+        n = cJSON_CreateString(strings[i]);
+        if(!n)
+        {
+            cJSON_Delete(a);
+            return NULL;
+        }
+        if(!i)
+        {
+            a->child = n;
+        }
+        else
+        {
+            suffix_object(p,n);
+        }
+        p = n;
+    }
+
+    return a;
+}
+
+/* Duplication */
+CJSON_PUBLIC(cJSON *) cJSON_Duplicate(const cJSON *item, cJSON_bool recurse)
+{
+    cJSON *newitem = NULL;
+    cJSON *child = NULL;
+    cJSON *next = NULL;
+    cJSON *newchild = NULL;
+
+    /* Bail on bad ptr */
+    if (!item)
+    {
+        goto fail;
+    }
+    /* Create new item */
+    newitem = cJSON_New_Item(&global_hooks);
+    if (!newitem)
+    {
+        goto fail;
+    }
+    /* Copy over all vars */
+    newitem->type = item->type & (~cJSON_IsReference);
+    newitem->valueint = item->valueint;
+    newitem->valuedouble = item->valuedouble;
+    if (item->valuestring)
+    {
+        newitem->valuestring = (char*)cJSON_strdup((unsigned char*)item->valuestring, &global_hooks);
+        if (!newitem->valuestring)
+        {
+            goto fail;
+        }
+    }
+    if (item->string)
+    {
+        newitem->string = (item->type&cJSON_StringIsConst) ? item->string : (char*)cJSON_strdup((unsigned char*)item->string, &global_hooks);
+        if (!newitem->string)
+        {
+            goto fail;
+        }
+    }
+    /* If non-recursive, then we're done! */
+    if (!recurse)
+    {
+        return newitem;
+    }
+    /* Walk the ->next chain for the child. */
+    child = item->child;
+    while (child != NULL)
+    {
+        newchild = cJSON_Duplicate(child, true); /* Duplicate (with recurse) each item in the ->next chain */
+        if (!newchild)
+        {
+            goto fail;
+        }
+        if (next != NULL)
+        {
+            /* If newitem->child already set, then crosswire ->prev and ->next and move on */
+            next->next = newchild;
+            newchild->prev = next;
+            next = newchild;
+        }
+        else
+        {
+            /* Set newitem->child and move to it */
+            newitem->child = newchild;
+            next = newchild;
+        }
+        child = child->next;
+    }
+
+    return newitem;
+
+fail:
+    if (newitem != NULL)
+    {
+        cJSON_Delete(newitem);
+    }
+
+    return NULL;
+}
+
+static void skip_oneline_comment(char **input)
+{
+    *input += static_strlen("//");
+
+    for (; (*input)[0] != '\0'; ++(*input))
+    {
+        if ((*input)[0] == '\n') {
+            *input += static_strlen("\n");
+            return;
+        }
+    }
+}
+
+static void skip_multiline_comment(char **input)
+{
+    *input += static_strlen("/*");
+
+    for (; (*input)[0] != '\0'; ++(*input))
+    {
+        if (((*input)[0] == '*') && ((*input)[1] == '/'))
+        {
+            *input += static_strlen("*/");
+            return;
+        }
+    }
+}
+
+static void minify_string(char **input, char **output) {
+    (*output)[0] = (*input)[0];
+    *input += static_strlen("\"");
+    *output += static_strlen("\"");
+
+
+    for (; (*input)[0] != '\0'; (void)++(*input), ++(*output)) {
+        (*output)[0] = (*input)[0];
+
+        if ((*input)[0] == '\"') {
+            (*output)[0] = '\"';
+            *input += static_strlen("\"");
+            *output += static_strlen("\"");
+            return;
+        } else if (((*input)[0] == '\\') && ((*input)[1] == '\"')) {
+            (*output)[1] = (*input)[1];
+            *input += static_strlen("\"");
+            *output += static_strlen("\"");
+        }
+    }
+}
+
+CJSON_PUBLIC(void) cJSON_Minify(char *json)
+{
+    char *into = json;
+
+    if (json == NULL)
+    {
+        return;
+    }
+
+    while (json[0] != '\0')
+    {
+        switch (json[0])
+        {
+            case ' ':
+            case '\t':
+            case '\r':
+            case '\n':
+                json++;
+                break;
+
+            case '/':
+                if (json[1] == '/')
+                {
+                    skip_oneline_comment(&json);
+                }
+                else if (json[1] == '*')
+                {
+                    skip_multiline_comment(&json);
+                } else {
+                    json++;
+                }
+                break;
+
+            case '\"':
+                minify_string(&json, (char**)&into);
+                break;
+
+            default:
+                into[0] = json[0];
+                json++;
+                into++;
+        }
+    }
+
+    /* and null-terminate. */
+    *into = '\0';
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsInvalid(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_Invalid;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsFalse(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_False;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsTrue(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xff) == cJSON_True;
+}
+
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsBool(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & (cJSON_True | cJSON_False)) != 0;
+}
+CJSON_PUBLIC(cJSON_bool) cJSON_IsNull(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_NULL;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsNumber(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_Number;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsString(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_String;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsArray(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_Array;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsObject(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_Object;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_IsRaw(const cJSON * const item)
+{
+    if (item == NULL)
+    {
+        return false;
+    }
+
+    return (item->type & 0xFF) == cJSON_Raw;
+}
+
+CJSON_PUBLIC(cJSON_bool) cJSON_Compare(const cJSON * const a, const cJSON * const b, const cJSON_bool case_sensitive)
+{
+    if ((a == NULL) || (b == NULL) || ((a->type & 0xFF) != (b->type & 0xFF)) || cJSON_IsInvalid(a))
+    {
+        return false;
+    }
+
+    /* check if type is valid */
+    switch (a->type & 0xFF)
+    {
+        case cJSON_False:
+        case cJSON_True:
+        case cJSON_NULL:
+        case cJSON_Number:
+        case cJSON_String:
+        case cJSON_Raw:
+        case cJSON_Array:
+        case cJSON_Object:
+            break;
+
+        default:
+            return false;
+    }
+
+    /* identical objects are equal */
+    if (a == b)
+    {
+        return true;
+    }
+
+    switch (a->type & 0xFF)
+    {
+        /* in these cases and equal type is enough */
+        case cJSON_False:
+        case cJSON_True:
+        case cJSON_NULL:
+            return true;
+
+        case cJSON_Number:
+            if (compare_double(a->valuedouble, b->valuedouble))
+            {
+                return true;
+            }
+            return false;
+
+        case cJSON_String:
+        case cJSON_Raw:
+            if ((a->valuestring == NULL) || (b->valuestring == NULL))
+            {
+                return false;
+            }
+            if (strcmp(a->valuestring, b->valuestring) == 0)
+            {
+                return true;
+            }
+
+            return false;
+
+        case cJSON_Array:
+        {
+            cJSON *a_element = a->child;
+            cJSON *b_element = b->child;
+
+            for (; (a_element != NULL) && (b_element != NULL);)
+            {
+                if (!cJSON_Compare(a_element, b_element, case_sensitive))
+                {
+                    return false;
+                }
+
+                a_element = a_element->next;
+                b_element = b_element->next;
+            }
+
+            /* one of the arrays is longer than the other */
+            if (a_element != b_element) {
+                return false;
+            }
+
+            return true;
+        }
+
+        case cJSON_Object:
+        {
+            cJSON *a_element = NULL;
+            cJSON *b_element = NULL;
+            cJSON_ArrayForEach(a_element, a)
+            {
+                /* TODO This has O(n^2) runtime, which is horrible! */
+                b_element = get_object_item(b, a_element->string, case_sensitive);
+                if (b_element == NULL)
+                {
+                    return false;
+                }
+
+                if (!cJSON_Compare(a_element, b_element, case_sensitive))
+                {
+                    return false;
+                }
+            }
+
+            /* doing this twice, once on a and b to prevent true comparison if a subset of b
+             * TODO: Do this the proper way, this is just a fix for now */
+            cJSON_ArrayForEach(b_element, b)
+            {
+                a_element = get_object_item(a, b_element->string, case_sensitive);
+                if (a_element == NULL)
+                {
+                    return false;
+                }
+
+                if (!cJSON_Compare(b_element, a_element, case_sensitive))
+                {
+                    return false;
+                }
+            }
+
+            return true;
+        }
+
+        default:
+            return false;
+    }
+}
+
+CJSON_PUBLIC(void *) cJSON_malloc(size_t size)
+{
+    return global_hooks.allocate(size);
+}
+
+CJSON_PUBLIC(void) cJSON_free(void *object)
+{
+    global_hooks.deallocate(object);
+}
+
diff --git a/src/external/cJSON.h b/src/external/cJSON.h
new file mode 100644
index 0000000..23d7851
--- /dev/null
+++ b/src/external/cJSON.h
@@ -0,0 +1,294 @@
+/*
+  Copyright (c) 2009-2017 Dave Gamble and cJSON contributors
+
+  Permission is hereby granted, free of charge, to any person obtaining a copy
+  of this software and associated documentation files (the "Software"), to deal
+  in the Software without restriction, including without limitation the rights
+  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+  copies of the Software, and to permit persons to whom the Software is
+  furnished to do so, subject to the following conditions:
+
+  The above copyright notice and this permission notice shall be included in
+  all copies or substantial portions of the Software.
+
+  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+  THE SOFTWARE.
+*/
+
+#ifndef cJSON__h
+#define cJSON__h
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+#if !defined(__WINDOWS__) && (defined(WIN32) || defined(WIN64) || defined(_MSC_VER) || defined(_WIN32))
+#define __WINDOWS__
+#endif
+
+#ifdef __WINDOWS__
+
+/* When compiling for windows, we specify a specific calling convention to avoid issues where we are being called from a project with a different default calling convention.  For windows you have 3 define options:
+
+CJSON_HIDE_SYMBOLS - Define this in the case where you don't want to ever dllexport symbols
+CJSON_EXPORT_SYMBOLS - Define this on library build when you want to dllexport symbols (default)
+CJSON_IMPORT_SYMBOLS - Define this if you want to dllimport symbol
+
+For *nix builds that support visibility attribute, you can define similar behavior by
+
+setting default visibility to hidden by adding
+-fvisibility=hidden (for gcc)
+or
+-xldscope=hidden (for sun cc)
+to CFLAGS
+
+then using the CJSON_API_VISIBILITY flag to "export" the same symbols the way CJSON_EXPORT_SYMBOLS does
+
+*/
+
+#define CJSON_CDECL __cdecl
+#define CJSON_STDCALL __stdcall
+
+/* export symbols by default, this is necessary for copy pasting the C and header file */
+#if !defined(CJSON_HIDE_SYMBOLS) && !defined(CJSON_IMPORT_SYMBOLS) && !defined(CJSON_EXPORT_SYMBOLS)
+#define CJSON_EXPORT_SYMBOLS
+#endif
+
+#if defined(CJSON_HIDE_SYMBOLS)
+#define CJSON_PUBLIC(type)   type CJSON_STDCALL
+#elif defined(CJSON_EXPORT_SYMBOLS)
+#define CJSON_PUBLIC(type)   __declspec(dllexport) type CJSON_STDCALL
+#elif defined(CJSON_IMPORT_SYMBOLS)
+#define CJSON_PUBLIC(type)   __declspec(dllimport) type CJSON_STDCALL
+#endif
+#else /* !__WINDOWS__ */
+#define CJSON_CDECL
+#define CJSON_STDCALL
+
+#if (defined(__GNUC__) || defined(__SUNPRO_CC) || defined (__SUNPRO_C)) && defined(CJSON_API_VISIBILITY)
+#define CJSON_PUBLIC(type)   __attribute__((visibility("default"))) type
+#else
+#define CJSON_PUBLIC(type) type
+#endif
+#endif
+
+/* project version */
+#define CJSON_VERSION_MAJOR 1
+#define CJSON_VERSION_MINOR 7
+#define CJSON_VERSION_PATCH 12
+
+#include <stddef.h>
+
+/* cJSON Types: */
+#define cJSON_Invalid (0)
+#define cJSON_False  (1 << 0)
+#define cJSON_True   (1 << 1)
+#define cJSON_NULL   (1 << 2)
+#define cJSON_Number (1 << 3)
+#define cJSON_String (1 << 4)
+#define cJSON_Array  (1 << 5)
+#define cJSON_Object (1 << 6)
+#define cJSON_Raw    (1 << 7) /* raw json */
+
+#define cJSON_IsReference 256
+#define cJSON_StringIsConst 512
+
+/* The cJSON structure: */
+typedef struct cJSON
+{
+    /* next/prev allow you to walk array/object chains. Alternatively, use GetArraySize/GetArrayItem/GetObjectItem */
+    struct cJSON *next;
+    struct cJSON *prev;
+    /* An array or object item will have a child pointer pointing to a chain of the items in the array/object. */
+    struct cJSON *child;
+
+    /* The type of the item, as above. */
+    int type;
+
+    /* The item's string, if type==cJSON_String  and type == cJSON_Raw */
+    char *valuestring;
+    /* writing to valueint is DEPRECATED, use cJSON_SetNumberValue instead */
+    int valueint;
+    /* The item's number, if type==cJSON_Number */
+    double valuedouble;
+
+    /* The item's name string, if this item is the child of, or is in the list of subitems of an object. */
+    char *string;
+} cJSON;
+
+typedef struct cJSON_Hooks
+{
+      /* malloc/free are CDECL on Windows regardless of the default calling convention of the compiler, so ensure the hooks allow passing those functions directly. */
+      void *(CJSON_CDECL *malloc_fn)(size_t sz);
+      void (CJSON_CDECL *free_fn)(void *ptr);
+} cJSON_Hooks;
+
+typedef int cJSON_bool;
+
+/* Limits how deeply nested arrays/objects can be before cJSON rejects to parse them.
+ * This is to prevent stack overflows. */
+#ifndef CJSON_NESTING_LIMIT
+#define CJSON_NESTING_LIMIT 1000
+#endif
+
+/* Precision of double variables comparison */
+#ifndef CJSON_DOUBLE_PRECISION
+#define CJSON_DOUBLE_PRECISION .0000000000000001
+#endif
+
+/* returns the version of cJSON as a string */
+CJSON_PUBLIC(const char*) cJSON_Version(void);
+
+/* Supply malloc, realloc and free functions to cJSON */
+CJSON_PUBLIC(void) cJSON_InitHooks(cJSON_Hooks* hooks);
+
+/* Memory Management: the caller is always responsible to free the results from all variants of cJSON_Parse (with cJSON_Delete) and cJSON_Print (with stdlib free, cJSON_Hooks.free_fn, or cJSON_free as appropriate). The exception is cJSON_PrintPreallocated, where the caller has full responsibility of the buffer. */
+/* Supply a block of JSON, and this returns a cJSON object you can interrogate. */
+CJSON_PUBLIC(cJSON *) cJSON_Parse(const char *value);
+/* ParseWithOpts allows you to require (and check) that the JSON is null terminated, and to retrieve the pointer to the final byte parsed. */
+/* If you supply a ptr in return_parse_end and parsing fails, then return_parse_end will contain a pointer to the error so will match cJSON_GetErrorPtr(). */
+CJSON_PUBLIC(cJSON *) cJSON_ParseWithOpts(const char *value, const char **return_parse_end, cJSON_bool require_null_terminated);
+
+/* Render a cJSON entity to text for transfer/storage. */
+CJSON_PUBLIC(char *) cJSON_Print(const cJSON *item);
+/* Render a cJSON entity to text for transfer/storage without any formatting. */
+CJSON_PUBLIC(char *) cJSON_PrintUnformatted(const cJSON *item);
+/* Render a cJSON entity to text using a buffered strategy. prebuffer is a guess at the final size. guessing well reduces reallocation. fmt=0 gives unformatted, =1 gives formatted */
+CJSON_PUBLIC(char *) cJSON_PrintBuffered(const cJSON *item, int prebuffer, cJSON_bool fmt);
+/* Render a cJSON entity to text using a buffer already allocated in memory with given length. Returns 1 on success and 0 on failure. */
+/* NOTE: cJSON is not always 100% accurate in estimating how much memory it will use, so to be safe allocate 5 bytes more than you actually need */
+CJSON_PUBLIC(cJSON_bool) cJSON_PrintPreallocated(cJSON *item, char *buffer, const int length, const cJSON_bool format);
+/* Delete a cJSON entity and all subentities. */
+CJSON_PUBLIC(void) cJSON_Delete(cJSON *item);
+
+/* Returns the number of items in an array (or object). */
+CJSON_PUBLIC(int) cJSON_GetArraySize(const cJSON *array);
+/* Retrieve item number "index" from array "array". Returns NULL if unsuccessful. */
+CJSON_PUBLIC(cJSON *) cJSON_GetArrayItem(const cJSON *array, int index);
+/* Get item "string" from object. Case insensitive. */
+CJSON_PUBLIC(cJSON *) cJSON_GetObjectItem(const cJSON * const object, const char * const string);
+CJSON_PUBLIC(cJSON *) cJSON_GetObjectItemCaseSensitive(const cJSON * const object, const char * const string);
+CJSON_PUBLIC(cJSON_bool) cJSON_HasObjectItem(const cJSON *object, const char *string);
+/* For analysing failed parses. This returns a pointer to the parse error. You'll probably need to look a few chars back to make sense of it. Defined when cJSON_Parse() returns 0. 0 when cJSON_Parse() succeeds. */
+CJSON_PUBLIC(const char *) cJSON_GetErrorPtr(void);
+
+/* Check if the item is a string and return its valuestring */
+CJSON_PUBLIC(char *) cJSON_GetStringValue(const cJSON * const item);
+
+/* These functions check the type of an item */
+CJSON_PUBLIC(cJSON_bool) cJSON_IsInvalid(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsFalse(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsTrue(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsBool(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsNull(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsNumber(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsString(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsArray(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsObject(const cJSON * const item);
+CJSON_PUBLIC(cJSON_bool) cJSON_IsRaw(const cJSON * const item);
+
+/* These calls create a cJSON item of the appropriate type. */
+CJSON_PUBLIC(cJSON *) cJSON_CreateNull(void);
+CJSON_PUBLIC(cJSON *) cJSON_CreateTrue(void);
+CJSON_PUBLIC(cJSON *) cJSON_CreateFalse(void);
+CJSON_PUBLIC(cJSON *) cJSON_CreateBool(cJSON_bool boolean);
+CJSON_PUBLIC(cJSON *) cJSON_CreateNumber(double num);
+CJSON_PUBLIC(cJSON *) cJSON_CreateString(const char *string);
+/* raw json */
+CJSON_PUBLIC(cJSON *) cJSON_CreateRaw(const char *raw);
+CJSON_PUBLIC(cJSON *) cJSON_CreateArray(void);
+CJSON_PUBLIC(cJSON *) cJSON_CreateObject(void);
+
+/* Create a string where valuestring references a string so
+ * it will not be freed by cJSON_Delete */
+CJSON_PUBLIC(cJSON *) cJSON_CreateStringReference(const char *string);
+/* Create an object/array that only references it's elements so
+ * they will not be freed by cJSON_Delete */
+CJSON_PUBLIC(cJSON *) cJSON_CreateObjectReference(const cJSON *child);
+CJSON_PUBLIC(cJSON *) cJSON_CreateArrayReference(const cJSON *child);
+
+/* These utilities create an Array of count items.
+ * The parameter count cannot be greater than the number of elements in the number array, otherwise array access will be out of bounds.*/
+CJSON_PUBLIC(cJSON *) cJSON_CreateIntArray(const int *numbers, int count);
+CJSON_PUBLIC(cJSON *) cJSON_CreateFloatArray(const float *numbers, int count);
+CJSON_PUBLIC(cJSON *) cJSON_CreateDoubleArray(const double *numbers, int count);
+CJSON_PUBLIC(cJSON *) cJSON_CreateStringArray(const char *const *strings, int count);
+
+/* Append item to the specified array/object. */
+CJSON_PUBLIC(void) cJSON_AddItemToArray(cJSON *array, cJSON *item);
+CJSON_PUBLIC(void) cJSON_AddItemToObject(cJSON *object, const char *string, cJSON *item);
+/* Use this when string is definitely const (i.e. a literal, or as good as), and will definitely survive the cJSON object.
+ * WARNING: When this function was used, make sure to always check that (item->type & cJSON_StringIsConst) is zero before
+ * writing to `item->string` */
+CJSON_PUBLIC(void) cJSON_AddItemToObjectCS(cJSON *object, const char *string, cJSON *item);
+/* Append reference to item to the specified array/object. Use this when you want to add an existing cJSON to a new cJSON, but don't want to corrupt your existing cJSON. */
+CJSON_PUBLIC(void) cJSON_AddItemReferenceToArray(cJSON *array, cJSON *item);
+CJSON_PUBLIC(void) cJSON_AddItemReferenceToObject(cJSON *object, const char *string, cJSON *item);
+
+/* Remove/Detach items from Arrays/Objects. */
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemViaPointer(cJSON *parent, cJSON * const item);
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromArray(cJSON *array, int which);
+CJSON_PUBLIC(void) cJSON_DeleteItemFromArray(cJSON *array, int which);
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromObject(cJSON *object, const char *string);
+CJSON_PUBLIC(cJSON *) cJSON_DetachItemFromObjectCaseSensitive(cJSON *object, const char *string);
+CJSON_PUBLIC(void) cJSON_DeleteItemFromObject(cJSON *object, const char *string);
+CJSON_PUBLIC(void) cJSON_DeleteItemFromObjectCaseSensitive(cJSON *object, const char *string);
+
+/* Update array items. */
+CJSON_PUBLIC(void) cJSON_InsertItemInArray(cJSON *array, int which, cJSON *newitem); /* Shifts pre-existing items to the right. */
+CJSON_PUBLIC(cJSON_bool) cJSON_ReplaceItemViaPointer(cJSON * const parent, cJSON * const item, cJSON * replacement);
+CJSON_PUBLIC(void) cJSON_ReplaceItemInArray(cJSON *array, int which, cJSON *newitem);
+CJSON_PUBLIC(void) cJSON_ReplaceItemInObject(cJSON *object,const char *string,cJSON *newitem);
+CJSON_PUBLIC(void) cJSON_ReplaceItemInObjectCaseSensitive(cJSON *object,const char *string,cJSON *newitem);
+
+/* Duplicate a cJSON item */
+CJSON_PUBLIC(cJSON *) cJSON_Duplicate(const cJSON *item, cJSON_bool recurse);
+/* Duplicate will create a new, identical cJSON item to the one you pass, in new memory that will
+ * need to be released. With recurse!=0, it will duplicate any children connected to the item.
+ * The item->next and ->prev pointers are always zero on return from Duplicate. */
+/* Recursively compare two cJSON items for equality. If either a or b is NULL or invalid, they will be considered unequal.
+ * case_sensitive determines if object keys are treated case sensitive (1) or case insensitive (0) */
+CJSON_PUBLIC(cJSON_bool) cJSON_Compare(const cJSON * const a, const cJSON * const b, const cJSON_bool case_sensitive);
+
+/* Minify a strings, remove blank characters(such as ' ', '\t', '\r', '\n') from strings.
+ * The input pointer json cannot point to a read-only address area, such as a string constant, 
+ * but should point to a readable and writable adress area. */
+CJSON_PUBLIC(void) cJSON_Minify(char *json);
+
+/* Helper functions for creating and adding items to an object at the same time.
+ * They return the added item or NULL on failure. */
+CJSON_PUBLIC(cJSON*) cJSON_AddNullToObject(cJSON * const object, const char * const name);
+CJSON_PUBLIC(cJSON*) cJSON_AddTrueToObject(cJSON * const object, const char * const name);
+CJSON_PUBLIC(cJSON*) cJSON_AddFalseToObject(cJSON * const object, const char * const name);
+CJSON_PUBLIC(cJSON*) cJSON_AddBoolToObject(cJSON * const object, const char * const name, const cJSON_bool boolean);
+CJSON_PUBLIC(cJSON*) cJSON_AddNumberToObject(cJSON * const object, const char * const name, const double number);
+CJSON_PUBLIC(cJSON*) cJSON_AddStringToObject(cJSON * const object, const char * const name, const char * const string);
+CJSON_PUBLIC(cJSON*) cJSON_AddRawToObject(cJSON * const object, const char * const name, const char * const raw);
+CJSON_PUBLIC(cJSON*) cJSON_AddObjectToObject(cJSON * const object, const char * const name);
+CJSON_PUBLIC(cJSON*) cJSON_AddArrayToObject(cJSON * const object, const char * const name);
+
+/* When assigning an integer value, it needs to be propagated to valuedouble too. */
+#define cJSON_SetIntValue(object, number) ((object) ? (object)->valueint = (object)->valuedouble = (number) : (number))
+/* helper for the cJSON_SetNumberValue macro */
+CJSON_PUBLIC(double) cJSON_SetNumberHelper(cJSON *object, double number);
+#define cJSON_SetNumberValue(object, number) ((object != NULL) ? cJSON_SetNumberHelper(object, (double)number) : (number))
+
+/* Macro for iterating over an array or object */
+#define cJSON_ArrayForEach(element, array) for(element = (array != NULL) ? (array)->child : NULL; element != NULL; element = element->next)
+
+/* malloc/free objects using the malloc/free functions that have been set with cJSON_InitHooks */
+CJSON_PUBLIC(void *) cJSON_malloc(size_t size);
+CJSON_PUBLIC(void) cJSON_free(void *object);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
+
diff --git a/src/external/stb_image.h b/src/external/stb_image.h
new file mode 100644
index 0000000..a6202a3
--- /dev/null
+++ b/src/external/stb_image.h
@@ -0,0 +1,7547 @@
+/* stb_image - v2.22 - public domain image loader - http://nothings.org/stb
+                                  no warranty implied; use at your own risk
+
+   Do this:
+      #define STB_IMAGE_IMPLEMENTATION
+   before you include this file in *one* C or C++ file to create the implementation.
+
+   // i.e. it should look like this:
+   #include ...
+   #include ...
+   #include ...
+   #define STB_IMAGE_IMPLEMENTATION
+   #include "stb_image.h"
+
+   You can #define STBI_ASSERT(x) before the #include to avoid using assert.h.
+   And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free
+
+
+   QUICK NOTES:
+      Primarily of interest to game developers and other people who can
+          avoid problematic images and only need the trivial interface
+
+      JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib)
+      PNG 1/2/4/8/16-bit-per-channel
+
+      TGA (not sure what subset, if a subset)
+      BMP non-1bpp, non-RLE
+      PSD (composited view only, no extra channels, 8/16 bit-per-channel)
+
+      GIF (*comp always reports as 4-channel)
+      HDR (radiance rgbE format)
+      PIC (Softimage PIC)
+      PNM (PPM and PGM binary only)
+
+      Animated GIF still needs a proper API, but here's one way to do it:
+          http://gist.github.com/urraka/685d9a6340b26b830d49
+
+      - decode from memory or through FILE (define STBI_NO_STDIO to remove code)
+      - decode from arbitrary I/O callbacks
+      - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON)
+
+   Full documentation under "DOCUMENTATION" below.
+
+
+LICENSE
+
+  See end of file for license information.
+
+RECENT REVISION HISTORY:
+
+      2.22  (2019-03-04) gif fixes, fix warnings
+      2.21  (2019-02-25) fix typo in comment
+      2.20  (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs 
+      2.19  (2018-02-11) fix warning
+      2.18  (2018-01-30) fix warnings
+      2.17  (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings
+      2.16  (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes
+      2.15  (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC
+      2.14  (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+      2.13  (2016-12-04) experimental 16-bit API, only for PNG so far; fixes
+      2.12  (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+      2.11  (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64
+                         RGB-format JPEG; remove white matting in PSD;
+                         allocate large structures on the stack;
+                         correct channel count for PNG & BMP
+      2.10  (2016-01-22) avoid warning introduced in 2.09
+      2.09  (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED
+
+   See end of file for full revision history.
+
+
+ ============================    Contributors    =========================
+
+ Image formats                          Extensions, features
+    Sean Barrett (jpeg, png, bmp)          Jetro Lauha (stbi_info)
+    Nicolas Schulz (hdr, psd)              Martin "SpartanJ" Golini (stbi_info)
+    Jonathan Dummer (tga)                  James "moose2000" Brown (iPhone PNG)
+    Jean-Marc Lienher (gif)                Ben "Disch" Wenger (io callbacks)
+    Tom Seddon (pic)                       Omar Cornut (1/2/4-bit PNG)
+    Thatcher Ulrich (psd)                  Nicolas Guillemot (vertical flip)
+    Ken Miller (pgm, ppm)                  Richard Mitton (16-bit PSD)
+    github:urraka (animated gif)           Junggon Kim (PNM comments)
+    Christopher Forseth (animated gif)     Daniel Gibson (16-bit TGA)
+                                           socks-the-fox (16-bit PNG)
+                                           Jeremy Sawicki (handle all ImageNet JPGs)
+ Optimizations & bugfixes                  Mikhail Morozov (1-bit BMP)
+    Fabian "ryg" Giesen                    Anael Seghezzi (is-16-bit query)
+    Arseny Kapoulkine
+    John-Mark Allen
+    Carmelo J Fdez-Aguera
+
+ Bug & warning fixes
+    Marc LeBlanc            David Woo          Guillaume George   Martins Mozeiko
+    Christpher Lloyd        Jerry Jansson      Joseph Thomson     Phil Jordan
+    Dave Moore              Roy Eltham         Hayaki Saito       Nathan Reed
+    Won Chun                Luke Graham        Johan Duparc       Nick Verigakis
+    the Horde3D community   Thomas Ruf         Ronny Chevalier    github:rlyeh
+    Janez Zemva             John Bartholomew   Michal Cichon      github:romigrou
+    Jonathan Blow           Ken Hamada         Tero Hanninen      github:svdijk
+    Laurent Gomila          Cort Stratton      Sergio Gonzalez    github:snagar
+    Aruelien Pocheville     Thibault Reuille   Cass Everitt       github:Zelex
+    Ryamond Barbiero        Paul Du Bois       Engin Manap        github:grim210
+    Aldo Culquicondor       Philipp Wiesemann  Dale Weiler        github:sammyhw
+    Oriol Ferrer Mesia      Josh Tobin         Matthew Gregan     github:phprus
+    Julian Raschke          Gregory Mullen     Baldur Karlsson    github:poppolopoppo
+    Christian Floisand      Kevin Schmidt      JR Smith           github:darealshinji
+    Blazej Dariusz Roszkowski                                     github:Michaelangel007
+*/
+
+#ifndef STBI_INCLUDE_STB_IMAGE_H
+#define STBI_INCLUDE_STB_IMAGE_H
+
+// DOCUMENTATION
+//
+// Limitations:
+//    - no 12-bit-per-channel JPEG
+//    - no JPEGs with arithmetic coding
+//    - GIF always returns *comp=4
+//
+// Basic usage (see HDR discussion below for HDR usage):
+//    int x,y,n;
+//    unsigned char *data = stbi_load(filename, &x, &y, &n, 0);
+//    // ... process data if not NULL ...
+//    // ... x = width, y = height, n = # 8-bit components per pixel ...
+//    // ... replace '0' with '1'..'4' to force that many components per pixel
+//    // ... but 'n' will always be the number that it would have been if you said 0
+//    stbi_image_free(data)
+//
+// Standard parameters:
+//    int *x                 -- outputs image width in pixels
+//    int *y                 -- outputs image height in pixels
+//    int *channels_in_file  -- outputs # of image components in image file
+//    int desired_channels   -- if non-zero, # of image components requested in result
+//
+// The return value from an image loader is an 'unsigned char *' which points
+// to the pixel data, or NULL on an allocation failure or if the image is
+// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels,
+// with each pixel consisting of N interleaved 8-bit components; the first
+// pixel pointed to is top-left-most in the image. There is no padding between
+// image scanlines or between pixels, regardless of format. The number of
+// components N is 'desired_channels' if desired_channels is non-zero, or
+// *channels_in_file otherwise. If desired_channels is non-zero,
+// *channels_in_file has the number of components that _would_ have been
+// output otherwise. E.g. if you set desired_channels to 4, you will always
+// get RGBA output, but you can check *channels_in_file to see if it's trivially
+// opaque because e.g. there were only 3 channels in the source image.
+//
+// An output image with N components has the following components interleaved
+// in this order in each pixel:
+//
+//     N=#comp     components
+//       1           grey
+//       2           grey, alpha
+//       3           red, green, blue
+//       4           red, green, blue, alpha
+//
+// If image loading fails for any reason, the return value will be NULL,
+// and *x, *y, *channels_in_file will be unchanged. The function
+// stbi_failure_reason() can be queried for an extremely brief, end-user
+// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS
+// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly
+// more user-friendly ones.
+//
+// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.
+//
+// ===========================================================================
+//
+// UNICODE:
+//
+//   If compiling for Windows and you wish to use Unicode filenames, compile
+//   with
+//       #define STBI_WINDOWS_UTF8
+//   and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert
+//   Windows wchar_t filenames to utf8.
+//
+// ===========================================================================
+//
+// Philosophy
+//
+// stb libraries are designed with the following priorities:
+//
+//    1. easy to use
+//    2. easy to maintain
+//    3. good performance
+//
+// Sometimes I let "good performance" creep up in priority over "easy to maintain",
+// and for best performance I may provide less-easy-to-use APIs that give higher
+// performance, in addition to the easy-to-use ones. Nevertheless, it's important
+// to keep in mind that from the standpoint of you, a client of this library,
+// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all.
+//
+// Some secondary priorities arise directly from the first two, some of which
+// provide more explicit reasons why performance can't be emphasized.
+//
+//    - Portable ("ease of use")
+//    - Small source code footprint ("easy to maintain")
+//    - No dependencies ("ease of use")
+//
+// ===========================================================================
+//
+// I/O callbacks
+//
+// I/O callbacks allow you to read from arbitrary sources, like packaged
+// files or some other source. Data read from callbacks are processed
+// through a small internal buffer (currently 128 bytes) to try to reduce
+// overhead.
+//
+// The three functions you must define are "read" (reads some bytes of data),
+// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end).
+//
+// ===========================================================================
+//
+// SIMD support
+//
+// The JPEG decoder will try to automatically use SIMD kernels on x86 when
+// supported by the compiler. For ARM Neon support, you must explicitly
+// request it.
+//
+// (The old do-it-yourself SIMD API is no longer supported in the current
+// code.)
+//
+// On x86, SSE2 will automatically be used when available based on a run-time
+// test; if not, the generic C versions are used as a fall-back. On ARM targets,
+// the typical path is to have separate builds for NEON and non-NEON devices
+// (at least this is true for iOS and Android). Therefore, the NEON support is
+// toggled by a build flag: define STBI_NEON to get NEON loops.
+//
+// If for some reason you do not want to use any of SIMD code, or if
+// you have issues compiling it, you can disable it entirely by
+// defining STBI_NO_SIMD.
+//
+// ===========================================================================
+//
+// HDR image support   (disable by defining STBI_NO_HDR)
+//
+// stb_image supports loading HDR images in general, and currently the Radiance
+// .HDR file format specifically. You can still load any file through the existing
+// interface; if you attempt to load an HDR file, it will be automatically remapped
+// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1;
+// both of these constants can be reconfigured through this interface:
+//
+//     stbi_hdr_to_ldr_gamma(2.2f);
+//     stbi_hdr_to_ldr_scale(1.0f);
+//
+// (note, do not use _inverse_ constants; stbi_image will invert them
+// appropriately).
+//
+// Additionally, there is a new, parallel interface for loading files as
+// (linear) floats to preserve the full dynamic range:
+//
+//    float *data = stbi_loadf(filename, &x, &y, &n, 0);
+//
+// If you load LDR images through this interface, those images will
+// be promoted to floating point values, run through the inverse of
+// constants corresponding to the above:
+//
+//     stbi_ldr_to_hdr_scale(1.0f);
+//     stbi_ldr_to_hdr_gamma(2.2f);
+//
+// Finally, given a filename (or an open file or memory block--see header
+// file for details) containing image data, you can query for the "most
+// appropriate" interface to use (that is, whether the image is HDR or
+// not), using:
+//
+//     stbi_is_hdr(char *filename);
+//
+// ===========================================================================
+//
+// iPhone PNG support:
+//
+// By default we convert iphone-formatted PNGs back to RGB, even though
+// they are internally encoded differently. You can disable this conversion
+// by calling stbi_convert_iphone_png_to_rgb(0), in which case
+// you will always just get the native iphone "format" through (which
+// is BGR stored in RGB).
+//
+// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per
+// pixel to remove any premultiplied alpha *only* if the image file explicitly
+// says there's premultiplied data (currently only happens in iPhone images,
+// and only if iPhone convert-to-rgb processing is on).
+//
+// ===========================================================================
+//
+// ADDITIONAL CONFIGURATION
+//
+//  - You can suppress implementation of any of the decoders to reduce
+//    your code footprint by #defining one or more of the following
+//    symbols before creating the implementation.
+//
+//        STBI_NO_JPEG
+//        STBI_NO_PNG
+//        STBI_NO_BMP
+//        STBI_NO_PSD
+//        STBI_NO_TGA
+//        STBI_NO_GIF
+//        STBI_NO_HDR
+//        STBI_NO_PIC
+//        STBI_NO_PNM   (.ppm and .pgm)
+//
+//  - You can request *only* certain decoders and suppress all other ones
+//    (this will be more forward-compatible, as addition of new decoders
+//    doesn't require you to disable them explicitly):
+//
+//        STBI_ONLY_JPEG
+//        STBI_ONLY_PNG
+//        STBI_ONLY_BMP
+//        STBI_ONLY_PSD
+//        STBI_ONLY_TGA
+//        STBI_ONLY_GIF
+//        STBI_ONLY_HDR
+//        STBI_ONLY_PIC
+//        STBI_ONLY_PNM   (.ppm and .pgm)
+//
+//   - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still
+//     want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB
+//
+
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif // STBI_NO_STDIO
+
+#define STBI_VERSION 1
+
+enum
+{
+   STBI_default = 0, // only used for desired_channels
+
+   STBI_grey       = 1,
+   STBI_grey_alpha = 2,
+   STBI_rgb        = 3,
+   STBI_rgb_alpha  = 4
+};
+
+#include <stdlib.h>
+typedef unsigned char stbi_uc;
+typedef unsigned short stbi_us;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef STBIDEF
+#ifdef STB_IMAGE_STATIC
+#define STBIDEF static
+#else
+#define STBIDEF extern
+#endif
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// PRIMARY API - works on images of any type
+//
+
+//
+// load image by filename, open file, or memory buffer
+//
+
+typedef struct
+{
+   int      (*read)  (void *user,char *data,int size);   // fill 'data' with 'size' bytes.  return number of bytes actually read
+   void     (*skip)  (void *user,int n);                 // skip the next 'n' bytes, or 'unget' the last -n bytes if negative
+   int      (*eof)   (void *user);                       // returns nonzero if we are at end of file/data
+} stbi_io_callbacks;
+
+////////////////////////////////////
+//
+// 8-bits-per-channel interface
+//
+
+STBIDEF stbi_uc *stbi_load_from_memory   (stbi_uc           const *buffer, int len   , int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk  , void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_uc *stbi_load            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+// for stbi_load_from_file, file pointer is left pointing immediately after image
+#endif
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+#endif
+
+#ifdef STBI_WINDOWS_UTF8
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input);
+#endif
+
+////////////////////////////////////
+//
+// 16-bits-per-channel interface
+//
+
+STBIDEF stbi_us *stbi_load_16_from_memory   (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
+#ifndef STBI_NO_STDIO
+STBIDEF stbi_us *stbi_load_16          (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+#endif
+
+////////////////////////////////////
+//
+// float-per-channel interface
+//
+#ifndef STBI_NO_LINEAR
+   STBIDEF float *stbi_loadf_from_memory     (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+   STBIDEF float *stbi_loadf_from_callbacks  (stbi_io_callbacks const *clbk, void *user, int *x, int *y,  int *channels_in_file, int desired_channels);
+
+   #ifndef STBI_NO_STDIO
+   STBIDEF float *stbi_loadf            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+   STBIDEF float *stbi_loadf_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+   #endif
+#endif
+
+#ifndef STBI_NO_HDR
+   STBIDEF void   stbi_hdr_to_ldr_gamma(float gamma);
+   STBIDEF void   stbi_hdr_to_ldr_scale(float scale);
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_LINEAR
+   STBIDEF void   stbi_ldr_to_hdr_gamma(float gamma);
+   STBIDEF void   stbi_ldr_to_hdr_scale(float scale);
+#endif // STBI_NO_LINEAR
+
+// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR
+STBIDEF int    stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+STBIDEF int    stbi_is_hdr_from_memory(stbi_uc const *buffer, int len);
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_is_hdr          (char const *filename);
+STBIDEF int      stbi_is_hdr_from_file(FILE *f);
+#endif // STBI_NO_STDIO
+
+
+// get a VERY brief reason for failure
+// NOT THREADSAFE
+STBIDEF const char *stbi_failure_reason  (void);
+
+// free the loaded image -- this is just free()
+STBIDEF void     stbi_image_free      (void *retval_from_stbi_load);
+
+// get image dimensions & components without fully decoding
+STBIDEF int      stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp);
+STBIDEF int      stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp);
+STBIDEF int      stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len);
+STBIDEF int      stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user);
+
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_info               (char const *filename,     int *x, int *y, int *comp);
+STBIDEF int      stbi_info_from_file     (FILE *f,                  int *x, int *y, int *comp);
+STBIDEF int      stbi_is_16_bit          (char const *filename);
+STBIDEF int      stbi_is_16_bit_from_file(FILE *f);
+#endif
+
+
+
+// for image formats that explicitly notate that they have premultiplied alpha,
+// we just return the colors as stored in the file. set this flag to force
+// unpremultiplication. results are undefined if the unpremultiply overflow.
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply);
+
+// indicate whether we should process iphone images back to canonical format,
+// or just pass them through "as-is"
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert);
+
+// flip the image vertically, so the first pixel in the output array is the bottom left
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip);
+
+// ZLIB client - used by PNG, available for other purposes
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);
+STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int   stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);
+STBIDEF int   stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+//
+//
+////   end header file   /////////////////////////////////////////////////////
+#endif // STBI_INCLUDE_STB_IMAGE_H
+
+#ifdef STB_IMAGE_IMPLEMENTATION
+
+#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \
+  || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \
+  || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \
+  || defined(STBI_ONLY_ZLIB)
+   #ifndef STBI_ONLY_JPEG
+   #define STBI_NO_JPEG
+   #endif
+   #ifndef STBI_ONLY_PNG
+   #define STBI_NO_PNG
+   #endif
+   #ifndef STBI_ONLY_BMP
+   #define STBI_NO_BMP
+   #endif
+   #ifndef STBI_ONLY_PSD
+   #define STBI_NO_PSD
+   #endif
+   #ifndef STBI_ONLY_TGA
+   #define STBI_NO_TGA
+   #endif
+   #ifndef STBI_ONLY_GIF
+   #define STBI_NO_GIF
+   #endif
+   #ifndef STBI_ONLY_HDR
+   #define STBI_NO_HDR
+   #endif
+   #ifndef STBI_ONLY_PIC
+   #define STBI_NO_PIC
+   #endif
+   #ifndef STBI_ONLY_PNM
+   #define STBI_NO_PNM
+   #endif
+#endif
+
+#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB)
+#define STBI_NO_ZLIB
+#endif
+
+
+#include <stdarg.h>
+#include <stddef.h> // ptrdiff_t on osx
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+#include <math.h>  // ldexp, pow
+#endif
+
+#ifndef STBI_NO_STDIO
+#include <stdio.h>
+#endif
+
+#ifndef STBI_ASSERT
+#include <assert.h>
+#define STBI_ASSERT(x) assert(x)
+#endif
+
+#ifdef __cplusplus
+#define STBI_EXTERN extern "C"
+#else
+#define STBI_EXTERN extern
+#endif
+
+
+#ifndef _MSC_VER
+   #ifdef __cplusplus
+   #define stbi_inline inline
+   #else
+   #define stbi_inline
+   #endif
+#else
+   #define stbi_inline __forceinline
+#endif
+
+
+#ifdef _MSC_VER
+typedef unsigned short stbi__uint16;
+typedef   signed short stbi__int16;
+typedef unsigned int   stbi__uint32;
+typedef   signed int   stbi__int32;
+#else
+#include <stdint.h>
+typedef uint16_t stbi__uint16;
+typedef int16_t  stbi__int16;
+typedef uint32_t stbi__uint32;
+typedef int32_t  stbi__int32;
+#endif
+
+// should produce compiler error if size is wrong
+typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1];
+
+#ifdef _MSC_VER
+#define STBI_NOTUSED(v)  (void)(v)
+#else
+#define STBI_NOTUSED(v)  (void)sizeof(v)
+#endif
+
+#ifdef _MSC_VER
+#define STBI_HAS_LROTL
+#endif
+
+#ifdef STBI_HAS_LROTL
+   #define stbi_lrot(x,y)  _lrotl(x,y)
+#else
+   #define stbi_lrot(x,y)  (((x) << (y)) | ((x) >> (32 - (y))))
+#endif
+
+#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED))
+// ok
+#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED)
+// ok
+#else
+#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)."
+#endif
+
+#ifndef STBI_MALLOC
+#define STBI_MALLOC(sz)           malloc(sz)
+#define STBI_REALLOC(p,newsz)     realloc(p,newsz)
+#define STBI_FREE(p)              free(p)
+#endif
+
+#ifndef STBI_REALLOC_SIZED
+#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz)
+#endif
+
+// x86/x64 detection
+#if defined(__x86_64__) || defined(_M_X64)
+#define STBI__X64_TARGET
+#elif defined(__i386) || defined(_M_IX86)
+#define STBI__X86_TARGET
+#endif
+
+#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD)
+// gcc doesn't support sse2 intrinsics unless you compile with -msse2,
+// which in turn means it gets to use SSE2 everywhere. This is unfortunate,
+// but previous attempts to provide the SSE2 functions with runtime
+// detection caused numerous issues. The way architecture extensions are
+// exposed in GCC/Clang is, sadly, not really suited for one-file libs.
+// New behavior: if compiled with -msse2, we use SSE2 without any
+// detection; if not, we don't use it at all.
+#define STBI_NO_SIMD
+#endif
+
+#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD)
+// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET
+//
+// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the
+// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant.
+// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not
+// simultaneously enabling "-mstackrealign".
+//
+// See https://github.com/nothings/stb/issues/81 for more information.
+//
+// So default to no SSE2 on 32-bit MinGW. If you've read this far and added
+// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2.
+#define STBI_NO_SIMD
+#endif
+
+#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET))
+#define STBI_SSE2
+#include <emmintrin.h>
+
+#ifdef _MSC_VER
+
+#if _MSC_VER >= 1400  // not VC6
+#include <intrin.h> // __cpuid
+static int stbi__cpuid3(void)
+{
+   int info[4];
+   __cpuid(info,1);
+   return info[3];
+}
+#else
+static int stbi__cpuid3(void)
+{
+   int res;
+   __asm {
+      mov  eax,1
+      cpuid
+      mov  res,edx
+   }
+   return res;
+}
+#endif
+
+#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+   int info3 = stbi__cpuid3();
+   return ((info3 >> 26) & 1) != 0;
+}
+#endif
+
+#else // assume GCC-style if not VC++
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+
+#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2)
+static int stbi__sse2_available(void)
+{
+   // If we're even attempting to compile this on GCC/Clang, that means
+   // -msse2 is on, which means the compiler is allowed to use SSE2
+   // instructions at will, and so are we.
+   return 1;
+}
+#endif
+
+#endif
+#endif
+
+// ARM NEON
+#if defined(STBI_NO_SIMD) && defined(STBI_NEON)
+#undef STBI_NEON
+#endif
+
+#ifdef STBI_NEON
+#include <arm_neon.h>
+// assume GCC or Clang on ARM targets
+#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
+#endif
+
+#ifndef STBI_SIMD_ALIGN
+#define STBI_SIMD_ALIGN(type, name) type name
+#endif
+
+///////////////////////////////////////////////
+//
+//  stbi__context struct and start_xxx functions
+
+// stbi__context structure is our basic context used by all images, so it
+// contains all the IO context, plus some basic image information
+typedef struct
+{
+   stbi__uint32 img_x, img_y;
+   int img_n, img_out_n;
+
+   stbi_io_callbacks io;
+   void *io_user_data;
+
+   int read_from_callbacks;
+   int buflen;
+   stbi_uc buffer_start[128];
+
+   stbi_uc *img_buffer, *img_buffer_end;
+   stbi_uc *img_buffer_original, *img_buffer_original_end;
+} stbi__context;
+
+
+static void stbi__refill_buffer(stbi__context *s);
+
+// initialize a memory-decode context
+static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len)
+{
+   s->io.read = NULL;
+   s->read_from_callbacks = 0;
+   s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer;
+   s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len;
+}
+
+// initialize a callback-based context
+static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)
+{
+   s->io = *c;
+   s->io_user_data = user;
+   s->buflen = sizeof(s->buffer_start);
+   s->read_from_callbacks = 1;
+   s->img_buffer_original = s->buffer_start;
+   stbi__refill_buffer(s);
+   s->img_buffer_original_end = s->img_buffer_end;
+}
+
+#ifndef STBI_NO_STDIO
+
+static int stbi__stdio_read(void *user, char *data, int size)
+{
+   return (int) fread(data,1,size,(FILE*) user);
+}
+
+static void stbi__stdio_skip(void *user, int n)
+{
+   fseek((FILE*) user, n, SEEK_CUR);
+}
+
+static int stbi__stdio_eof(void *user)
+{
+   return feof((FILE*) user);
+}
+
+static stbi_io_callbacks stbi__stdio_callbacks =
+{
+   stbi__stdio_read,
+   stbi__stdio_skip,
+   stbi__stdio_eof,
+};
+
+static void stbi__start_file(stbi__context *s, FILE *f)
+{
+   stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);
+}
+
+//static void stop_file(stbi__context *s) { }
+
+#endif // !STBI_NO_STDIO
+
+static void stbi__rewind(stbi__context *s)
+{
+   // conceptually rewind SHOULD rewind to the beginning of the stream,
+   // but we just rewind to the beginning of the initial buffer, because
+   // we only use it after doing 'test', which only ever looks at at most 92 bytes
+   s->img_buffer = s->img_buffer_original;
+   s->img_buffer_end = s->img_buffer_original_end;
+}
+
+enum
+{
+   STBI_ORDER_RGB,
+   STBI_ORDER_BGR
+};
+
+typedef struct
+{
+   int bits_per_channel;
+   int num_channels;
+   int channel_order;
+} stbi__result_info;
+
+#ifndef STBI_NO_JPEG
+static int      stbi__jpeg_test(stbi__context *s);
+static void    *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNG
+static int      stbi__png_test(stbi__context *s);
+static void    *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__png_info(stbi__context *s, int *x, int *y, int *comp);
+static int      stbi__png_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_BMP
+static int      stbi__bmp_test(stbi__context *s);
+static void    *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_TGA
+static int      stbi__tga_test(stbi__context *s);
+static void    *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__tga_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PSD
+static int      stbi__psd_test(stbi__context *s);
+static void    *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc);
+static int      stbi__psd_info(stbi__context *s, int *x, int *y, int *comp);
+static int      stbi__psd_is16(stbi__context *s);
+#endif
+
+#ifndef STBI_NO_HDR
+static int      stbi__hdr_test(stbi__context *s);
+static float   *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PIC
+static int      stbi__pic_test(stbi__context *s);
+static void    *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__pic_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_GIF
+static int      stbi__gif_test(stbi__context *s);
+static void    *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static void    *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp);
+static int      stbi__gif_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+#ifndef STBI_NO_PNM
+static int      stbi__pnm_test(stbi__context *s);
+static void    *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri);
+static int      stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp);
+#endif
+
+// this is not threadsafe
+static const char *stbi__g_failure_reason;
+
+STBIDEF const char *stbi_failure_reason(void)
+{
+   return stbi__g_failure_reason;
+}
+
+static int stbi__err(const char *str)
+{
+   stbi__g_failure_reason = str;
+   return 0;
+}
+
+static void *stbi__malloc(size_t size)
+{
+    return STBI_MALLOC(size);
+}
+
+// stb_image uses ints pervasively, including for offset calculations.
+// therefore the largest decoded image size we can support with the
+// current code, even on 64-bit targets, is INT_MAX. this is not a
+// significant limitation for the intended use case.
+//
+// we do, however, need to make sure our size calculations don't
+// overflow. hence a few helper functions for size calculations that
+// multiply integers together, making sure that they're non-negative
+// and no overflow occurs.
+
+// return 1 if the sum is valid, 0 on overflow.
+// negative terms are considered invalid.
+static int stbi__addsizes_valid(int a, int b)
+{
+   if (b < 0) return 0;
+   // now 0 <= b <= INT_MAX, hence also
+   // 0 <= INT_MAX - b <= INTMAX.
+   // And "a + b <= INT_MAX" (which might overflow) is the
+   // same as a <= INT_MAX - b (no overflow)
+   return a <= INT_MAX - b;
+}
+
+// returns 1 if the product is valid, 0 on overflow.
+// negative factors are considered invalid.
+static int stbi__mul2sizes_valid(int a, int b)
+{
+   if (a < 0 || b < 0) return 0;
+   if (b == 0) return 1; // mul-by-0 is always safe
+   // portable way to check for no overflows in a*b
+   return a <= INT_MAX/b;
+}
+
+// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad2sizes_valid(int a, int b, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add);
+}
+
+// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow
+static int stbi__mad3sizes_valid(int a, int b, int c, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+      stbi__addsizes_valid(a*b*c, add);
+}
+
+// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add)
+{
+   return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
+      stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add);
+}
+#endif
+
+// mallocs with size overflow checking
+static void *stbi__malloc_mad2(int a, int b, int add)
+{
+   if (!stbi__mad2sizes_valid(a, b, add)) return NULL;
+   return stbi__malloc(a*b + add);
+}
+
+static void *stbi__malloc_mad3(int a, int b, int c, int add)
+{
+   if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL;
+   return stbi__malloc(a*b*c + add);
+}
+
+#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)
+static void *stbi__malloc_mad4(int a, int b, int c, int d, int add)
+{
+   if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL;
+   return stbi__malloc(a*b*c*d + add);
+}
+#endif
+
+// stbi__err - error
+// stbi__errpf - error returning pointer to float
+// stbi__errpuc - error returning pointer to unsigned char
+
+#ifdef STBI_NO_FAILURE_STRINGS
+   #define stbi__err(x,y)  0
+#elif defined(STBI_FAILURE_USERMSG)
+   #define stbi__err(x,y)  stbi__err(y)
+#else
+   #define stbi__err(x,y)  stbi__err(x)
+#endif
+
+#define stbi__errpf(x,y)   ((float *)(size_t) (stbi__err(x,y)?NULL:NULL))
+#define stbi__errpuc(x,y)  ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))
+
+STBIDEF void stbi_image_free(void *retval_from_stbi_load)
+{
+   STBI_FREE(retval_from_stbi_load);
+}
+
+#ifndef STBI_NO_LINEAR
+static float   *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp);
+#endif
+
+#ifndef STBI_NO_HDR
+static stbi_uc *stbi__hdr_to_ldr(float   *data, int x, int y, int comp);
+#endif
+
+static int stbi__vertically_flip_on_load = 0;
+
+STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip)
+{
+    stbi__vertically_flip_on_load = flag_true_if_should_flip;
+}
+
+static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+   memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields
+   ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed
+   ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order
+   ri->num_channels = 0;
+
+   #ifndef STBI_NO_JPEG
+   if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_PNG
+   if (stbi__png_test(s))  return stbi__png_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_BMP
+   if (stbi__bmp_test(s))  return stbi__bmp_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_GIF
+   if (stbi__gif_test(s))  return stbi__gif_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_test(s))  return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc);
+   #endif
+   #ifndef STBI_NO_PIC
+   if (stbi__pic_test(s))  return stbi__pic_load(s,x,y,comp,req_comp, ri);
+   #endif
+   #ifndef STBI_NO_PNM
+   if (stbi__pnm_test(s))  return stbi__pnm_load(s,x,y,comp,req_comp, ri);
+   #endif
+
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_test(s)) {
+      float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri);
+      return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp);
+   }
+   #endif
+
+   #ifndef STBI_NO_TGA
+   // test tga last because it's a crappy test!
+   if (stbi__tga_test(s))
+      return stbi__tga_load(s,x,y,comp,req_comp, ri);
+   #endif
+
+   return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels)
+{
+   int i;
+   int img_len = w * h * channels;
+   stbi_uc *reduced;
+
+   reduced = (stbi_uc *) stbi__malloc(img_len);
+   if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory");
+
+   for (i = 0; i < img_len; ++i)
+      reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling
+
+   STBI_FREE(orig);
+   return reduced;
+}
+
+static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels)
+{
+   int i;
+   int img_len = w * h * channels;
+   stbi__uint16 *enlarged;
+
+   enlarged = (stbi__uint16 *) stbi__malloc(img_len*2);
+   if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+
+   for (i = 0; i < img_len; ++i)
+      enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff
+
+   STBI_FREE(orig);
+   return enlarged;
+}
+
+static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel)
+{
+   int row;
+   size_t bytes_per_row = (size_t)w * bytes_per_pixel;
+   stbi_uc temp[2048];
+   stbi_uc *bytes = (stbi_uc *)image;
+
+   for (row = 0; row < (h>>1); row++) {
+      stbi_uc *row0 = bytes + row*bytes_per_row;
+      stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row;
+      // swap row0 with row1
+      size_t bytes_left = bytes_per_row;
+      while (bytes_left) {
+         size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);
+         memcpy(temp, row0, bytes_copy);
+         memcpy(row0, row1, bytes_copy);
+         memcpy(row1, temp, bytes_copy);
+         row0 += bytes_copy;
+         row1 += bytes_copy;
+         bytes_left -= bytes_copy;
+      }
+   }
+}
+
+#ifndef STBI_NO_GIF
+static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel)
+{
+   int slice;
+   int slice_size = w * h * bytes_per_pixel;
+
+   stbi_uc *bytes = (stbi_uc *)image;
+   for (slice = 0; slice < z; ++slice) {
+      stbi__vertical_flip(bytes, w, h, bytes_per_pixel); 
+      bytes += slice_size; 
+   }
+}
+#endif
+
+static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__result_info ri;
+   void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8);
+
+   if (result == NULL)
+      return NULL;
+
+   if (ri.bits_per_channel != 8) {
+      STBI_ASSERT(ri.bits_per_channel == 16);
+      result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+      ri.bits_per_channel = 8;
+   }
+
+   // @TODO: move stbi__convert_format to here
+
+   if (stbi__vertically_flip_on_load) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc));
+   }
+
+   return (unsigned char *) result;
+}
+
+static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__result_info ri;
+   void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16);
+
+   if (result == NULL)
+      return NULL;
+
+   if (ri.bits_per_channel != 16) {
+      STBI_ASSERT(ri.bits_per_channel == 8);
+      result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp);
+      ri.bits_per_channel = 16;
+   }
+
+   // @TODO: move stbi__convert_format16 to here
+   // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision
+
+   if (stbi__vertically_flip_on_load) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16));
+   }
+
+   return (stbi__uint16 *) result;
+}
+
+#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR)
+static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)
+{
+   if (stbi__vertically_flip_on_load && result != NULL) {
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(float));
+   }
+}
+#endif
+
+#ifndef STBI_NO_STDIO
+
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide);
+STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default);
+#endif
+
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input)
+{
+	return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL);
+}
+#endif
+
+static FILE *stbi__fopen(char const *filename, char const *mode)
+{
+   FILE *f;
+#if defined(_MSC_VER) && defined(STBI_WINDOWS_UTF8)
+   wchar_t wMode[64];
+   wchar_t wFilename[1024];
+	if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)))
+      return 0;
+	
+	if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)))
+      return 0;
+
+#if _MSC_VER >= 1400
+	if (0 != _wfopen_s(&f, wFilename, wMode))
+		f = 0;
+#else
+   f = _wfopen(wFilename, wMode);
+#endif
+
+#elif defined(_MSC_VER) && _MSC_VER >= 1400
+   if (0 != fopen_s(&f, filename, mode))
+      f=0;
+#else
+   f = fopen(filename, mode);
+#endif
+   return f;
+}
+
+
+STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   unsigned char *result;
+   if (!f) return stbi__errpuc("can't fopen", "Unable to open file");
+   result = stbi_load_from_file(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   unsigned char *result;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+   if (result) {
+      // need to 'unget' all the characters in the IO buffer
+      fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+   }
+   return result;
+}
+
+STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__uint16 *result;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp);
+   if (result) {
+      // need to 'unget' all the characters in the IO buffer
+      fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
+   }
+   return result;
+}
+
+STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   stbi__uint16 *result;
+   if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file");
+   result = stbi_load_from_file_16(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+
+#endif //!STBI_NO_STDIO
+
+STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_GIF
+STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+   unsigned char *result;
+   stbi__context s; 
+   stbi__start_mem(&s,buffer,len); 
+   
+   result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp);
+   if (stbi__vertically_flip_on_load) {
+      stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); 
+   }
+
+   return result; 
+}
+#endif
+
+#ifndef STBI_NO_LINEAR
+static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)
+{
+   unsigned char *data;
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_test(s)) {
+      stbi__result_info ri;
+      float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri);
+      if (hdr_data)
+         stbi__float_postprocess(hdr_data,x,y,comp,req_comp);
+      return hdr_data;
+   }
+   #endif
+   data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp);
+   if (data)
+      return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp);
+   return stbi__errpf("unknown image type", "Image not of any known type, or corrupt");
+}
+
+STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp)
+{
+   float *result;
+   FILE *f = stbi__fopen(filename, "rb");
+   if (!f) return stbi__errpf("can't fopen", "Unable to open file");
+   result = stbi_loadf_from_file(f,x,y,comp,req_comp);
+   fclose(f);
+   return result;
+}
+
+STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)
+{
+   stbi__context s;
+   stbi__start_file(&s,f);
+   return stbi__loadf_main(&s,x,y,comp,req_comp);
+}
+#endif // !STBI_NO_STDIO
+
+#endif // !STBI_NO_LINEAR
+
+// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is
+// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always
+// reports false!
+
+STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len)
+{
+   #ifndef STBI_NO_HDR
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__hdr_test(&s);
+   #else
+   STBI_NOTUSED(buffer);
+   STBI_NOTUSED(len);
+   return 0;
+   #endif
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int      stbi_is_hdr          (char const *filename)
+{
+   FILE *f = stbi__fopen(filename, "rb");
+   int result=0;
+   if (f) {
+      result = stbi_is_hdr_from_file(f);
+      fclose(f);
+   }
+   return result;
+}
+
+STBIDEF int stbi_is_hdr_from_file(FILE *f)
+{
+   #ifndef STBI_NO_HDR
+   long pos = ftell(f);
+   int res;
+   stbi__context s;
+   stbi__start_file(&s,f);
+   res = stbi__hdr_test(&s);
+   fseek(f, pos, SEEK_SET);
+   return res;
+   #else
+   STBI_NOTUSED(f);
+   return 0;
+   #endif
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int      stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user)
+{
+   #ifndef STBI_NO_HDR
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);
+   return stbi__hdr_test(&s);
+   #else
+   STBI_NOTUSED(clbk);
+   STBI_NOTUSED(user);
+   return 0;
+   #endif
+}
+
+#ifndef STBI_NO_LINEAR
+static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f;
+
+STBIDEF void   stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; }
+STBIDEF void   stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; }
+#endif
+
+static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f;
+
+STBIDEF void   stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; }
+STBIDEF void   stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; }
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Common code used by all image loaders
+//
+
+enum
+{
+   STBI__SCAN_load=0,
+   STBI__SCAN_type,
+   STBI__SCAN_header
+};
+
+static void stbi__refill_buffer(stbi__context *s)
+{
+   int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);
+   if (n == 0) {
+      // at end of file, treat same as if from memory, but need to handle case
+      // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file
+      s->read_from_callbacks = 0;
+      s->img_buffer = s->buffer_start;
+      s->img_buffer_end = s->buffer_start+1;
+      *s->img_buffer = 0;
+   } else {
+      s->img_buffer = s->buffer_start;
+      s->img_buffer_end = s->buffer_start + n;
+   }
+}
+
+stbi_inline static stbi_uc stbi__get8(stbi__context *s)
+{
+   if (s->img_buffer < s->img_buffer_end)
+      return *s->img_buffer++;
+   if (s->read_from_callbacks) {
+      stbi__refill_buffer(s);
+      return *s->img_buffer++;
+   }
+   return 0;
+}
+
+stbi_inline static int stbi__at_eof(stbi__context *s)
+{
+   if (s->io.read) {
+      if (!(s->io.eof)(s->io_user_data)) return 0;
+      // if feof() is true, check if buffer = end
+      // special case: we've only got the special 0 character at the end
+      if (s->read_from_callbacks == 0) return 1;
+   }
+
+   return s->img_buffer >= s->img_buffer_end;
+}
+
+static void stbi__skip(stbi__context *s, int n)
+{
+   if (n < 0) {
+      s->img_buffer = s->img_buffer_end;
+      return;
+   }
+   if (s->io.read) {
+      int blen = (int) (s->img_buffer_end - s->img_buffer);
+      if (blen < n) {
+         s->img_buffer = s->img_buffer_end;
+         (s->io.skip)(s->io_user_data, n - blen);
+         return;
+      }
+   }
+   s->img_buffer += n;
+}
+
+static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n)
+{
+   if (s->io.read) {
+      int blen = (int) (s->img_buffer_end - s->img_buffer);
+      if (blen < n) {
+         int res, count;
+
+         memcpy(buffer, s->img_buffer, blen);
+
+         count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);
+         res = (count == (n-blen));
+         s->img_buffer = s->img_buffer_end;
+         return res;
+      }
+   }
+
+   if (s->img_buffer+n <= s->img_buffer_end) {
+      memcpy(buffer, s->img_buffer, n);
+      s->img_buffer += n;
+      return 1;
+   } else
+      return 0;
+}
+
+static int stbi__get16be(stbi__context *s)
+{
+   int z = stbi__get8(s);
+   return (z << 8) + stbi__get8(s);
+}
+
+static stbi__uint32 stbi__get32be(stbi__context *s)
+{
+   stbi__uint32 z = stbi__get16be(s);
+   return (z << 16) + stbi__get16be(s);
+}
+
+#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF)
+// nothing
+#else
+static int stbi__get16le(stbi__context *s)
+{
+   int z = stbi__get8(s);
+   return z + (stbi__get8(s) << 8);
+}
+#endif
+
+#ifndef STBI_NO_BMP
+static stbi__uint32 stbi__get32le(stbi__context *s)
+{
+   stbi__uint32 z = stbi__get16le(s);
+   return z + (stbi__get16le(s) << 16);
+}
+#endif
+
+#define STBI__BYTECAST(x)  ((stbi_uc) ((x) & 255))  // truncate int to byte without warnings
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+//  generic converter from built-in img_n to req_comp
+//    individual types do this automatically as much as possible (e.g. jpeg
+//    does all cases internally since it needs to colorspace convert anyway,
+//    and it never has alpha, so very few cases ). png can automatically
+//    interleave an alpha=255 channel, but falls back to this for other cases
+//
+//  assume data buffer is malloced, so malloc a new one and free that one
+//  only failure mode is malloc failing
+
+static stbi_uc stbi__compute_y(int r, int g, int b)
+{
+   return (stbi_uc) (((r*77) + (g*150) +  (29*b)) >> 8);
+}
+
+static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+   int i,j;
+   unsigned char *good;
+
+   if (req_comp == img_n) return data;
+   STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+   good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0);
+   if (good == NULL) {
+      STBI_FREE(data);
+      return stbi__errpuc("outofmem", "Out of memory");
+   }
+
+   for (j=0; j < (int) y; ++j) {
+      unsigned char *src  = data + j * x * img_n   ;
+      unsigned char *dest = good + j * x * req_comp;
+
+      #define STBI__COMBO(a,b)  ((a)*8+(b))
+      #define STBI__CASE(a,b)   case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+      // convert source image with img_n components to one with req_comp components;
+      // avoid switch per pixel, so use switch per scanline and massive macros
+      switch (STBI__COMBO(img_n, req_comp)) {
+         STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255;                                     } break;
+         STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0];                                  } break;
+         STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255;                     } break;
+         STBI__CASE(2,1) { dest[0]=src[0];                                                  } break;
+         STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0];                                  } break;
+         STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1];                  } break;
+         STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255;        } break;
+         STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255;    } break;
+         STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+         STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];                    } break;
+         default: STBI_ASSERT(0);
+      }
+      #undef STBI__CASE
+   }
+
+   STBI_FREE(data);
+   return good;
+}
+
+static stbi__uint16 stbi__compute_y_16(int r, int g, int b)
+{
+   return (stbi__uint16) (((r*77) + (g*150) +  (29*b)) >> 8);
+}
+
+static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y)
+{
+   int i,j;
+   stbi__uint16 *good;
+
+   if (req_comp == img_n) return data;
+   STBI_ASSERT(req_comp >= 1 && req_comp <= 4);
+
+   good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2);
+   if (good == NULL) {
+      STBI_FREE(data);
+      return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory");
+   }
+
+   for (j=0; j < (int) y; ++j) {
+      stbi__uint16 *src  = data + j * x * img_n   ;
+      stbi__uint16 *dest = good + j * x * req_comp;
+
+      #define STBI__COMBO(a,b)  ((a)*8+(b))
+      #define STBI__CASE(a,b)   case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
+      // convert source image with img_n components to one with req_comp components;
+      // avoid switch per pixel, so use switch per scanline and massive macros
+      switch (STBI__COMBO(img_n, req_comp)) {
+         STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff;                                     } break;
+         STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0];                                     } break;
+         STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff;                     } break;
+         STBI__CASE(2,1) { dest[0]=src[0];                                                     } break;
+         STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0];                                     } break;
+         STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1];                     } break;
+         STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff;        } break;
+         STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break;
+         STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]);                   } break;
+         STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break;
+         STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];                       } break;
+         default: STBI_ASSERT(0);
+      }
+      #undef STBI__CASE
+   }
+
+   STBI_FREE(data);
+   return good;
+}
+
+#ifndef STBI_NO_LINEAR
+static float   *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp)
+{
+   int i,k,n;
+   float *output;
+   if (!data) return NULL;
+   output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0);
+   if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); }
+   // compute number of non-alpha components
+   if (comp & 1) n = comp; else n = comp-1;
+   for (i=0; i < x*y; ++i) {
+      for (k=0; k < n; ++k) {
+         output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale);
+      }
+   }
+   if (n < comp) {
+      for (i=0; i < x*y; ++i) {
+         output[i*comp + n] = data[i*comp + n]/255.0f;
+      }
+   }
+   STBI_FREE(data);
+   return output;
+}
+#endif
+
+#ifndef STBI_NO_HDR
+#define stbi__float2int(x)   ((int) (x))
+static stbi_uc *stbi__hdr_to_ldr(float   *data, int x, int y, int comp)
+{
+   int i,k,n;
+   stbi_uc *output;
+   if (!data) return NULL;
+   output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0);
+   if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); }
+   // compute number of non-alpha components
+   if (comp & 1) n = comp; else n = comp-1;
+   for (i=0; i < x*y; ++i) {
+      for (k=0; k < n; ++k) {
+         float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f;
+         if (z < 0) z = 0;
+         if (z > 255) z = 255;
+         output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+      }
+      if (k < comp) {
+         float z = data[i*comp+k] * 255 + 0.5f;
+         if (z < 0) z = 0;
+         if (z > 255) z = 255;
+         output[i*comp + k] = (stbi_uc) stbi__float2int(z);
+      }
+   }
+   STBI_FREE(data);
+   return output;
+}
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+//  "baseline" JPEG/JFIF decoder
+//
+//    simple implementation
+//      - doesn't support delayed output of y-dimension
+//      - simple interface (only one output format: 8-bit interleaved RGB)
+//      - doesn't try to recover corrupt jpegs
+//      - doesn't allow partial loading, loading multiple at once
+//      - still fast on x86 (copying globals into locals doesn't help x86)
+//      - allocates lots of intermediate memory (full size of all components)
+//        - non-interleaved case requires this anyway
+//        - allows good upsampling (see next)
+//    high-quality
+//      - upsampled channels are bilinearly interpolated, even across blocks
+//      - quality integer IDCT derived from IJG's 'slow'
+//    performance
+//      - fast huffman; reasonable integer IDCT
+//      - some SIMD kernels for common paths on targets with SSE2/NEON
+//      - uses a lot of intermediate memory, could cache poorly
+
+#ifndef STBI_NO_JPEG
+
+// huffman decoding acceleration
+#define FAST_BITS   9  // larger handles more cases; smaller stomps less cache
+
+typedef struct
+{
+   stbi_uc  fast[1 << FAST_BITS];
+   // weirdly, repacking this into AoS is a 10% speed loss, instead of a win
+   stbi__uint16 code[256];
+   stbi_uc  values[256];
+   stbi_uc  size[257];
+   unsigned int maxcode[18];
+   int    delta[17];   // old 'firstsymbol' - old 'firstcode'
+} stbi__huffman;
+
+typedef struct
+{
+   stbi__context *s;
+   stbi__huffman huff_dc[4];
+   stbi__huffman huff_ac[4];
+   stbi__uint16 dequant[4][64];
+   stbi__int16 fast_ac[4][1 << FAST_BITS];
+
+// sizes for components, interleaved MCUs
+   int img_h_max, img_v_max;
+   int img_mcu_x, img_mcu_y;
+   int img_mcu_w, img_mcu_h;
+
+// definition of jpeg image component
+   struct
+   {
+      int id;
+      int h,v;
+      int tq;
+      int hd,ha;
+      int dc_pred;
+
+      int x,y,w2,h2;
+      stbi_uc *data;
+      void *raw_data, *raw_coeff;
+      stbi_uc *linebuf;
+      short   *coeff;   // progressive only
+      int      coeff_w, coeff_h; // number of 8x8 coefficient blocks
+   } img_comp[4];
+
+   stbi__uint32   code_buffer; // jpeg entropy-coded buffer
+   int            code_bits;   // number of valid bits
+   unsigned char  marker;      // marker seen while filling entropy buffer
+   int            nomore;      // flag if we saw a marker so must stop
+
+   int            progressive;
+   int            spec_start;
+   int            spec_end;
+   int            succ_high;
+   int            succ_low;
+   int            eob_run;
+   int            jfif;
+   int            app14_color_transform; // Adobe APP14 tag
+   int            rgb;
+
+   int scan_n, order[4];
+   int restart_interval, todo;
+
+// kernels
+   void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]);
+   void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step);
+   stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs);
+} stbi__jpeg;
+
+static int stbi__build_huffman(stbi__huffman *h, int *count)
+{
+   int i,j,k=0;
+   unsigned int code;
+   // build size list for each symbol (from JPEG spec)
+   for (i=0; i < 16; ++i)
+      for (j=0; j < count[i]; ++j)
+         h->size[k++] = (stbi_uc) (i+1);
+   h->size[k] = 0;
+
+   // compute actual symbols (from jpeg spec)
+   code = 0;
+   k = 0;
+   for(j=1; j <= 16; ++j) {
+      // compute delta to add to code to compute symbol id
+      h->delta[j] = k - code;
+      if (h->size[k] == j) {
+         while (h->size[k] == j)
+            h->code[k++] = (stbi__uint16) (code++);
+         if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG");
+      }
+      // compute largest code + 1 for this size, preshifted as needed later
+      h->maxcode[j] = code << (16-j);
+      code <<= 1;
+   }
+   h->maxcode[j] = 0xffffffff;
+
+   // build non-spec acceleration table; 255 is flag for not-accelerated
+   memset(h->fast, 255, 1 << FAST_BITS);
+   for (i=0; i < k; ++i) {
+      int s = h->size[i];
+      if (s <= FAST_BITS) {
+         int c = h->code[i] << (FAST_BITS-s);
+         int m = 1 << (FAST_BITS-s);
+         for (j=0; j < m; ++j) {
+            h->fast[c+j] = (stbi_uc) i;
+         }
+      }
+   }
+   return 1;
+}
+
+// build a table that decodes both magnitude and value of small ACs in
+// one go.
+static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h)
+{
+   int i;
+   for (i=0; i < (1 << FAST_BITS); ++i) {
+      stbi_uc fast = h->fast[i];
+      fast_ac[i] = 0;
+      if (fast < 255) {
+         int rs = h->values[fast];
+         int run = (rs >> 4) & 15;
+         int magbits = rs & 15;
+         int len = h->size[fast];
+
+         if (magbits && len + magbits <= FAST_BITS) {
+            // magnitude code followed by receive_extend code
+            int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits);
+            int m = 1 << (magbits - 1);
+            if (k < m) k += (~0U << magbits) + 1;
+            // if the result is small enough, we can fit it in fast_ac table
+            if (k >= -128 && k <= 127)
+               fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits));
+         }
+      }
+   }
+}
+
+static void stbi__grow_buffer_unsafe(stbi__jpeg *j)
+{
+   do {
+      unsigned int b = j->nomore ? 0 : stbi__get8(j->s);
+      if (b == 0xff) {
+         int c = stbi__get8(j->s);
+         while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes
+         if (c != 0) {
+            j->marker = (unsigned char) c;
+            j->nomore = 1;
+            return;
+         }
+      }
+      j->code_buffer |= b << (24 - j->code_bits);
+      j->code_bits += 8;
+   } while (j->code_bits <= 24);
+}
+
+// (1 << n) - 1
+static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535};
+
+// decode a jpeg huffman value from the bitstream
+stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h)
+{
+   unsigned int temp;
+   int c,k;
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+   // look at the top FAST_BITS and determine what symbol ID it is,
+   // if the code is <= FAST_BITS
+   c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+   k = h->fast[c];
+   if (k < 255) {
+      int s = h->size[k];
+      if (s > j->code_bits)
+         return -1;
+      j->code_buffer <<= s;
+      j->code_bits -= s;
+      return h->values[k];
+   }
+
+   // naive test is to shift the code_buffer down so k bits are
+   // valid, then test against maxcode. To speed this up, we've
+   // preshifted maxcode left so that it has (16-k) 0s at the
+   // end; in other words, regardless of the number of bits, it
+   // wants to be compared against something shifted to have 16;
+   // that way we don't need to shift inside the loop.
+   temp = j->code_buffer >> 16;
+   for (k=FAST_BITS+1 ; ; ++k)
+      if (temp < h->maxcode[k])
+         break;
+   if (k == 17) {
+      // error! code not found
+      j->code_bits -= 16;
+      return -1;
+   }
+
+   if (k > j->code_bits)
+      return -1;
+
+   // convert the huffman code to the symbol id
+   c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k];
+   STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]);
+
+   // convert the id to a symbol
+   j->code_bits -= k;
+   j->code_buffer <<= k;
+   return h->values[c];
+}
+
+// bias[n] = (-1<<n) + 1
+static const int stbi__jbias[16] = {0,-1,-3,-7,-15,-31,-63,-127,-255,-511,-1023,-2047,-4095,-8191,-16383,-32767};
+
+// combined JPEG 'receive' and JPEG 'extend', since baseline
+// always extends everything it receives.
+stbi_inline static int stbi__extend_receive(stbi__jpeg *j, int n)
+{
+   unsigned int k;
+   int sgn;
+   if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+
+   sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB
+   k = stbi_lrot(j->code_buffer, n);
+   STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask)));
+   j->code_buffer = k & ~stbi__bmask[n];
+   k &= stbi__bmask[n];
+   j->code_bits -= n;
+   return k + (stbi__jbias[n] & ~sgn);
+}
+
+// get some unsigned bits
+stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n)
+{
+   unsigned int k;
+   if (j->code_bits < n) stbi__grow_buffer_unsafe(j);
+   k = stbi_lrot(j->code_buffer, n);
+   j->code_buffer = k & ~stbi__bmask[n];
+   k &= stbi__bmask[n];
+   j->code_bits -= n;
+   return k;
+}
+
+stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j)
+{
+   unsigned int k;
+   if (j->code_bits < 1) stbi__grow_buffer_unsafe(j);
+   k = j->code_buffer;
+   j->code_buffer <<= 1;
+   --j->code_bits;
+   return k & 0x80000000;
+}
+
+// given a value that's at position X in the zigzag stream,
+// where does it appear in the 8x8 matrix coded as row-major?
+static const stbi_uc stbi__jpeg_dezigzag[64+15] =
+{
+    0,  1,  8, 16,  9,  2,  3, 10,
+   17, 24, 32, 25, 18, 11,  4,  5,
+   12, 19, 26, 33, 40, 48, 41, 34,
+   27, 20, 13,  6,  7, 14, 21, 28,
+   35, 42, 49, 56, 57, 50, 43, 36,
+   29, 22, 15, 23, 30, 37, 44, 51,
+   58, 59, 52, 45, 38, 31, 39, 46,
+   53, 60, 61, 54, 47, 55, 62, 63,
+   // let corrupt input sample past end
+   63, 63, 63, 63, 63, 63, 63, 63,
+   63, 63, 63, 63, 63, 63, 63
+};
+
+// decode one 64-entry block--
+static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant)
+{
+   int diff,dc,k;
+   int t;
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+   t = stbi__jpeg_huff_decode(j, hdc);
+   if (t < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+
+   // 0 all the ac values now so we can do it 32-bits at a time
+   memset(data,0,64*sizeof(data[0]));
+
+   diff = t ? stbi__extend_receive(j, t) : 0;
+   dc = j->img_comp[b].dc_pred + diff;
+   j->img_comp[b].dc_pred = dc;
+   data[0] = (short) (dc * dequant[0]);
+
+   // decode AC components, see JPEG spec
+   k = 1;
+   do {
+      unsigned int zig;
+      int c,r,s;
+      if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+      c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+      r = fac[c];
+      if (r) { // fast-AC path
+         k += (r >> 4) & 15; // run
+         s = r & 15; // combined length
+         j->code_buffer <<= s;
+         j->code_bits -= s;
+         // decode into unzigzag'd location
+         zig = stbi__jpeg_dezigzag[k++];
+         data[zig] = (short) ((r >> 8) * dequant[zig]);
+      } else {
+         int rs = stbi__jpeg_huff_decode(j, hac);
+         if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+         s = rs & 15;
+         r = rs >> 4;
+         if (s == 0) {
+            if (rs != 0xf0) break; // end block
+            k += 16;
+         } else {
+            k += r;
+            // decode into unzigzag'd location
+            zig = stbi__jpeg_dezigzag[k++];
+            data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]);
+         }
+      }
+   } while (k < 64);
+   return 1;
+}
+
+static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b)
+{
+   int diff,dc;
+   int t;
+   if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+   if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+
+   if (j->succ_high == 0) {
+      // first scan for DC coefficient, must be first
+      memset(data,0,64*sizeof(data[0])); // 0 all the ac values now
+      t = stbi__jpeg_huff_decode(j, hdc);
+      diff = t ? stbi__extend_receive(j, t) : 0;
+
+      dc = j->img_comp[b].dc_pred + diff;
+      j->img_comp[b].dc_pred = dc;
+      data[0] = (short) (dc << j->succ_low);
+   } else {
+      // refinement scan for DC coefficient
+      if (stbi__jpeg_get_bit(j))
+         data[0] += (short) (1 << j->succ_low);
+   }
+   return 1;
+}
+
+// @OPTIMIZE: store non-zigzagged during the decode passes,
+// and only de-zigzag when dequantizing
+static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac)
+{
+   int k;
+   if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG");
+
+   if (j->succ_high == 0) {
+      int shift = j->succ_low;
+
+      if (j->eob_run) {
+         --j->eob_run;
+         return 1;
+      }
+
+      k = j->spec_start;
+      do {
+         unsigned int zig;
+         int c,r,s;
+         if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);
+         c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);
+         r = fac[c];
+         if (r) { // fast-AC path
+            k += (r >> 4) & 15; // run
+            s = r & 15; // combined length
+            j->code_buffer <<= s;
+            j->code_bits -= s;
+            zig = stbi__jpeg_dezigzag[k++];
+            data[zig] = (short) ((r >> 8) << shift);
+         } else {
+            int rs = stbi__jpeg_huff_decode(j, hac);
+            if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+            s = rs & 15;
+            r = rs >> 4;
+            if (s == 0) {
+               if (r < 15) {
+                  j->eob_run = (1 << r);
+                  if (r)
+                     j->eob_run += stbi__jpeg_get_bits(j, r);
+                  --j->eob_run;
+                  break;
+               }
+               k += 16;
+            } else {
+               k += r;
+               zig = stbi__jpeg_dezigzag[k++];
+               data[zig] = (short) (stbi__extend_receive(j,s) << shift);
+            }
+         }
+      } while (k <= j->spec_end);
+   } else {
+      // refinement scan for these AC coefficients
+
+      short bit = (short) (1 << j->succ_low);
+
+      if (j->eob_run) {
+         --j->eob_run;
+         for (k = j->spec_start; k <= j->spec_end; ++k) {
+            short *p = &data[stbi__jpeg_dezigzag[k]];
+            if (*p != 0)
+               if (stbi__jpeg_get_bit(j))
+                  if ((*p & bit)==0) {
+                     if (*p > 0)
+                        *p += bit;
+                     else
+                        *p -= bit;
+                  }
+         }
+      } else {
+         k = j->spec_start;
+         do {
+            int r,s;
+            int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh
+            if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG");
+            s = rs & 15;
+            r = rs >> 4;
+            if (s == 0) {
+               if (r < 15) {
+                  j->eob_run = (1 << r) - 1;
+                  if (r)
+                     j->eob_run += stbi__jpeg_get_bits(j, r);
+                  r = 64; // force end of block
+               } else {
+                  // r=15 s=0 should write 16 0s, so we just do
+                  // a run of 15 0s and then write s (which is 0),
+                  // so we don't have to do anything special here
+               }
+            } else {
+               if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG");
+               // sign bit
+               if (stbi__jpeg_get_bit(j))
+                  s = bit;
+               else
+                  s = -bit;
+            }
+
+            // advance by r
+            while (k <= j->spec_end) {
+               short *p = &data[stbi__jpeg_dezigzag[k++]];
+               if (*p != 0) {
+                  if (stbi__jpeg_get_bit(j))
+                     if ((*p & bit)==0) {
+                        if (*p > 0)
+                           *p += bit;
+                        else
+                           *p -= bit;
+                     }
+               } else {
+                  if (r == 0) {
+                     *p = (short) s;
+                     break;
+                  }
+                  --r;
+               }
+            }
+         } while (k <= j->spec_end);
+      }
+   }
+   return 1;
+}
+
+// take a -128..127 value and stbi__clamp it and convert to 0..255
+stbi_inline static stbi_uc stbi__clamp(int x)
+{
+   // trick to use a single test to catch both cases
+   if ((unsigned int) x > 255) {
+      if (x < 0) return 0;
+      if (x > 255) return 255;
+   }
+   return (stbi_uc) x;
+}
+
+#define stbi__f2f(x)  ((int) (((x) * 4096 + 0.5)))
+#define stbi__fsh(x)  ((x) * 4096)
+
+// derived from jidctint -- DCT_ISLOW
+#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \
+   int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \
+   p2 = s2;                                    \
+   p3 = s6;                                    \
+   p1 = (p2+p3) * stbi__f2f(0.5411961f);       \
+   t2 = p1 + p3*stbi__f2f(-1.847759065f);      \
+   t3 = p1 + p2*stbi__f2f( 0.765366865f);      \
+   p2 = s0;                                    \
+   p3 = s4;                                    \
+   t0 = stbi__fsh(p2+p3);                      \
+   t1 = stbi__fsh(p2-p3);                      \
+   x0 = t0+t3;                                 \
+   x3 = t0-t3;                                 \
+   x1 = t1+t2;                                 \
+   x2 = t1-t2;                                 \
+   t0 = s7;                                    \
+   t1 = s5;                                    \
+   t2 = s3;                                    \
+   t3 = s1;                                    \
+   p3 = t0+t2;                                 \
+   p4 = t1+t3;                                 \
+   p1 = t0+t3;                                 \
+   p2 = t1+t2;                                 \
+   p5 = (p3+p4)*stbi__f2f( 1.175875602f);      \
+   t0 = t0*stbi__f2f( 0.298631336f);           \
+   t1 = t1*stbi__f2f( 2.053119869f);           \
+   t2 = t2*stbi__f2f( 3.072711026f);           \
+   t3 = t3*stbi__f2f( 1.501321110f);           \
+   p1 = p5 + p1*stbi__f2f(-0.899976223f);      \
+   p2 = p5 + p2*stbi__f2f(-2.562915447f);      \
+   p3 = p3*stbi__f2f(-1.961570560f);           \
+   p4 = p4*stbi__f2f(-0.390180644f);           \
+   t3 += p1+p4;                                \
+   t2 += p2+p3;                                \
+   t1 += p2+p4;                                \
+   t0 += p1+p3;
+
+static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64])
+{
+   int i,val[64],*v=val;
+   stbi_uc *o;
+   short *d = data;
+
+   // columns
+   for (i=0; i < 8; ++i,++d, ++v) {
+      // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing
+      if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0
+           && d[40]==0 && d[48]==0 && d[56]==0) {
+         //    no shortcut                 0     seconds
+         //    (1|2|3|4|5|6|7)==0          0     seconds
+         //    all separate               -0.047 seconds
+         //    1 && 2|3 && 4|5 && 6|7:    -0.047 seconds
+         int dcterm = d[0]*4;
+         v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm;
+      } else {
+         STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56])
+         // constants scaled things up by 1<<12; let's bring them back
+         // down, but keep 2 extra bits of precision
+         x0 += 512; x1 += 512; x2 += 512; x3 += 512;
+         v[ 0] = (x0+t3) >> 10;
+         v[56] = (x0-t3) >> 10;
+         v[ 8] = (x1+t2) >> 10;
+         v[48] = (x1-t2) >> 10;
+         v[16] = (x2+t1) >> 10;
+         v[40] = (x2-t1) >> 10;
+         v[24] = (x3+t0) >> 10;
+         v[32] = (x3-t0) >> 10;
+      }
+   }
+
+   for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) {
+      // no fast case since the first 1D IDCT spread components out
+      STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7])
+      // constants scaled things up by 1<<12, plus we had 1<<2 from first
+      // loop, plus horizontal and vertical each scale by sqrt(8) so together
+      // we've got an extra 1<<3, so 1<<17 total we need to remove.
+      // so we want to round that, which means adding 0.5 * 1<<17,
+      // aka 65536. Also, we'll end up with -128 to 127 that we want
+      // to encode as 0..255 by adding 128, so we'll add that before the shift
+      x0 += 65536 + (128<<17);
+      x1 += 65536 + (128<<17);
+      x2 += 65536 + (128<<17);
+      x3 += 65536 + (128<<17);
+      // tried computing the shifts into temps, or'ing the temps to see
+      // if any were out of range, but that was slower
+      o[0] = stbi__clamp((x0+t3) >> 17);
+      o[7] = stbi__clamp((x0-t3) >> 17);
+      o[1] = stbi__clamp((x1+t2) >> 17);
+      o[6] = stbi__clamp((x1-t2) >> 17);
+      o[2] = stbi__clamp((x2+t1) >> 17);
+      o[5] = stbi__clamp((x2-t1) >> 17);
+      o[3] = stbi__clamp((x3+t0) >> 17);
+      o[4] = stbi__clamp((x3-t0) >> 17);
+   }
+}
+
+#ifdef STBI_SSE2
+// sse2 integer IDCT. not the fastest possible implementation but it
+// produces bit-identical results to the generic C version so it's
+// fully "transparent".
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+   // This is constructed to match our regular (generic) integer IDCT exactly.
+   __m128i row0, row1, row2, row3, row4, row5, row6, row7;
+   __m128i tmp;
+
+   // dot product constant: even elems=x, odd elems=y
+   #define dct_const(x,y)  _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y))
+
+   // out(0) = c0[even]*x + c0[odd]*y   (c0, x, y 16-bit, out 32-bit)
+   // out(1) = c1[even]*x + c1[odd]*y
+   #define dct_rot(out0,out1, x,y,c0,c1) \
+      __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \
+      __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \
+      __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \
+      __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \
+      __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \
+      __m128i out1##_h = _mm_madd_epi16(c0##hi, c1)
+
+   // out = in << 12  (in 16-bit, out 32-bit)
+   #define dct_widen(out, in) \
+      __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \
+      __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4)
+
+   // wide add
+   #define dct_wadd(out, a, b) \
+      __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \
+      __m128i out##_h = _mm_add_epi32(a##_h, b##_h)
+
+   // wide sub
+   #define dct_wsub(out, a, b) \
+      __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \
+      __m128i out##_h = _mm_sub_epi32(a##_h, b##_h)
+
+   // butterfly a/b, add bias, then shift by "s" and pack
+   #define dct_bfly32o(out0, out1, a,b,bias,s) \
+      { \
+         __m128i abiased_l = _mm_add_epi32(a##_l, bias); \
+         __m128i abiased_h = _mm_add_epi32(a##_h, bias); \
+         dct_wadd(sum, abiased, b); \
+         dct_wsub(dif, abiased, b); \
+         out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \
+         out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \
+      }
+
+   // 8-bit interleave step (for transposes)
+   #define dct_interleave8(a, b) \
+      tmp = a; \
+      a = _mm_unpacklo_epi8(a, b); \
+      b = _mm_unpackhi_epi8(tmp, b)
+
+   // 16-bit interleave step (for transposes)
+   #define dct_interleave16(a, b) \
+      tmp = a; \
+      a = _mm_unpacklo_epi16(a, b); \
+      b = _mm_unpackhi_epi16(tmp, b)
+
+   #define dct_pass(bias,shift) \
+      { \
+         /* even part */ \
+         dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \
+         __m128i sum04 = _mm_add_epi16(row0, row4); \
+         __m128i dif04 = _mm_sub_epi16(row0, row4); \
+         dct_widen(t0e, sum04); \
+         dct_widen(t1e, dif04); \
+         dct_wadd(x0, t0e, t3e); \
+         dct_wsub(x3, t0e, t3e); \
+         dct_wadd(x1, t1e, t2e); \
+         dct_wsub(x2, t1e, t2e); \
+         /* odd part */ \
+         dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \
+         dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \
+         __m128i sum17 = _mm_add_epi16(row1, row7); \
+         __m128i sum35 = _mm_add_epi16(row3, row5); \
+         dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \
+         dct_wadd(x4, y0o, y4o); \
+         dct_wadd(x5, y1o, y5o); \
+         dct_wadd(x6, y2o, y5o); \
+         dct_wadd(x7, y3o, y4o); \
+         dct_bfly32o(row0,row7, x0,x7,bias,shift); \
+         dct_bfly32o(row1,row6, x1,x6,bias,shift); \
+         dct_bfly32o(row2,row5, x2,x5,bias,shift); \
+         dct_bfly32o(row3,row4, x3,x4,bias,shift); \
+      }
+
+   __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f));
+   __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f));
+   __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f));
+   __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f));
+   __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f));
+   __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f));
+   __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f));
+   __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f));
+
+   // rounding biases in column/row passes, see stbi__idct_block for explanation.
+   __m128i bias_0 = _mm_set1_epi32(512);
+   __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17));
+
+   // load
+   row0 = _mm_load_si128((const __m128i *) (data + 0*8));
+   row1 = _mm_load_si128((const __m128i *) (data + 1*8));
+   row2 = _mm_load_si128((const __m128i *) (data + 2*8));
+   row3 = _mm_load_si128((const __m128i *) (data + 3*8));
+   row4 = _mm_load_si128((const __m128i *) (data + 4*8));
+   row5 = _mm_load_si128((const __m128i *) (data + 5*8));
+   row6 = _mm_load_si128((const __m128i *) (data + 6*8));
+   row7 = _mm_load_si128((const __m128i *) (data + 7*8));
+
+   // column pass
+   dct_pass(bias_0, 10);
+
+   {
+      // 16bit 8x8 transpose pass 1
+      dct_interleave16(row0, row4);
+      dct_interleave16(row1, row5);
+      dct_interleave16(row2, row6);
+      dct_interleave16(row3, row7);
+
+      // transpose pass 2
+      dct_interleave16(row0, row2);
+      dct_interleave16(row1, row3);
+      dct_interleave16(row4, row6);
+      dct_interleave16(row5, row7);
+
+      // transpose pass 3
+      dct_interleave16(row0, row1);
+      dct_interleave16(row2, row3);
+      dct_interleave16(row4, row5);
+      dct_interleave16(row6, row7);
+   }
+
+   // row pass
+   dct_pass(bias_1, 17);
+
+   {
+      // pack
+      __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7
+      __m128i p1 = _mm_packus_epi16(row2, row3);
+      __m128i p2 = _mm_packus_epi16(row4, row5);
+      __m128i p3 = _mm_packus_epi16(row6, row7);
+
+      // 8bit 8x8 transpose pass 1
+      dct_interleave8(p0, p2); // a0e0a1e1...
+      dct_interleave8(p1, p3); // c0g0c1g1...
+
+      // transpose pass 2
+      dct_interleave8(p0, p1); // a0c0e0g0...
+      dct_interleave8(p2, p3); // b0d0f0h0...
+
+      // transpose pass 3
+      dct_interleave8(p0, p2); // a0b0c0d0...
+      dct_interleave8(p1, p3); // a4b4c4d4...
+
+      // store
+      _mm_storel_epi64((__m128i *) out, p0); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p2); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p1); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, p3); out += out_stride;
+      _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e));
+   }
+
+#undef dct_const
+#undef dct_rot
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_interleave8
+#undef dct_interleave16
+#undef dct_pass
+}
+
+#endif // STBI_SSE2
+
+#ifdef STBI_NEON
+
+// NEON integer IDCT. should produce bit-identical
+// results to the generic C version.
+static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])
+{
+   int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;
+
+   int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f));
+   int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f));
+   int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f));
+   int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f));
+   int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f));
+   int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f));
+   int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f));
+   int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f));
+   int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f));
+   int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f));
+   int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f));
+   int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f));
+
+#define dct_long_mul(out, inq, coeff) \
+   int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \
+   int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff)
+
+#define dct_long_mac(out, acc, inq, coeff) \
+   int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \
+   int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff)
+
+#define dct_widen(out, inq) \
+   int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \
+   int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12)
+
+// wide add
+#define dct_wadd(out, a, b) \
+   int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \
+   int32x4_t out##_h = vaddq_s32(a##_h, b##_h)
+
+// wide sub
+#define dct_wsub(out, a, b) \
+   int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \
+   int32x4_t out##_h = vsubq_s32(a##_h, b##_h)
+
+// butterfly a/b, then shift using "shiftop" by "s" and pack
+#define dct_bfly32o(out0,out1, a,b,shiftop,s) \
+   { \
+      dct_wadd(sum, a, b); \
+      dct_wsub(dif, a, b); \
+      out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \
+      out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \
+   }
+
+#define dct_pass(shiftop, shift) \
+   { \
+      /* even part */ \
+      int16x8_t sum26 = vaddq_s16(row2, row6); \
+      dct_long_mul(p1e, sum26, rot0_0); \
+      dct_long_mac(t2e, p1e, row6, rot0_1); \
+      dct_long_mac(t3e, p1e, row2, rot0_2); \
+      int16x8_t sum04 = vaddq_s16(row0, row4); \
+      int16x8_t dif04 = vsubq_s16(row0, row4); \
+      dct_widen(t0e, sum04); \
+      dct_widen(t1e, dif04); \
+      dct_wadd(x0, t0e, t3e); \
+      dct_wsub(x3, t0e, t3e); \
+      dct_wadd(x1, t1e, t2e); \
+      dct_wsub(x2, t1e, t2e); \
+      /* odd part */ \
+      int16x8_t sum15 = vaddq_s16(row1, row5); \
+      int16x8_t sum17 = vaddq_s16(row1, row7); \
+      int16x8_t sum35 = vaddq_s16(row3, row5); \
+      int16x8_t sum37 = vaddq_s16(row3, row7); \
+      int16x8_t sumodd = vaddq_s16(sum17, sum35); \
+      dct_long_mul(p5o, sumodd, rot1_0); \
+      dct_long_mac(p1o, p5o, sum17, rot1_1); \
+      dct_long_mac(p2o, p5o, sum35, rot1_2); \
+      dct_long_mul(p3o, sum37, rot2_0); \
+      dct_long_mul(p4o, sum15, rot2_1); \
+      dct_wadd(sump13o, p1o, p3o); \
+      dct_wadd(sump24o, p2o, p4o); \
+      dct_wadd(sump23o, p2o, p3o); \
+      dct_wadd(sump14o, p1o, p4o); \
+      dct_long_mac(x4, sump13o, row7, rot3_0); \
+      dct_long_mac(x5, sump24o, row5, rot3_1); \
+      dct_long_mac(x6, sump23o, row3, rot3_2); \
+      dct_long_mac(x7, sump14o, row1, rot3_3); \
+      dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \
+      dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \
+      dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \
+      dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \
+   }
+
+   // load
+   row0 = vld1q_s16(data + 0*8);
+   row1 = vld1q_s16(data + 1*8);
+   row2 = vld1q_s16(data + 2*8);
+   row3 = vld1q_s16(data + 3*8);
+   row4 = vld1q_s16(data + 4*8);
+   row5 = vld1q_s16(data + 5*8);
+   row6 = vld1q_s16(data + 6*8);
+   row7 = vld1q_s16(data + 7*8);
+
+   // add DC bias
+   row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0));
+
+   // column pass
+   dct_pass(vrshrn_n_s32, 10);
+
+   // 16bit 8x8 transpose
+   {
+// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively.
+// whether compilers actually get this is another story, sadly.
+#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); }
+#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); }
+
+      // pass 1
+      dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6
+      dct_trn16(row2, row3);
+      dct_trn16(row4, row5);
+      dct_trn16(row6, row7);
+
+      // pass 2
+      dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4
+      dct_trn32(row1, row3);
+      dct_trn32(row4, row6);
+      dct_trn32(row5, row7);
+
+      // pass 3
+      dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0
+      dct_trn64(row1, row5);
+      dct_trn64(row2, row6);
+      dct_trn64(row3, row7);
+
+#undef dct_trn16
+#undef dct_trn32
+#undef dct_trn64
+   }
+
+   // row pass
+   // vrshrn_n_s32 only supports shifts up to 16, we need
+   // 17. so do a non-rounding shift of 16 first then follow
+   // up with a rounding shift by 1.
+   dct_pass(vshrn_n_s32, 16);
+
+   {
+      // pack and round
+      uint8x8_t p0 = vqrshrun_n_s16(row0, 1);
+      uint8x8_t p1 = vqrshrun_n_s16(row1, 1);
+      uint8x8_t p2 = vqrshrun_n_s16(row2, 1);
+      uint8x8_t p3 = vqrshrun_n_s16(row3, 1);
+      uint8x8_t p4 = vqrshrun_n_s16(row4, 1);
+      uint8x8_t p5 = vqrshrun_n_s16(row5, 1);
+      uint8x8_t p6 = vqrshrun_n_s16(row6, 1);
+      uint8x8_t p7 = vqrshrun_n_s16(row7, 1);
+
+      // again, these can translate into one instruction, but often don't.
+#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; }
+#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); }
+#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); }
+
+      // sadly can't use interleaved stores here since we only write
+      // 8 bytes to each scan line!
+
+      // 8x8 8-bit transpose pass 1
+      dct_trn8_8(p0, p1);
+      dct_trn8_8(p2, p3);
+      dct_trn8_8(p4, p5);
+      dct_trn8_8(p6, p7);
+
+      // pass 2
+      dct_trn8_16(p0, p2);
+      dct_trn8_16(p1, p3);
+      dct_trn8_16(p4, p6);
+      dct_trn8_16(p5, p7);
+
+      // pass 3
+      dct_trn8_32(p0, p4);
+      dct_trn8_32(p1, p5);
+      dct_trn8_32(p2, p6);
+      dct_trn8_32(p3, p7);
+
+      // store
+      vst1_u8(out, p0); out += out_stride;
+      vst1_u8(out, p1); out += out_stride;
+      vst1_u8(out, p2); out += out_stride;
+      vst1_u8(out, p3); out += out_stride;
+      vst1_u8(out, p4); out += out_stride;
+      vst1_u8(out, p5); out += out_stride;
+      vst1_u8(out, p6); out += out_stride;
+      vst1_u8(out, p7);
+
+#undef dct_trn8_8
+#undef dct_trn8_16
+#undef dct_trn8_32
+   }
+
+#undef dct_long_mul
+#undef dct_long_mac
+#undef dct_widen
+#undef dct_wadd
+#undef dct_wsub
+#undef dct_bfly32o
+#undef dct_pass
+}
+
+#endif // STBI_NEON
+
+#define STBI__MARKER_none  0xff
+// if there's a pending marker from the entropy stream, return that
+// otherwise, fetch from the stream and get a marker. if there's no
+// marker, return 0xff, which is never a valid marker value
+static stbi_uc stbi__get_marker(stbi__jpeg *j)
+{
+   stbi_uc x;
+   if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; }
+   x = stbi__get8(j->s);
+   if (x != 0xff) return STBI__MARKER_none;
+   while (x == 0xff)
+      x = stbi__get8(j->s); // consume repeated 0xff fill bytes
+   return x;
+}
+
+// in each scan, we'll have scan_n components, and the order
+// of the components is specified by order[]
+#define STBI__RESTART(x)     ((x) >= 0xd0 && (x) <= 0xd7)
+
+// after a restart interval, stbi__jpeg_reset the entropy decoder and
+// the dc prediction
+static void stbi__jpeg_reset(stbi__jpeg *j)
+{
+   j->code_bits = 0;
+   j->code_buffer = 0;
+   j->nomore = 0;
+   j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0;
+   j->marker = STBI__MARKER_none;
+   j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff;
+   j->eob_run = 0;
+   // no more than 1<<31 MCUs if no restart_interal? that's plenty safe,
+   // since we don't even allow 1<<30 pixels
+}
+
+static int stbi__parse_entropy_coded_data(stbi__jpeg *z)
+{
+   stbi__jpeg_reset(z);
+   if (!z->progressive) {
+      if (z->scan_n == 1) {
+         int i,j;
+         STBI_SIMD_ALIGN(short, data[64]);
+         int n = z->order[0];
+         // non-interleaved data, we just need to process one block at a time,
+         // in trivial scanline order
+         // number of blocks to do just depends on how many actual "pixels" this
+         // component has, independent of interleaved MCU blocking and such
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               int ha = z->img_comp[n].ha;
+               if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+               z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+               // every data block is an MCU, so countdown the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  // if it's NOT a restart, then just bail, so we get corrupt data
+                  // rather than no data
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      } else { // interleaved
+         int i,j,k,x,y;
+         STBI_SIMD_ALIGN(short, data[64]);
+         for (j=0; j < z->img_mcu_y; ++j) {
+            for (i=0; i < z->img_mcu_x; ++i) {
+               // scan an interleaved mcu... process scan_n components in order
+               for (k=0; k < z->scan_n; ++k) {
+                  int n = z->order[k];
+                  // scan out an mcu's worth of this component; that's just determined
+                  // by the basic H and V specified for the component
+                  for (y=0; y < z->img_comp[n].v; ++y) {
+                     for (x=0; x < z->img_comp[n].h; ++x) {
+                        int x2 = (i*z->img_comp[n].h + x)*8;
+                        int y2 = (j*z->img_comp[n].v + y)*8;
+                        int ha = z->img_comp[n].ha;
+                        if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;
+                        z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data);
+                     }
+                  }
+               }
+               // after all interleaved components, that's an interleaved MCU,
+               // so now count down the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      }
+   } else {
+      if (z->scan_n == 1) {
+         int i,j;
+         int n = z->order[0];
+         // non-interleaved data, we just need to process one block at a time,
+         // in trivial scanline order
+         // number of blocks to do just depends on how many actual "pixels" this
+         // component has, independent of interleaved MCU blocking and such
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+               if (z->spec_start == 0) {
+                  if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+                     return 0;
+               } else {
+                  int ha = z->img_comp[n].ha;
+                  if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha]))
+                     return 0;
+               }
+               // every data block is an MCU, so countdown the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      } else { // interleaved
+         int i,j,k,x,y;
+         for (j=0; j < z->img_mcu_y; ++j) {
+            for (i=0; i < z->img_mcu_x; ++i) {
+               // scan an interleaved mcu... process scan_n components in order
+               for (k=0; k < z->scan_n; ++k) {
+                  int n = z->order[k];
+                  // scan out an mcu's worth of this component; that's just determined
+                  // by the basic H and V specified for the component
+                  for (y=0; y < z->img_comp[n].v; ++y) {
+                     for (x=0; x < z->img_comp[n].h; ++x) {
+                        int x2 = (i*z->img_comp[n].h + x);
+                        int y2 = (j*z->img_comp[n].v + y);
+                        short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w);
+                        if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))
+                           return 0;
+                     }
+                  }
+               }
+               // after all interleaved components, that's an interleaved MCU,
+               // so now count down the restart interval
+               if (--z->todo <= 0) {
+                  if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);
+                  if (!STBI__RESTART(z->marker)) return 1;
+                  stbi__jpeg_reset(z);
+               }
+            }
+         }
+         return 1;
+      }
+   }
+}
+
+static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant)
+{
+   int i;
+   for (i=0; i < 64; ++i)
+      data[i] *= dequant[i];
+}
+
+static void stbi__jpeg_finish(stbi__jpeg *z)
+{
+   if (z->progressive) {
+      // dequantize and idct the data
+      int i,j,n;
+      for (n=0; n < z->s->img_n; ++n) {
+         int w = (z->img_comp[n].x+7) >> 3;
+         int h = (z->img_comp[n].y+7) >> 3;
+         for (j=0; j < h; ++j) {
+            for (i=0; i < w; ++i) {
+               short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);
+               stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]);
+               z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);
+            }
+         }
+      }
+   }
+}
+
+static int stbi__process_marker(stbi__jpeg *z, int m)
+{
+   int L;
+   switch (m) {
+      case STBI__MARKER_none: // no marker found
+         return stbi__err("expected marker","Corrupt JPEG");
+
+      case 0xDD: // DRI - specify restart interval
+         if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG");
+         z->restart_interval = stbi__get16be(z->s);
+         return 1;
+
+      case 0xDB: // DQT - define quantization table
+         L = stbi__get16be(z->s)-2;
+         while (L > 0) {
+            int q = stbi__get8(z->s);
+            int p = q >> 4, sixteen = (p != 0);
+            int t = q & 15,i;
+            if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG");
+            if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG");
+
+            for (i=0; i < 64; ++i)
+               z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s));
+            L -= (sixteen ? 129 : 65);
+         }
+         return L==0;
+
+      case 0xC4: // DHT - define huffman table
+         L = stbi__get16be(z->s)-2;
+         while (L > 0) {
+            stbi_uc *v;
+            int sizes[16],i,n=0;
+            int q = stbi__get8(z->s);
+            int tc = q >> 4;
+            int th = q & 15;
+            if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG");
+            for (i=0; i < 16; ++i) {
+               sizes[i] = stbi__get8(z->s);
+               n += sizes[i];
+            }
+            L -= 17;
+            if (tc == 0) {
+               if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0;
+               v = z->huff_dc[th].values;
+            } else {
+               if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0;
+               v = z->huff_ac[th].values;
+            }
+            for (i=0; i < n; ++i)
+               v[i] = stbi__get8(z->s);
+            if (tc != 0)
+               stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th);
+            L -= n;
+         }
+         return L==0;
+   }
+
+   // check for comment block or APP blocks
+   if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) {
+      L = stbi__get16be(z->s);
+      if (L < 2) {
+         if (m == 0xFE)
+            return stbi__err("bad COM len","Corrupt JPEG");
+         else
+            return stbi__err("bad APP len","Corrupt JPEG");
+      }
+      L -= 2;
+
+      if (m == 0xE0 && L >= 5) { // JFIF APP0 segment
+         static const unsigned char tag[5] = {'J','F','I','F','\0'};
+         int ok = 1;
+         int i;
+         for (i=0; i < 5; ++i)
+            if (stbi__get8(z->s) != tag[i])
+               ok = 0;
+         L -= 5;
+         if (ok)
+            z->jfif = 1;
+      } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment
+         static const unsigned char tag[6] = {'A','d','o','b','e','\0'};
+         int ok = 1;
+         int i;
+         for (i=0; i < 6; ++i)
+            if (stbi__get8(z->s) != tag[i])
+               ok = 0;
+         L -= 6;
+         if (ok) {
+            stbi__get8(z->s); // version
+            stbi__get16be(z->s); // flags0
+            stbi__get16be(z->s); // flags1
+            z->app14_color_transform = stbi__get8(z->s); // color transform
+            L -= 6;
+         }
+      }
+
+      stbi__skip(z->s, L);
+      return 1;
+   }
+
+   return stbi__err("unknown marker","Corrupt JPEG");
+}
+
+// after we see SOS
+static int stbi__process_scan_header(stbi__jpeg *z)
+{
+   int i;
+   int Ls = stbi__get16be(z->s);
+   z->scan_n = stbi__get8(z->s);
+   if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG");
+   if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG");
+   for (i=0; i < z->scan_n; ++i) {
+      int id = stbi__get8(z->s), which;
+      int q = stbi__get8(z->s);
+      for (which = 0; which < z->s->img_n; ++which)
+         if (z->img_comp[which].id == id)
+            break;
+      if (which == z->s->img_n) return 0; // no match
+      z->img_comp[which].hd = q >> 4;   if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG");
+      z->img_comp[which].ha = q & 15;   if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG");
+      z->order[i] = which;
+   }
+
+   {
+      int aa;
+      z->spec_start = stbi__get8(z->s);
+      z->spec_end   = stbi__get8(z->s); // should be 63, but might be 0
+      aa = stbi__get8(z->s);
+      z->succ_high = (aa >> 4);
+      z->succ_low  = (aa & 15);
+      if (z->progressive) {
+         if (z->spec_start > 63 || z->spec_end > 63  || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13)
+            return stbi__err("bad SOS", "Corrupt JPEG");
+      } else {
+         if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG");
+         if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG");
+         z->spec_end = 63;
+      }
+   }
+
+   return 1;
+}
+
+static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why)
+{
+   int i;
+   for (i=0; i < ncomp; ++i) {
+      if (z->img_comp[i].raw_data) {
+         STBI_FREE(z->img_comp[i].raw_data);
+         z->img_comp[i].raw_data = NULL;
+         z->img_comp[i].data = NULL;
+      }
+      if (z->img_comp[i].raw_coeff) {
+         STBI_FREE(z->img_comp[i].raw_coeff);
+         z->img_comp[i].raw_coeff = 0;
+         z->img_comp[i].coeff = 0;
+      }
+      if (z->img_comp[i].linebuf) {
+         STBI_FREE(z->img_comp[i].linebuf);
+         z->img_comp[i].linebuf = NULL;
+      }
+   }
+   return why;
+}
+
+static int stbi__process_frame_header(stbi__jpeg *z, int scan)
+{
+   stbi__context *s = z->s;
+   int Lf,p,i,q, h_max=1,v_max=1,c;
+   Lf = stbi__get16be(s);         if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG
+   p  = stbi__get8(s);            if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline
+   s->img_y = stbi__get16be(s);   if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG
+   s->img_x = stbi__get16be(s);   if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires
+   c = stbi__get8(s);
+   if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG");
+   s->img_n = c;
+   for (i=0; i < c; ++i) {
+      z->img_comp[i].data = NULL;
+      z->img_comp[i].linebuf = NULL;
+   }
+
+   if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG");
+
+   z->rgb = 0;
+   for (i=0; i < s->img_n; ++i) {
+      static const unsigned char rgb[3] = { 'R', 'G', 'B' };
+      z->img_comp[i].id = stbi__get8(s);
+      if (s->img_n == 3 && z->img_comp[i].id == rgb[i])
+         ++z->rgb;
+      q = stbi__get8(s);
+      z->img_comp[i].h = (q >> 4);  if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG");
+      z->img_comp[i].v = q & 15;    if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG");
+      z->img_comp[i].tq = stbi__get8(s);  if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG");
+   }
+
+   if (scan != STBI__SCAN_load) return 1;
+
+   if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode");
+
+   for (i=0; i < s->img_n; ++i) {
+      if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h;
+      if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v;
+   }
+
+   // compute interleaved mcu info
+   z->img_h_max = h_max;
+   z->img_v_max = v_max;
+   z->img_mcu_w = h_max * 8;
+   z->img_mcu_h = v_max * 8;
+   // these sizes can't be more than 17 bits
+   z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w;
+   z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h;
+
+   for (i=0; i < s->img_n; ++i) {
+      // number of effective pixels (e.g. for non-interleaved MCU)
+      z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max;
+      z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max;
+      // to simplify generation, we'll allocate enough memory to decode
+      // the bogus oversized data from using interleaved MCUs and their
+      // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't
+      // discard the extra data until colorspace conversion
+      //
+      // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier)
+      // so these muls can't overflow with 32-bit ints (which we require)
+      z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8;
+      z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8;
+      z->img_comp[i].coeff = 0;
+      z->img_comp[i].raw_coeff = 0;
+      z->img_comp[i].linebuf = NULL;
+      z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15);
+      if (z->img_comp[i].raw_data == NULL)
+         return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+      // align blocks for idct using mmx/sse
+      z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15);
+      if (z->progressive) {
+         // w2, h2 are multiples of 8 (see above)
+         z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8;
+         z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8;
+         z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15);
+         if (z->img_comp[i].raw_coeff == NULL)
+            return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory"));
+         z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15);
+      }
+   }
+
+   return 1;
+}
+
+// use comparisons since in some cases we handle more than one case (e.g. SOF)
+#define stbi__DNL(x)         ((x) == 0xdc)
+#define stbi__SOI(x)         ((x) == 0xd8)
+#define stbi__EOI(x)         ((x) == 0xd9)
+#define stbi__SOF(x)         ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2)
+#define stbi__SOS(x)         ((x) == 0xda)
+
+#define stbi__SOF_progressive(x)   ((x) == 0xc2)
+
+static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan)
+{
+   int m;
+   z->jfif = 0;
+   z->app14_color_transform = -1; // valid values are 0,1,2
+   z->marker = STBI__MARKER_none; // initialize cached marker to empty
+   m = stbi__get_marker(z);
+   if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG");
+   if (scan == STBI__SCAN_type) return 1;
+   m = stbi__get_marker(z);
+   while (!stbi__SOF(m)) {
+      if (!stbi__process_marker(z,m)) return 0;
+      m = stbi__get_marker(z);
+      while (m == STBI__MARKER_none) {
+         // some files have extra padding after their blocks, so ok, we'll scan
+         if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG");
+         m = stbi__get_marker(z);
+      }
+   }
+   z->progressive = stbi__SOF_progressive(m);
+   if (!stbi__process_frame_header(z, scan)) return 0;
+   return 1;
+}
+
+// decode image to YCbCr format
+static int stbi__decode_jpeg_image(stbi__jpeg *j)
+{
+   int m;
+   for (m = 0; m < 4; m++) {
+      j->img_comp[m].raw_data = NULL;
+      j->img_comp[m].raw_coeff = NULL;
+   }
+   j->restart_interval = 0;
+   if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0;
+   m = stbi__get_marker(j);
+   while (!stbi__EOI(m)) {
+      if (stbi__SOS(m)) {
+         if (!stbi__process_scan_header(j)) return 0;
+         if (!stbi__parse_entropy_coded_data(j)) return 0;
+         if (j->marker == STBI__MARKER_none ) {
+            // handle 0s at the end of image data from IP Kamera 9060
+            while (!stbi__at_eof(j->s)) {
+               int x = stbi__get8(j->s);
+               if (x == 255) {
+                  j->marker = stbi__get8(j->s);
+                  break;
+               }
+            }
+            // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0
+         }
+      } else if (stbi__DNL(m)) {
+         int Ld = stbi__get16be(j->s);
+         stbi__uint32 NL = stbi__get16be(j->s);
+         if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG");
+         if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG");
+      } else {
+         if (!stbi__process_marker(j, m)) return 0;
+      }
+      m = stbi__get_marker(j);
+   }
+   if (j->progressive)
+      stbi__jpeg_finish(j);
+   return 1;
+}
+
+// static jfif-centered resampling (across block boundaries)
+
+typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1,
+                                    int w, int hs);
+
+#define stbi__div4(x) ((stbi_uc) ((x) >> 2))
+
+static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   STBI_NOTUSED(out);
+   STBI_NOTUSED(in_far);
+   STBI_NOTUSED(w);
+   STBI_NOTUSED(hs);
+   return in_near;
+}
+
+static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate two samples vertically for every one in input
+   int i;
+   STBI_NOTUSED(hs);
+   for (i=0; i < w; ++i)
+      out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2);
+   return out;
+}
+
+static stbi_uc*  stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate two samples horizontally for every one in input
+   int i;
+   stbi_uc *input = in_near;
+
+   if (w == 1) {
+      // if only one sample, can't do any interpolation
+      out[0] = out[1] = input[0];
+      return out;
+   }
+
+   out[0] = input[0];
+   out[1] = stbi__div4(input[0]*3 + input[1] + 2);
+   for (i=1; i < w-1; ++i) {
+      int n = 3*input[i]+2;
+      out[i*2+0] = stbi__div4(n+input[i-1]);
+      out[i*2+1] = stbi__div4(n+input[i+1]);
+   }
+   out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2);
+   out[i*2+1] = input[w-1];
+
+   STBI_NOTUSED(in_far);
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+
+#define stbi__div16(x) ((stbi_uc) ((x) >> 4))
+
+static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate 2x2 samples for every one in input
+   int i,t0,t1;
+   if (w == 1) {
+      out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+      return out;
+   }
+
+   t1 = 3*in_near[0] + in_far[0];
+   out[0] = stbi__div4(t1+2);
+   for (i=1; i < w; ++i) {
+      t0 = t1;
+      t1 = 3*in_near[i]+in_far[i];
+      out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+      out[i*2  ] = stbi__div16(3*t1 + t0 + 8);
+   }
+   out[w*2-1] = stbi__div4(t1+2);
+
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // need to generate 2x2 samples for every one in input
+   int i=0,t0,t1;
+
+   if (w == 1) {
+      out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);
+      return out;
+   }
+
+   t1 = 3*in_near[0] + in_far[0];
+   // process groups of 8 pixels for as long as we can.
+   // note we can't handle the last pixel in a row in this loop
+   // because we need to handle the filter boundary conditions.
+   for (; i < ((w-1) & ~7); i += 8) {
+#if defined(STBI_SSE2)
+      // load and perform the vertical filtering pass
+      // this uses 3*x + y = 4*x + (y - x)
+      __m128i zero  = _mm_setzero_si128();
+      __m128i farb  = _mm_loadl_epi64((__m128i *) (in_far + i));
+      __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i));
+      __m128i farw  = _mm_unpacklo_epi8(farb, zero);
+      __m128i nearw = _mm_unpacklo_epi8(nearb, zero);
+      __m128i diff  = _mm_sub_epi16(farw, nearw);
+      __m128i nears = _mm_slli_epi16(nearw, 2);
+      __m128i curr  = _mm_add_epi16(nears, diff); // current row
+
+      // horizontal filter works the same based on shifted vers of current
+      // row. "prev" is current row shifted right by 1 pixel; we need to
+      // insert the previous pixel value (from t1).
+      // "next" is current row shifted left by 1 pixel, with first pixel
+      // of next block of 8 pixels added in.
+      __m128i prv0 = _mm_slli_si128(curr, 2);
+      __m128i nxt0 = _mm_srli_si128(curr, 2);
+      __m128i prev = _mm_insert_epi16(prv0, t1, 0);
+      __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7);
+
+      // horizontal filter, polyphase implementation since it's convenient:
+      // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+      // odd  pixels = 3*cur + next = cur*4 + (next - cur)
+      // note the shared term.
+      __m128i bias  = _mm_set1_epi16(8);
+      __m128i curs = _mm_slli_epi16(curr, 2);
+      __m128i prvd = _mm_sub_epi16(prev, curr);
+      __m128i nxtd = _mm_sub_epi16(next, curr);
+      __m128i curb = _mm_add_epi16(curs, bias);
+      __m128i even = _mm_add_epi16(prvd, curb);
+      __m128i odd  = _mm_add_epi16(nxtd, curb);
+
+      // interleave even and odd pixels, then undo scaling.
+      __m128i int0 = _mm_unpacklo_epi16(even, odd);
+      __m128i int1 = _mm_unpackhi_epi16(even, odd);
+      __m128i de0  = _mm_srli_epi16(int0, 4);
+      __m128i de1  = _mm_srli_epi16(int1, 4);
+
+      // pack and write output
+      __m128i outv = _mm_packus_epi16(de0, de1);
+      _mm_storeu_si128((__m128i *) (out + i*2), outv);
+#elif defined(STBI_NEON)
+      // load and perform the vertical filtering pass
+      // this uses 3*x + y = 4*x + (y - x)
+      uint8x8_t farb  = vld1_u8(in_far + i);
+      uint8x8_t nearb = vld1_u8(in_near + i);
+      int16x8_t diff  = vreinterpretq_s16_u16(vsubl_u8(farb, nearb));
+      int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2));
+      int16x8_t curr  = vaddq_s16(nears, diff); // current row
+
+      // horizontal filter works the same based on shifted vers of current
+      // row. "prev" is current row shifted right by 1 pixel; we need to
+      // insert the previous pixel value (from t1).
+      // "next" is current row shifted left by 1 pixel, with first pixel
+      // of next block of 8 pixels added in.
+      int16x8_t prv0 = vextq_s16(curr, curr, 7);
+      int16x8_t nxt0 = vextq_s16(curr, curr, 1);
+      int16x8_t prev = vsetq_lane_s16(t1, prv0, 0);
+      int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7);
+
+      // horizontal filter, polyphase implementation since it's convenient:
+      // even pixels = 3*cur + prev = cur*4 + (prev - cur)
+      // odd  pixels = 3*cur + next = cur*4 + (next - cur)
+      // note the shared term.
+      int16x8_t curs = vshlq_n_s16(curr, 2);
+      int16x8_t prvd = vsubq_s16(prev, curr);
+      int16x8_t nxtd = vsubq_s16(next, curr);
+      int16x8_t even = vaddq_s16(curs, prvd);
+      int16x8_t odd  = vaddq_s16(curs, nxtd);
+
+      // undo scaling and round, then store with even/odd phases interleaved
+      uint8x8x2_t o;
+      o.val[0] = vqrshrun_n_s16(even, 4);
+      o.val[1] = vqrshrun_n_s16(odd,  4);
+      vst2_u8(out + i*2, o);
+#endif
+
+      // "previous" value for next iter
+      t1 = 3*in_near[i+7] + in_far[i+7];
+   }
+
+   t0 = t1;
+   t1 = 3*in_near[i] + in_far[i];
+   out[i*2] = stbi__div16(3*t1 + t0 + 8);
+
+   for (++i; i < w; ++i) {
+      t0 = t1;
+      t1 = 3*in_near[i]+in_far[i];
+      out[i*2-1] = stbi__div16(3*t0 + t1 + 8);
+      out[i*2  ] = stbi__div16(3*t1 + t0 + 8);
+   }
+   out[w*2-1] = stbi__div4(t1+2);
+
+   STBI_NOTUSED(hs);
+
+   return out;
+}
+#endif
+
+static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)
+{
+   // resample with nearest-neighbor
+   int i,j;
+   STBI_NOTUSED(in_far);
+   for (i=0; i < w; ++i)
+      for (j=0; j < hs; ++j)
+         out[i*hs+j] = in_near[i];
+   return out;
+}
+
+// this is a reduced-precision calculation of YCbCr-to-RGB introduced
+// to make sure the code produces the same results in both SIMD and scalar
+#define stbi__float2fixed(x)  (((int) ((x) * 4096.0f + 0.5f)) << 8)
+static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)
+{
+   int i;
+   for (i=0; i < count; ++i) {
+      int y_fixed = (y[i] << 20) + (1<<19); // rounding
+      int r,g,b;
+      int cr = pcr[i] - 128;
+      int cb = pcb[i] - 128;
+      r = y_fixed +  cr* stbi__float2fixed(1.40200f);
+      g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+      b = y_fixed                                     +   cb* stbi__float2fixed(1.77200f);
+      r >>= 20;
+      g >>= 20;
+      b >>= 20;
+      if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+      if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+      if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+      out[0] = (stbi_uc)r;
+      out[1] = (stbi_uc)g;
+      out[2] = (stbi_uc)b;
+      out[3] = 255;
+      out += step;
+   }
+}
+
+#if defined(STBI_SSE2) || defined(STBI_NEON)
+static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step)
+{
+   int i = 0;
+
+#ifdef STBI_SSE2
+   // step == 3 is pretty ugly on the final interleave, and i'm not convinced
+   // it's useful in practice (you wouldn't use it for textures, for example).
+   // so just accelerate step == 4 case.
+   if (step == 4) {
+      // this is a fairly straightforward implementation and not super-optimized.
+      __m128i signflip  = _mm_set1_epi8(-0x80);
+      __m128i cr_const0 = _mm_set1_epi16(   (short) ( 1.40200f*4096.0f+0.5f));
+      __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f));
+      __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f));
+      __m128i cb_const1 = _mm_set1_epi16(   (short) ( 1.77200f*4096.0f+0.5f));
+      __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128);
+      __m128i xw = _mm_set1_epi16(255); // alpha channel
+
+      for (; i+7 < count; i += 8) {
+         // load
+         __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i));
+         __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i));
+         __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i));
+         __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128
+         __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128
+
+         // unpack to short (and left-shift cr, cb by 8)
+         __m128i yw  = _mm_unpacklo_epi8(y_bias, y_bytes);
+         __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased);
+         __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased);
+
+         // color transform
+         __m128i yws = _mm_srli_epi16(yw, 4);
+         __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw);
+         __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw);
+         __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1);
+         __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1);
+         __m128i rws = _mm_add_epi16(cr0, yws);
+         __m128i gwt = _mm_add_epi16(cb0, yws);
+         __m128i bws = _mm_add_epi16(yws, cb1);
+         __m128i gws = _mm_add_epi16(gwt, cr1);
+
+         // descale
+         __m128i rw = _mm_srai_epi16(rws, 4);
+         __m128i bw = _mm_srai_epi16(bws, 4);
+         __m128i gw = _mm_srai_epi16(gws, 4);
+
+         // back to byte, set up for transpose
+         __m128i brb = _mm_packus_epi16(rw, bw);
+         __m128i gxb = _mm_packus_epi16(gw, xw);
+
+         // transpose to interleave channels
+         __m128i t0 = _mm_unpacklo_epi8(brb, gxb);
+         __m128i t1 = _mm_unpackhi_epi8(brb, gxb);
+         __m128i o0 = _mm_unpacklo_epi16(t0, t1);
+         __m128i o1 = _mm_unpackhi_epi16(t0, t1);
+
+         // store
+         _mm_storeu_si128((__m128i *) (out + 0), o0);
+         _mm_storeu_si128((__m128i *) (out + 16), o1);
+         out += 32;
+      }
+   }
+#endif
+
+#ifdef STBI_NEON
+   // in this version, step=3 support would be easy to add. but is there demand?
+   if (step == 4) {
+      // this is a fairly straightforward implementation and not super-optimized.
+      uint8x8_t signflip = vdup_n_u8(0x80);
+      int16x8_t cr_const0 = vdupq_n_s16(   (short) ( 1.40200f*4096.0f+0.5f));
+      int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f));
+      int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f));
+      int16x8_t cb_const1 = vdupq_n_s16(   (short) ( 1.77200f*4096.0f+0.5f));
+
+      for (; i+7 < count; i += 8) {
+         // load
+         uint8x8_t y_bytes  = vld1_u8(y + i);
+         uint8x8_t cr_bytes = vld1_u8(pcr + i);
+         uint8x8_t cb_bytes = vld1_u8(pcb + i);
+         int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip));
+         int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip));
+
+         // expand to s16
+         int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4));
+         int16x8_t crw = vshll_n_s8(cr_biased, 7);
+         int16x8_t cbw = vshll_n_s8(cb_biased, 7);
+
+         // color transform
+         int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0);
+         int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0);
+         int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1);
+         int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1);
+         int16x8_t rws = vaddq_s16(yws, cr0);
+         int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1);
+         int16x8_t bws = vaddq_s16(yws, cb1);
+
+         // undo scaling, round, convert to byte
+         uint8x8x4_t o;
+         o.val[0] = vqrshrun_n_s16(rws, 4);
+         o.val[1] = vqrshrun_n_s16(gws, 4);
+         o.val[2] = vqrshrun_n_s16(bws, 4);
+         o.val[3] = vdup_n_u8(255);
+
+         // store, interleaving r/g/b/a
+         vst4_u8(out, o);
+         out += 8*4;
+      }
+   }
+#endif
+
+   for (; i < count; ++i) {
+      int y_fixed = (y[i] << 20) + (1<<19); // rounding
+      int r,g,b;
+      int cr = pcr[i] - 128;
+      int cb = pcb[i] - 128;
+      r = y_fixed + cr* stbi__float2fixed(1.40200f);
+      g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000);
+      b = y_fixed                                   +   cb* stbi__float2fixed(1.77200f);
+      r >>= 20;
+      g >>= 20;
+      b >>= 20;
+      if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }
+      if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }
+      if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }
+      out[0] = (stbi_uc)r;
+      out[1] = (stbi_uc)g;
+      out[2] = (stbi_uc)b;
+      out[3] = 255;
+      out += step;
+   }
+}
+#endif
+
+// set up the kernels
+static void stbi__setup_jpeg(stbi__jpeg *j)
+{
+   j->idct_block_kernel = stbi__idct_block;
+   j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row;
+   j->resample_row_hv_2_kernel = stbi__resample_row_hv_2;
+
+#ifdef STBI_SSE2
+   if (stbi__sse2_available()) {
+      j->idct_block_kernel = stbi__idct_simd;
+      j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+      j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+   }
+#endif
+
+#ifdef STBI_NEON
+   j->idct_block_kernel = stbi__idct_simd;
+   j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;
+   j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;
+#endif
+}
+
+// clean up the temporary component buffers
+static void stbi__cleanup_jpeg(stbi__jpeg *j)
+{
+   stbi__free_jpeg_components(j, j->s->img_n, 0);
+}
+
+typedef struct
+{
+   resample_row_func resample;
+   stbi_uc *line0,*line1;
+   int hs,vs;   // expansion factor in each axis
+   int w_lores; // horizontal pixels pre-expansion
+   int ystep;   // how far through vertical expansion we are
+   int ypos;    // which pre-expansion row we're on
+} stbi__resample;
+
+// fast 0..255 * 0..255 => 0..255 rounded multiplication
+static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y)
+{
+   unsigned int t = x*y + 128;
+   return (stbi_uc) ((t + (t >>8)) >> 8);
+}
+
+static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp)
+{
+   int n, decode_n, is_rgb;
+   z->s->img_n = 0; // make stbi__cleanup_jpeg safe
+
+   // validate req_comp
+   if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+
+   // load a jpeg image from whichever source, but leave in YCbCr format
+   if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; }
+
+   // determine actual number of components to generate
+   n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1;
+
+   is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif));
+
+   if (z->s->img_n == 3 && n < 3 && !is_rgb)
+      decode_n = 1;
+   else
+      decode_n = z->s->img_n;
+
+   // resample and color-convert
+   {
+      int k;
+      unsigned int i,j;
+      stbi_uc *output;
+      stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL };
+
+      stbi__resample res_comp[4];
+
+      for (k=0; k < decode_n; ++k) {
+         stbi__resample *r = &res_comp[k];
+
+         // allocate line buffer big enough for upsampling off the edges
+         // with upsample factor of 4
+         z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3);
+         if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+         r->hs      = z->img_h_max / z->img_comp[k].h;
+         r->vs      = z->img_v_max / z->img_comp[k].v;
+         r->ystep   = r->vs >> 1;
+         r->w_lores = (z->s->img_x + r->hs-1) / r->hs;
+         r->ypos    = 0;
+         r->line0   = r->line1 = z->img_comp[k].data;
+
+         if      (r->hs == 1 && r->vs == 1) r->resample = resample_row_1;
+         else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2;
+         else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2;
+         else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel;
+         else                               r->resample = stbi__resample_row_generic;
+      }
+
+      // can't error after this so, this is safe
+      output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1);
+      if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); }
+
+      // now go ahead and resample
+      for (j=0; j < z->s->img_y; ++j) {
+         stbi_uc *out = output + n * z->s->img_x * j;
+         for (k=0; k < decode_n; ++k) {
+            stbi__resample *r = &res_comp[k];
+            int y_bot = r->ystep >= (r->vs >> 1);
+            coutput[k] = r->resample(z->img_comp[k].linebuf,
+                                     y_bot ? r->line1 : r->line0,
+                                     y_bot ? r->line0 : r->line1,
+                                     r->w_lores, r->hs);
+            if (++r->ystep >= r->vs) {
+               r->ystep = 0;
+               r->line0 = r->line1;
+               if (++r->ypos < z->img_comp[k].y)
+                  r->line1 += z->img_comp[k].w2;
+            }
+         }
+         if (n >= 3) {
+            stbi_uc *y = coutput[0];
+            if (z->s->img_n == 3) {
+               if (is_rgb) {
+                  for (i=0; i < z->s->img_x; ++i) {
+                     out[0] = y[i];
+                     out[1] = coutput[1][i];
+                     out[2] = coutput[2][i];
+                     out[3] = 255;
+                     out += n;
+                  }
+               } else {
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+               }
+            } else if (z->s->img_n == 4) {
+               if (z->app14_color_transform == 0) { // CMYK
+                  for (i=0; i < z->s->img_x; ++i) {
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(coutput[0][i], m);
+                     out[1] = stbi__blinn_8x8(coutput[1][i], m);
+                     out[2] = stbi__blinn_8x8(coutput[2][i], m);
+                     out[3] = 255;
+                     out += n;
+                  }
+               } else if (z->app14_color_transform == 2) { // YCCK
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+                  for (i=0; i < z->s->img_x; ++i) {
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(255 - out[0], m);
+                     out[1] = stbi__blinn_8x8(255 - out[1], m);
+                     out[2] = stbi__blinn_8x8(255 - out[2], m);
+                     out += n;
+                  }
+               } else { // YCbCr + alpha?  Ignore the fourth channel for now
+                  z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
+               }
+            } else
+               for (i=0; i < z->s->img_x; ++i) {
+                  out[0] = out[1] = out[2] = y[i];
+                  out[3] = 255; // not used if n==3
+                  out += n;
+               }
+         } else {
+            if (is_rgb) {
+               if (n == 1)
+                  for (i=0; i < z->s->img_x; ++i)
+                     *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+               else {
+                  for (i=0; i < z->s->img_x; ++i, out += 2) {
+                     out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]);
+                     out[1] = 255;
+                  }
+               }
+            } else if (z->s->img_n == 4 && z->app14_color_transform == 0) {
+               for (i=0; i < z->s->img_x; ++i) {
+                  stbi_uc m = coutput[3][i];
+                  stbi_uc r = stbi__blinn_8x8(coutput[0][i], m);
+                  stbi_uc g = stbi__blinn_8x8(coutput[1][i], m);
+                  stbi_uc b = stbi__blinn_8x8(coutput[2][i], m);
+                  out[0] = stbi__compute_y(r, g, b);
+                  out[1] = 255;
+                  out += n;
+               }
+            } else if (z->s->img_n == 4 && z->app14_color_transform == 2) {
+               for (i=0; i < z->s->img_x; ++i) {
+                  out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]);
+                  out[1] = 255;
+                  out += n;
+               }
+            } else {
+               stbi_uc *y = coutput[0];
+               if (n == 1)
+                  for (i=0; i < z->s->img_x; ++i) out[i] = y[i];
+               else
+                  for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; }
+            }
+         }
+      }
+      stbi__cleanup_jpeg(z);
+      *out_x = z->s->img_x;
+      *out_y = z->s->img_y;
+      if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output
+      return output;
+   }
+}
+
+static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   unsigned char* result;
+   stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg));
+   STBI_NOTUSED(ri);
+   j->s = s;
+   stbi__setup_jpeg(j);
+   result = load_jpeg_image(j, x,y,comp,req_comp);
+   STBI_FREE(j);
+   return result;
+}
+
+static int stbi__jpeg_test(stbi__context *s)
+{
+   int r;
+   stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg));
+   j->s = s;
+   stbi__setup_jpeg(j);
+   r = stbi__decode_jpeg_header(j, STBI__SCAN_type);
+   stbi__rewind(s);
+   STBI_FREE(j);
+   return r;
+}
+
+static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp)
+{
+   if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) {
+      stbi__rewind( j->s );
+      return 0;
+   }
+   if (x) *x = j->s->img_x;
+   if (y) *y = j->s->img_y;
+   if (comp) *comp = j->s->img_n >= 3 ? 3 : 1;
+   return 1;
+}
+
+static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int result;
+   stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg)));
+   j->s = s;
+   result = stbi__jpeg_info_raw(j, x, y, comp);
+   STBI_FREE(j);
+   return result;
+}
+#endif
+
+// public domain zlib decode    v0.2  Sean Barrett 2006-11-18
+//    simple implementation
+//      - all input must be provided in an upfront buffer
+//      - all output is written to a single output buffer (can malloc/realloc)
+//    performance
+//      - fast huffman
+
+#ifndef STBI_NO_ZLIB
+
+// fast-way is faster to check than jpeg huffman, but slow way is slower
+#define STBI__ZFAST_BITS  9 // accelerate all cases in default tables
+#define STBI__ZFAST_MASK  ((1 << STBI__ZFAST_BITS) - 1)
+
+// zlib-style huffman encoding
+// (jpegs packs from left, zlib from right, so can't share code)
+typedef struct
+{
+   stbi__uint16 fast[1 << STBI__ZFAST_BITS];
+   stbi__uint16 firstcode[16];
+   int maxcode[17];
+   stbi__uint16 firstsymbol[16];
+   stbi_uc  size[288];
+   stbi__uint16 value[288];
+} stbi__zhuffman;
+
+stbi_inline static int stbi__bitreverse16(int n)
+{
+  n = ((n & 0xAAAA) >>  1) | ((n & 0x5555) << 1);
+  n = ((n & 0xCCCC) >>  2) | ((n & 0x3333) << 2);
+  n = ((n & 0xF0F0) >>  4) | ((n & 0x0F0F) << 4);
+  n = ((n & 0xFF00) >>  8) | ((n & 0x00FF) << 8);
+  return n;
+}
+
+stbi_inline static int stbi__bit_reverse(int v, int bits)
+{
+   STBI_ASSERT(bits <= 16);
+   // to bit reverse n bits, reverse 16 and shift
+   // e.g. 11 bits, bit reverse and shift away 5
+   return stbi__bitreverse16(v) >> (16-bits);
+}
+
+static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num)
+{
+   int i,k=0;
+   int code, next_code[16], sizes[17];
+
+   // DEFLATE spec for generating codes
+   memset(sizes, 0, sizeof(sizes));
+   memset(z->fast, 0, sizeof(z->fast));
+   for (i=0; i < num; ++i)
+      ++sizes[sizelist[i]];
+   sizes[0] = 0;
+   for (i=1; i < 16; ++i)
+      if (sizes[i] > (1 << i))
+         return stbi__err("bad sizes", "Corrupt PNG");
+   code = 0;
+   for (i=1; i < 16; ++i) {
+      next_code[i] = code;
+      z->firstcode[i] = (stbi__uint16) code;
+      z->firstsymbol[i] = (stbi__uint16) k;
+      code = (code + sizes[i]);
+      if (sizes[i])
+         if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG");
+      z->maxcode[i] = code << (16-i); // preshift for inner loop
+      code <<= 1;
+      k += sizes[i];
+   }
+   z->maxcode[16] = 0x10000; // sentinel
+   for (i=0; i < num; ++i) {
+      int s = sizelist[i];
+      if (s) {
+         int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];
+         stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i);
+         z->size [c] = (stbi_uc     ) s;
+         z->value[c] = (stbi__uint16) i;
+         if (s <= STBI__ZFAST_BITS) {
+            int j = stbi__bit_reverse(next_code[s],s);
+            while (j < (1 << STBI__ZFAST_BITS)) {
+               z->fast[j] = fastv;
+               j += (1 << s);
+            }
+         }
+         ++next_code[s];
+      }
+   }
+   return 1;
+}
+
+// zlib-from-memory implementation for PNG reading
+//    because PNG allows splitting the zlib stream arbitrarily,
+//    and it's annoying structurally to have PNG call ZLIB call PNG,
+//    we require PNG read all the IDATs and combine them into a single
+//    memory buffer
+
+typedef struct
+{
+   stbi_uc *zbuffer, *zbuffer_end;
+   int num_bits;
+   stbi__uint32 code_buffer;
+
+   char *zout;
+   char *zout_start;
+   char *zout_end;
+   int   z_expandable;
+
+   stbi__zhuffman z_length, z_distance;
+} stbi__zbuf;
+
+stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z)
+{
+   if (z->zbuffer >= z->zbuffer_end) return 0;
+   return *z->zbuffer++;
+}
+
+static void stbi__fill_bits(stbi__zbuf *z)
+{
+   do {
+      STBI_ASSERT(z->code_buffer < (1U << z->num_bits));
+      z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits;
+      z->num_bits += 8;
+   } while (z->num_bits <= 24);
+}
+
+stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n)
+{
+   unsigned int k;
+   if (z->num_bits < n) stbi__fill_bits(z);
+   k = z->code_buffer & ((1 << n) - 1);
+   z->code_buffer >>= n;
+   z->num_bits -= n;
+   return k;
+}
+
+static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)
+{
+   int b,s,k;
+   // not resolved by fast table, so compute it the slow way
+   // use jpeg approach, which requires MSbits at top
+   k = stbi__bit_reverse(a->code_buffer, 16);
+   for (s=STBI__ZFAST_BITS+1; ; ++s)
+      if (k < z->maxcode[s])
+         break;
+   if (s == 16) return -1; // invalid code!
+   // code size is s, so:
+   b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];
+   STBI_ASSERT(z->size[b] == s);
+   a->code_buffer >>= s;
+   a->num_bits -= s;
+   return z->value[b];
+}
+
+stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)
+{
+   int b,s;
+   if (a->num_bits < 16) stbi__fill_bits(a);
+   b = z->fast[a->code_buffer & STBI__ZFAST_MASK];
+   if (b) {
+      s = b >> 9;
+      a->code_buffer >>= s;
+      a->num_bits -= s;
+      return b & 511;
+   }
+   return stbi__zhuffman_decode_slowpath(a, z);
+}
+
+static int stbi__zexpand(stbi__zbuf *z, char *zout, int n)  // need to make room for n bytes
+{
+   char *q;
+   int cur, limit, old_limit;
+   z->zout = zout;
+   if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG");
+   cur   = (int) (z->zout     - z->zout_start);
+   limit = old_limit = (int) (z->zout_end - z->zout_start);
+   while (cur + n > limit)
+      limit *= 2;
+   q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit);
+   STBI_NOTUSED(old_limit);
+   if (q == NULL) return stbi__err("outofmem", "Out of memory");
+   z->zout_start = q;
+   z->zout       = q + cur;
+   z->zout_end   = q + limit;
+   return 1;
+}
+
+static const int stbi__zlength_base[31] = {
+   3,4,5,6,7,8,9,10,11,13,
+   15,17,19,23,27,31,35,43,51,59,
+   67,83,99,115,131,163,195,227,258,0,0 };
+
+static const int stbi__zlength_extra[31]=
+{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };
+
+static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,
+257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};
+
+static const int stbi__zdist_extra[32] =
+{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+static int stbi__parse_huffman_block(stbi__zbuf *a)
+{
+   char *zout = a->zout;
+   for(;;) {
+      int z = stbi__zhuffman_decode(a, &a->z_length);
+      if (z < 256) {
+         if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes
+         if (zout >= a->zout_end) {
+            if (!stbi__zexpand(a, zout, 1)) return 0;
+            zout = a->zout;
+         }
+         *zout++ = (char) z;
+      } else {
+         stbi_uc *p;
+         int len,dist;
+         if (z == 256) {
+            a->zout = zout;
+            return 1;
+         }
+         z -= 257;
+         len = stbi__zlength_base[z];
+         if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);
+         z = stbi__zhuffman_decode(a, &a->z_distance);
+         if (z < 0) return stbi__err("bad huffman code","Corrupt PNG");
+         dist = stbi__zdist_base[z];
+         if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);
+         if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG");
+         if (zout + len > a->zout_end) {
+            if (!stbi__zexpand(a, zout, len)) return 0;
+            zout = a->zout;
+         }
+         p = (stbi_uc *) (zout - dist);
+         if (dist == 1) { // run of one byte; common in images.
+            stbi_uc v = *p;
+            if (len) { do *zout++ = v; while (--len); }
+         } else {
+            if (len) { do *zout++ = *p++; while (--len); }
+         }
+      }
+   }
+}
+
+static int stbi__compute_huffman_codes(stbi__zbuf *a)
+{
+   static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };
+   stbi__zhuffman z_codelength;
+   stbi_uc lencodes[286+32+137];//padding for maximum single op
+   stbi_uc codelength_sizes[19];
+   int i,n;
+
+   int hlit  = stbi__zreceive(a,5) + 257;
+   int hdist = stbi__zreceive(a,5) + 1;
+   int hclen = stbi__zreceive(a,4) + 4;
+   int ntot  = hlit + hdist;
+
+   memset(codelength_sizes, 0, sizeof(codelength_sizes));
+   for (i=0; i < hclen; ++i) {
+      int s = stbi__zreceive(a,3);
+      codelength_sizes[length_dezigzag[i]] = (stbi_uc) s;
+   }
+   if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;
+
+   n = 0;
+   while (n < ntot) {
+      int c = stbi__zhuffman_decode(a, &z_codelength);
+      if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG");
+      if (c < 16)
+         lencodes[n++] = (stbi_uc) c;
+      else {
+         stbi_uc fill = 0;
+         if (c == 16) {
+            c = stbi__zreceive(a,2)+3;
+            if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG");
+            fill = lencodes[n-1];
+         } else if (c == 17)
+            c = stbi__zreceive(a,3)+3;
+         else {
+            STBI_ASSERT(c == 18);
+            c = stbi__zreceive(a,7)+11;
+         }
+         if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG");
+         memset(lencodes+n, fill, c);
+         n += c;
+      }
+   }
+   if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG");
+   if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;
+   if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;
+   return 1;
+}
+
+static int stbi__parse_uncompressed_block(stbi__zbuf *a)
+{
+   stbi_uc header[4];
+   int len,nlen,k;
+   if (a->num_bits & 7)
+      stbi__zreceive(a, a->num_bits & 7); // discard
+   // drain the bit-packed data into header
+   k = 0;
+   while (a->num_bits > 0) {
+      header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check
+      a->code_buffer >>= 8;
+      a->num_bits -= 8;
+   }
+   STBI_ASSERT(a->num_bits == 0);
+   // now fill header the normal way
+   while (k < 4)
+      header[k++] = stbi__zget8(a);
+   len  = header[1] * 256 + header[0];
+   nlen = header[3] * 256 + header[2];
+   if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG");
+   if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG");
+   if (a->zout + len > a->zout_end)
+      if (!stbi__zexpand(a, a->zout, len)) return 0;
+   memcpy(a->zout, a->zbuffer, len);
+   a->zbuffer += len;
+   a->zout += len;
+   return 1;
+}
+
+static int stbi__parse_zlib_header(stbi__zbuf *a)
+{
+   int cmf   = stbi__zget8(a);
+   int cm    = cmf & 15;
+   /* int cinfo = cmf >> 4; */
+   int flg   = stbi__zget8(a);
+   if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
+   if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png
+   if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png
+   // window = 1 << (8 + cinfo)... but who cares, we fully buffer output
+   return 1;
+}
+
+static const stbi_uc stbi__zdefault_length[288] =
+{
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
+   8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8
+};
+static const stbi_uc stbi__zdefault_distance[32] =
+{
+   5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5
+};
+/*
+Init algorithm:
+{
+   int i;   // use <= to match clearly with spec
+   for (i=0; i <= 143; ++i)     stbi__zdefault_length[i]   = 8;
+   for (   ; i <= 255; ++i)     stbi__zdefault_length[i]   = 9;
+   for (   ; i <= 279; ++i)     stbi__zdefault_length[i]   = 7;
+   for (   ; i <= 287; ++i)     stbi__zdefault_length[i]   = 8;
+
+   for (i=0; i <=  31; ++i)     stbi__zdefault_distance[i] = 5;
+}
+*/
+
+static int stbi__parse_zlib(stbi__zbuf *a, int parse_header)
+{
+   int final, type;
+   if (parse_header)
+      if (!stbi__parse_zlib_header(a)) return 0;
+   a->num_bits = 0;
+   a->code_buffer = 0;
+   do {
+      final = stbi__zreceive(a,1);
+      type = stbi__zreceive(a,2);
+      if (type == 0) {
+         if (!stbi__parse_uncompressed_block(a)) return 0;
+      } else if (type == 3) {
+         return 0;
+      } else {
+         if (type == 1) {
+            // use fixed code lengths
+            if (!stbi__zbuild_huffman(&a->z_length  , stbi__zdefault_length  , 288)) return 0;
+            if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance,  32)) return 0;
+         } else {
+            if (!stbi__compute_huffman_codes(a)) return 0;
+         }
+         if (!stbi__parse_huffman_block(a)) return 0;
+      }
+   } while (!final);
+   return 1;
+}
+
+static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)
+{
+   a->zout_start = obuf;
+   a->zout       = obuf;
+   a->zout_end   = obuf + olen;
+   a->z_expandable = exp;
+
+   return stbi__parse_zlib(a, parse_header);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(initial_size);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer + len;
+   if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)
+{
+   return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);
+}
+
+STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(initial_size);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer + len;
+   if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)
+{
+   stbi__zbuf a;
+   a.zbuffer = (stbi_uc *) ibuffer;
+   a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+   if (stbi__do_zlib(&a, obuffer, olen, 0, 1))
+      return (int) (a.zout - a.zout_start);
+   else
+      return -1;
+}
+
+STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)
+{
+   stbi__zbuf a;
+   char *p = (char *) stbi__malloc(16384);
+   if (p == NULL) return NULL;
+   a.zbuffer = (stbi_uc *) buffer;
+   a.zbuffer_end = (stbi_uc *) buffer+len;
+   if (stbi__do_zlib(&a, p, 16384, 1, 0)) {
+      if (outlen) *outlen = (int) (a.zout - a.zout_start);
+      return a.zout_start;
+   } else {
+      STBI_FREE(a.zout_start);
+      return NULL;
+   }
+}
+
+STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)
+{
+   stbi__zbuf a;
+   a.zbuffer = (stbi_uc *) ibuffer;
+   a.zbuffer_end = (stbi_uc *) ibuffer + ilen;
+   if (stbi__do_zlib(&a, obuffer, olen, 0, 0))
+      return (int) (a.zout - a.zout_start);
+   else
+      return -1;
+}
+#endif
+
+// public domain "baseline" PNG decoder   v0.10  Sean Barrett 2006-11-18
+//    simple implementation
+//      - only 8-bit samples
+//      - no CRC checking
+//      - allocates lots of intermediate memory
+//        - avoids problem of streaming data between subsystems
+//        - avoids explicit window management
+//    performance
+//      - uses stb_zlib, a PD zlib implementation with fast huffman decoding
+
+#ifndef STBI_NO_PNG
+typedef struct
+{
+   stbi__uint32 length;
+   stbi__uint32 type;
+} stbi__pngchunk;
+
+static stbi__pngchunk stbi__get_chunk_header(stbi__context *s)
+{
+   stbi__pngchunk c;
+   c.length = stbi__get32be(s);
+   c.type   = stbi__get32be(s);
+   return c;
+}
+
+static int stbi__check_png_header(stbi__context *s)
+{
+   static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 };
+   int i;
+   for (i=0; i < 8; ++i)
+      if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG");
+   return 1;
+}
+
+typedef struct
+{
+   stbi__context *s;
+   stbi_uc *idata, *expanded, *out;
+   int depth;
+} stbi__png;
+
+
+enum {
+   STBI__F_none=0,
+   STBI__F_sub=1,
+   STBI__F_up=2,
+   STBI__F_avg=3,
+   STBI__F_paeth=4,
+   // synthetic filters used for first scanline to avoid needing a dummy row of 0s
+   STBI__F_avg_first,
+   STBI__F_paeth_first
+};
+
+static stbi_uc first_row_filter[5] =
+{
+   STBI__F_none,
+   STBI__F_sub,
+   STBI__F_none,
+   STBI__F_avg_first,
+   STBI__F_paeth_first
+};
+
+static int stbi__paeth(int a, int b, int c)
+{
+   int p = a + b - c;
+   int pa = abs(p-a);
+   int pb = abs(p-b);
+   int pc = abs(p-c);
+   if (pa <= pb && pa <= pc) return a;
+   if (pb <= pc) return b;
+   return c;
+}
+
+static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 };
+
+// create the png data from post-deflated data
+static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color)
+{
+   int bytes = (depth == 16? 2 : 1);
+   stbi__context *s = a->s;
+   stbi__uint32 i,j,stride = x*out_n*bytes;
+   stbi__uint32 img_len, img_width_bytes;
+   int k;
+   int img_n = s->img_n; // copy it into a local for later
+
+   int output_bytes = out_n*bytes;
+   int filter_bytes = img_n*bytes;
+   int width = x;
+
+   STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1);
+   a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into
+   if (!a->out) return stbi__err("outofmem", "Out of memory");
+
+   if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG");
+   img_width_bytes = (((img_n * x * depth) + 7) >> 3);
+   img_len = (img_width_bytes + 1) * y;
+
+   // we used to check for exact match between raw_len and img_len on non-interlaced PNGs,
+   // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),
+   // so just check for raw_len < img_len always.
+   if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
+
+   for (j=0; j < y; ++j) {
+      stbi_uc *cur = a->out + stride*j;
+      stbi_uc *prior;
+      int filter = *raw++;
+
+      if (filter > 4)
+         return stbi__err("invalid filter","Corrupt PNG");
+
+      if (depth < 8) {
+         STBI_ASSERT(img_width_bytes <= x);
+         cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place
+         filter_bytes = 1;
+         width = img_width_bytes;
+      }
+      prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above
+
+      // if first row, use special filter that doesn't sample previous row
+      if (j == 0) filter = first_row_filter[filter];
+
+      // handle first byte explicitly
+      for (k=0; k < filter_bytes; ++k) {
+         switch (filter) {
+            case STBI__F_none       : cur[k] = raw[k]; break;
+            case STBI__F_sub        : cur[k] = raw[k]; break;
+            case STBI__F_up         : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;
+            case STBI__F_avg        : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break;
+            case STBI__F_paeth      : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;
+            case STBI__F_avg_first  : cur[k] = raw[k]; break;
+            case STBI__F_paeth_first: cur[k] = raw[k]; break;
+         }
+      }
+
+      if (depth == 8) {
+         if (img_n != out_n)
+            cur[img_n] = 255; // first pixel
+         raw += img_n;
+         cur += out_n;
+         prior += out_n;
+      } else if (depth == 16) {
+         if (img_n != out_n) {
+            cur[filter_bytes]   = 255; // first pixel top byte
+            cur[filter_bytes+1] = 255; // first pixel bottom byte
+         }
+         raw += filter_bytes;
+         cur += output_bytes;
+         prior += output_bytes;
+      } else {
+         raw += 1;
+         cur += 1;
+         prior += 1;
+      }
+
+      // this is a little gross, so that we don't switch per-pixel or per-component
+      if (depth < 8 || img_n == out_n) {
+         int nk = (width - 1)*filter_bytes;
+         #define STBI__CASE(f) \
+             case f:     \
+                for (k=0; k < nk; ++k)
+         switch (filter) {
+            // "none" filter turns into a memcpy here; make that explicit.
+            case STBI__F_none:         memcpy(cur, raw, nk); break;
+            STBI__CASE(STBI__F_sub)          { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break;
+            STBI__CASE(STBI__F_up)           { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+            STBI__CASE(STBI__F_avg)          { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break;
+            STBI__CASE(STBI__F_paeth)        { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break;
+            STBI__CASE(STBI__F_avg_first)    { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break;
+            STBI__CASE(STBI__F_paeth_first)  { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break;
+         }
+         #undef STBI__CASE
+         raw += nk;
+      } else {
+         STBI_ASSERT(img_n+1 == out_n);
+         #define STBI__CASE(f) \
+             case f:     \
+                for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \
+                   for (k=0; k < filter_bytes; ++k)
+         switch (filter) {
+            STBI__CASE(STBI__F_none)         { cur[k] = raw[k]; } break;
+            STBI__CASE(STBI__F_sub)          { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break;
+            STBI__CASE(STBI__F_up)           { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
+            STBI__CASE(STBI__F_avg)          { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break;
+            STBI__CASE(STBI__F_paeth)        { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break;
+            STBI__CASE(STBI__F_avg_first)    { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break;
+            STBI__CASE(STBI__F_paeth_first)  { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break;
+         }
+         #undef STBI__CASE
+
+         // the loop above sets the high byte of the pixels' alpha, but for
+         // 16 bit png files we also need the low byte set. we'll do that here.
+         if (depth == 16) {
+            cur = a->out + stride*j; // start at the beginning of the row again
+            for (i=0; i < x; ++i,cur+=output_bytes) {
+               cur[filter_bytes+1] = 255;
+            }
+         }
+      }
+   }
+
+   // we make a separate pass to expand bits to pixels; for performance,
+   // this could run two scanlines behind the above code, so it won't
+   // intefere with filtering but will still be in the cache.
+   if (depth < 8) {
+      for (j=0; j < y; ++j) {
+         stbi_uc *cur = a->out + stride*j;
+         stbi_uc *in  = a->out + stride*j + x*out_n - img_width_bytes;
+         // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit
+         // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop
+         stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range
+
+         // note that the final byte might overshoot and write more data than desired.
+         // we can allocate enough data that this never writes out of memory, but it
+         // could also overwrite the next scanline. can it overwrite non-empty data
+         // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel.
+         // so we need to explicitly clamp the final ones
+
+         if (depth == 4) {
+            for (k=x*img_n; k >= 2; k-=2, ++in) {
+               *cur++ = scale * ((*in >> 4)       );
+               *cur++ = scale * ((*in     ) & 0x0f);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 4)       );
+         } else if (depth == 2) {
+            for (k=x*img_n; k >= 4; k-=4, ++in) {
+               *cur++ = scale * ((*in >> 6)       );
+               *cur++ = scale * ((*in >> 4) & 0x03);
+               *cur++ = scale * ((*in >> 2) & 0x03);
+               *cur++ = scale * ((*in     ) & 0x03);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 6)       );
+            if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03);
+            if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03);
+         } else if (depth == 1) {
+            for (k=x*img_n; k >= 8; k-=8, ++in) {
+               *cur++ = scale * ((*in >> 7)       );
+               *cur++ = scale * ((*in >> 6) & 0x01);
+               *cur++ = scale * ((*in >> 5) & 0x01);
+               *cur++ = scale * ((*in >> 4) & 0x01);
+               *cur++ = scale * ((*in >> 3) & 0x01);
+               *cur++ = scale * ((*in >> 2) & 0x01);
+               *cur++ = scale * ((*in >> 1) & 0x01);
+               *cur++ = scale * ((*in     ) & 0x01);
+            }
+            if (k > 0) *cur++ = scale * ((*in >> 7)       );
+            if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01);
+            if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01);
+            if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01);
+            if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01);
+            if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01);
+            if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01);
+         }
+         if (img_n != out_n) {
+            int q;
+            // insert alpha = 255
+            cur = a->out + stride*j;
+            if (img_n == 1) {
+               for (q=x-1; q >= 0; --q) {
+                  cur[q*2+1] = 255;
+                  cur[q*2+0] = cur[q];
+               }
+            } else {
+               STBI_ASSERT(img_n == 3);
+               for (q=x-1; q >= 0; --q) {
+                  cur[q*4+3] = 255;
+                  cur[q*4+2] = cur[q*3+2];
+                  cur[q*4+1] = cur[q*3+1];
+                  cur[q*4+0] = cur[q*3+0];
+               }
+            }
+         }
+      }
+   } else if (depth == 16) {
+      // force the image data from big-endian to platform-native.
+      // this is done in a separate pass due to the decoding relying
+      // on the data being untouched, but could probably be done
+      // per-line during decode if care is taken.
+      stbi_uc *cur = a->out;
+      stbi__uint16 *cur16 = (stbi__uint16*)cur;
+
+      for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) {
+         *cur16 = (cur[0] << 8) | cur[1];
+      }
+   }
+
+   return 1;
+}
+
+static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced)
+{
+   int bytes = (depth == 16 ? 2 : 1);
+   int out_bytes = out_n * bytes;
+   stbi_uc *final;
+   int p;
+   if (!interlaced)
+      return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);
+
+   // de-interlacing
+   final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0);
+   for (p=0; p < 7; ++p) {
+      int xorig[] = { 0,4,0,2,0,1,0 };
+      int yorig[] = { 0,0,4,0,2,0,1 };
+      int xspc[]  = { 8,8,4,4,2,2,1 };
+      int yspc[]  = { 8,8,8,4,4,2,2 };
+      int i,j,x,y;
+      // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1
+      x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];
+      y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];
+      if (x && y) {
+         stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;
+         if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {
+            STBI_FREE(final);
+            return 0;
+         }
+         for (j=0; j < y; ++j) {
+            for (i=0; i < x; ++i) {
+               int out_y = j*yspc[p]+yorig[p];
+               int out_x = i*xspc[p]+xorig[p];
+               memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes,
+                      a->out + (j*x+i)*out_bytes, out_bytes);
+            }
+         }
+         STBI_FREE(a->out);
+         image_data += img_len;
+         image_data_len -= img_len;
+      }
+   }
+   a->out = final;
+
+   return 1;
+}
+
+static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi_uc *p = z->out;
+
+   // compute color-based transparency, assuming we've
+   // already got 255 as the alpha value in the output
+   STBI_ASSERT(out_n == 2 || out_n == 4);
+
+   if (out_n == 2) {
+      for (i=0; i < pixel_count; ++i) {
+         p[1] = (p[0] == tc[0] ? 0 : 255);
+         p += 2;
+      }
+   } else {
+      for (i=0; i < pixel_count; ++i) {
+         if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+            p[3] = 0;
+         p += 4;
+      }
+   }
+   return 1;
+}
+
+static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi__uint16 *p = (stbi__uint16*) z->out;
+
+   // compute color-based transparency, assuming we've
+   // already got 65535 as the alpha value in the output
+   STBI_ASSERT(out_n == 2 || out_n == 4);
+
+   if (out_n == 2) {
+      for (i = 0; i < pixel_count; ++i) {
+         p[1] = (p[0] == tc[0] ? 0 : 65535);
+         p += 2;
+      }
+   } else {
+      for (i = 0; i < pixel_count; ++i) {
+         if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
+            p[3] = 0;
+         p += 4;
+      }
+   }
+   return 1;
+}
+
+static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n)
+{
+   stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y;
+   stbi_uc *p, *temp_out, *orig = a->out;
+
+   p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0);
+   if (p == NULL) return stbi__err("outofmem", "Out of memory");
+
+   // between here and free(out) below, exitting would leak
+   temp_out = p;
+
+   if (pal_img_n == 3) {
+      for (i=0; i < pixel_count; ++i) {
+         int n = orig[i]*4;
+         p[0] = palette[n  ];
+         p[1] = palette[n+1];
+         p[2] = palette[n+2];
+         p += 3;
+      }
+   } else {
+      for (i=0; i < pixel_count; ++i) {
+         int n = orig[i]*4;
+         p[0] = palette[n  ];
+         p[1] = palette[n+1];
+         p[2] = palette[n+2];
+         p[3] = palette[n+3];
+         p += 4;
+      }
+   }
+   STBI_FREE(a->out);
+   a->out = temp_out;
+
+   STBI_NOTUSED(len);
+
+   return 1;
+}
+
+static int stbi__unpremultiply_on_load = 0;
+static int stbi__de_iphone_flag = 0;
+
+STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply)
+{
+   stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply;
+}
+
+STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert)
+{
+   stbi__de_iphone_flag = flag_true_if_should_convert;
+}
+
+static void stbi__de_iphone(stbi__png *z)
+{
+   stbi__context *s = z->s;
+   stbi__uint32 i, pixel_count = s->img_x * s->img_y;
+   stbi_uc *p = z->out;
+
+   if (s->img_out_n == 3) {  // convert bgr to rgb
+      for (i=0; i < pixel_count; ++i) {
+         stbi_uc t = p[0];
+         p[0] = p[2];
+         p[2] = t;
+         p += 3;
+      }
+   } else {
+      STBI_ASSERT(s->img_out_n == 4);
+      if (stbi__unpremultiply_on_load) {
+         // convert bgr to rgb and unpremultiply
+         for (i=0; i < pixel_count; ++i) {
+            stbi_uc a = p[3];
+            stbi_uc t = p[0];
+            if (a) {
+               stbi_uc half = a / 2;
+               p[0] = (p[2] * 255 + half) / a;
+               p[1] = (p[1] * 255 + half) / a;
+               p[2] = ( t   * 255 + half) / a;
+            } else {
+               p[0] = p[2];
+               p[2] = t;
+            }
+            p += 4;
+         }
+      } else {
+         // convert bgr to rgb
+         for (i=0; i < pixel_count; ++i) {
+            stbi_uc t = p[0];
+            p[0] = p[2];
+            p[2] = t;
+            p += 4;
+         }
+      }
+   }
+}
+
+#define STBI__PNG_TYPE(a,b,c,d)  (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d))
+
+static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)
+{
+   stbi_uc palette[1024], pal_img_n=0;
+   stbi_uc has_trans=0, tc[3]={0};
+   stbi__uint16 tc16[3];
+   stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0;
+   int first=1,k,interlace=0, color=0, is_iphone=0;
+   stbi__context *s = z->s;
+
+   z->expanded = NULL;
+   z->idata = NULL;
+   z->out = NULL;
+
+   if (!stbi__check_png_header(s)) return 0;
+
+   if (scan == STBI__SCAN_type) return 1;
+
+   for (;;) {
+      stbi__pngchunk c = stbi__get_chunk_header(s);
+      switch (c.type) {
+         case STBI__PNG_TYPE('C','g','B','I'):
+            is_iphone = 1;
+            stbi__skip(s, c.length);
+            break;
+         case STBI__PNG_TYPE('I','H','D','R'): {
+            int comp,filter;
+            if (!first) return stbi__err("multiple IHDR","Corrupt PNG");
+            first = 0;
+            if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG");
+            s->img_x = stbi__get32be(s); if (s->img_x > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");
+            s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)");
+            z->depth = stbi__get8(s);  if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16)  return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only");
+            color = stbi__get8(s);  if (color > 6)         return stbi__err("bad ctype","Corrupt PNG");
+            if (color == 3 && z->depth == 16)                  return stbi__err("bad ctype","Corrupt PNG");
+            if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG");
+            comp  = stbi__get8(s);  if (comp) return stbi__err("bad comp method","Corrupt PNG");
+            filter= stbi__get8(s);  if (filter) return stbi__err("bad filter method","Corrupt PNG");
+            interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG");
+            if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG");
+            if (!pal_img_n) {
+               s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);
+               if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode");
+               if (scan == STBI__SCAN_header) return 1;
+            } else {
+               // if paletted, then pal_n is our final components, and
+               // img_n is # components to decompress/filter.
+               s->img_n = 1;
+               if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG");
+               // if SCAN_header, have to scan to see if we have a tRNS
+            }
+            break;
+         }
+
+         case STBI__PNG_TYPE('P','L','T','E'):  {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG");
+            pal_len = c.length / 3;
+            if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG");
+            for (i=0; i < pal_len; ++i) {
+               palette[i*4+0] = stbi__get8(s);
+               palette[i*4+1] = stbi__get8(s);
+               palette[i*4+2] = stbi__get8(s);
+               palette[i*4+3] = 255;
+            }
+            break;
+         }
+
+         case STBI__PNG_TYPE('t','R','N','S'): {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG");
+            if (pal_img_n) {
+               if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }
+               if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG");
+               if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG");
+               pal_img_n = 4;
+               for (i=0; i < c.length; ++i)
+                  palette[i*4+3] = stbi__get8(s);
+            } else {
+               if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG");
+               if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG");
+               has_trans = 1;
+               if (z->depth == 16) {
+                  for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is
+               } else {
+                  for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger
+               }
+            }
+            break;
+         }
+
+         case STBI__PNG_TYPE('I','D','A','T'): {
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG");
+            if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; }
+            if ((int)(ioff + c.length) < (int)ioff) return 0;
+            if (ioff + c.length > idata_limit) {
+               stbi__uint32 idata_limit_old = idata_limit;
+               stbi_uc *p;
+               if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;
+               while (ioff + c.length > idata_limit)
+                  idata_limit *= 2;
+               STBI_NOTUSED(idata_limit_old);
+               p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory");
+               z->idata = p;
+            }
+            if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG");
+            ioff += c.length;
+            break;
+         }
+
+         case STBI__PNG_TYPE('I','E','N','D'): {
+            stbi__uint32 raw_len, bpl;
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if (scan != STBI__SCAN_load) return 1;
+            if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG");
+            // initial guess for decoded data size to avoid unnecessary reallocs
+            bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component
+            raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;
+            z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);
+            if (z->expanded == NULL) return 0; // zlib should set error
+            STBI_FREE(z->idata); z->idata = NULL;
+            if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans)
+               s->img_out_n = s->img_n+1;
+            else
+               s->img_out_n = s->img_n;
+            if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0;
+            if (has_trans) {
+               if (z->depth == 16) {
+                  if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0;
+               } else {
+                  if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;
+               }
+            }
+            if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)
+               stbi__de_iphone(z);
+            if (pal_img_n) {
+               // pal_img_n == 3 or 4
+               s->img_n = pal_img_n; // record the actual colors we had
+               s->img_out_n = pal_img_n;
+               if (req_comp >= 3) s->img_out_n = req_comp;
+               if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))
+                  return 0;
+            } else if (has_trans) {
+               // non-paletted image with tRNS -> source image has (constant) alpha
+               ++s->img_n;
+            }
+            STBI_FREE(z->expanded); z->expanded = NULL;
+            return 1;
+         }
+
+         default:
+            // if critical, fail
+            if (first) return stbi__err("first not IHDR", "Corrupt PNG");
+            if ((c.type & (1 << 29)) == 0) {
+               #ifndef STBI_NO_FAILURE_STRINGS
+               // not threadsafe
+               static char invalid_chunk[] = "XXXX PNG chunk not known";
+               invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);
+               invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);
+               invalid_chunk[2] = STBI__BYTECAST(c.type >>  8);
+               invalid_chunk[3] = STBI__BYTECAST(c.type >>  0);
+               #endif
+               return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type");
+            }
+            stbi__skip(s, c.length);
+            break;
+      }
+      // end of PNG chunk, read and skip CRC
+      stbi__get32be(s);
+   }
+}
+
+static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri)
+{
+   void *result=NULL;
+   if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error");
+   if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) {
+      if (p->depth < 8)
+         ri->bits_per_channel = 8;
+      else
+         ri->bits_per_channel = p->depth;
+      result = p->out;
+      p->out = NULL;
+      if (req_comp && req_comp != p->s->img_out_n) {
+         if (ri->bits_per_channel == 8)
+            result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+         else
+            result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);
+         p->s->img_out_n = req_comp;
+         if (result == NULL) return result;
+      }
+      *x = p->s->img_x;
+      *y = p->s->img_y;
+      if (n) *n = p->s->img_n;
+   }
+   STBI_FREE(p->out);      p->out      = NULL;
+   STBI_FREE(p->expanded); p->expanded = NULL;
+   STBI_FREE(p->idata);    p->idata    = NULL;
+
+   return result;
+}
+
+static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi__png p;
+   p.s = s;
+   return stbi__do_png(&p, x,y,comp,req_comp, ri);
+}
+
+static int stbi__png_test(stbi__context *s)
+{
+   int r;
+   r = stbi__check_png_header(s);
+   stbi__rewind(s);
+   return r;
+}
+
+static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp)
+{
+   if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) {
+      stbi__rewind( p->s );
+      return 0;
+   }
+   if (x) *x = p->s->img_x;
+   if (y) *y = p->s->img_y;
+   if (comp) *comp = p->s->img_n;
+   return 1;
+}
+
+static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   stbi__png p;
+   p.s = s;
+   return stbi__png_info_raw(&p, x, y, comp);
+}
+
+static int stbi__png_is16(stbi__context *s)
+{
+   stbi__png p;
+   p.s = s;
+   if (!stbi__png_info_raw(&p, NULL, NULL, NULL))
+	   return 0;
+   if (p.depth != 16) {
+      stbi__rewind(p.s);
+      return 0;
+   }
+   return 1;
+}
+#endif
+
+// Microsoft/Windows BMP image
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_test_raw(stbi__context *s)
+{
+   int r;
+   int sz;
+   if (stbi__get8(s) != 'B') return 0;
+   if (stbi__get8(s) != 'M') return 0;
+   stbi__get32le(s); // discard filesize
+   stbi__get16le(s); // discard reserved
+   stbi__get16le(s); // discard reserved
+   stbi__get32le(s); // discard data offset
+   sz = stbi__get32le(s);
+   r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124);
+   return r;
+}
+
+static int stbi__bmp_test(stbi__context *s)
+{
+   int r = stbi__bmp_test_raw(s);
+   stbi__rewind(s);
+   return r;
+}
+
+
+// returns 0..31 for the highest set bit
+static int stbi__high_bit(unsigned int z)
+{
+   int n=0;
+   if (z == 0) return -1;
+   if (z >= 0x10000) { n += 16; z >>= 16; }
+   if (z >= 0x00100) { n +=  8; z >>=  8; }
+   if (z >= 0x00010) { n +=  4; z >>=  4; }
+   if (z >= 0x00004) { n +=  2; z >>=  2; }
+   if (z >= 0x00002) { n +=  1; z >>=  1; }
+   return n;
+}
+
+static int stbi__bitcount(unsigned int a)
+{
+   a = (a & 0x55555555) + ((a >>  1) & 0x55555555); // max 2
+   a = (a & 0x33333333) + ((a >>  2) & 0x33333333); // max 4
+   a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits
+   a = (a + (a >> 8)); // max 16 per 8 bits
+   a = (a + (a >> 16)); // max 32 per 8 bits
+   return a & 0xff;
+}
+
+// extract an arbitrarily-aligned N-bit value (N=bits)
+// from v, and then make it 8-bits long and fractionally
+// extend it to full full range.
+static int stbi__shiftsigned(unsigned int v, int shift, int bits)
+{
+   static unsigned int mul_table[9] = {
+      0,
+      0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/,
+      0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/,
+   };
+   static unsigned int shift_table[9] = {
+      0, 0,0,1,0,2,4,6,0,
+   };
+   if (shift < 0)
+      v <<= -shift;
+   else
+      v >>= shift;
+   STBI_ASSERT(v >= 0 && v < 256);
+   v >>= (8-bits);
+   STBI_ASSERT(bits >= 0 && bits <= 8);
+   return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits];
+}
+
+typedef struct
+{
+   int bpp, offset, hsz;
+   unsigned int mr,mg,mb,ma, all_a;
+} stbi__bmp_data;
+
+static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info)
+{
+   int hsz;
+   if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP");
+   stbi__get32le(s); // discard filesize
+   stbi__get16le(s); // discard reserved
+   stbi__get16le(s); // discard reserved
+   info->offset = stbi__get32le(s);
+   info->hsz = hsz = stbi__get32le(s);
+   info->mr = info->mg = info->mb = info->ma = 0;
+
+   if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown");
+   if (hsz == 12) {
+      s->img_x = stbi__get16le(s);
+      s->img_y = stbi__get16le(s);
+   } else {
+      s->img_x = stbi__get32le(s);
+      s->img_y = stbi__get32le(s);
+   }
+   if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP");
+   info->bpp = stbi__get16le(s);
+   if (hsz != 12) {
+      int compress = stbi__get32le(s);
+      if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE");
+      stbi__get32le(s); // discard sizeof
+      stbi__get32le(s); // discard hres
+      stbi__get32le(s); // discard vres
+      stbi__get32le(s); // discard colorsused
+      stbi__get32le(s); // discard max important
+      if (hsz == 40 || hsz == 56) {
+         if (hsz == 56) {
+            stbi__get32le(s);
+            stbi__get32le(s);
+            stbi__get32le(s);
+            stbi__get32le(s);
+         }
+         if (info->bpp == 16 || info->bpp == 32) {
+            if (compress == 0) {
+               if (info->bpp == 32) {
+                  info->mr = 0xffu << 16;
+                  info->mg = 0xffu <<  8;
+                  info->mb = 0xffu <<  0;
+                  info->ma = 0xffu << 24;
+                  info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0
+               } else {
+                  info->mr = 31u << 10;
+                  info->mg = 31u <<  5;
+                  info->mb = 31u <<  0;
+               }
+            } else if (compress == 3) {
+               info->mr = stbi__get32le(s);
+               info->mg = stbi__get32le(s);
+               info->mb = stbi__get32le(s);
+               // not documented, but generated by photoshop and handled by mspaint
+               if (info->mr == info->mg && info->mg == info->mb) {
+                  // ?!?!?
+                  return stbi__errpuc("bad BMP", "bad BMP");
+               }
+            } else
+               return stbi__errpuc("bad BMP", "bad BMP");
+         }
+      } else {
+         int i;
+         if (hsz != 108 && hsz != 124)
+            return stbi__errpuc("bad BMP", "bad BMP");
+         info->mr = stbi__get32le(s);
+         info->mg = stbi__get32le(s);
+         info->mb = stbi__get32le(s);
+         info->ma = stbi__get32le(s);
+         stbi__get32le(s); // discard color space
+         for (i=0; i < 12; ++i)
+            stbi__get32le(s); // discard color space parameters
+         if (hsz == 124) {
+            stbi__get32le(s); // discard rendering intent
+            stbi__get32le(s); // discard offset of profile data
+            stbi__get32le(s); // discard size of profile data
+            stbi__get32le(s); // discard reserved
+         }
+      }
+   }
+   return (void *) 1;
+}
+
+
+static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *out;
+   unsigned int mr=0,mg=0,mb=0,ma=0, all_a;
+   stbi_uc pal[256][4];
+   int psize=0,i,j,width;
+   int flip_vertically, pad, target;
+   stbi__bmp_data info;
+   STBI_NOTUSED(ri);
+
+   info.all_a = 255;
+   if (stbi__bmp_parse_header(s, &info) == NULL)
+      return NULL; // error code already set
+
+   flip_vertically = ((int) s->img_y) > 0;
+   s->img_y = abs((int) s->img_y);
+
+   mr = info.mr;
+   mg = info.mg;
+   mb = info.mb;
+   ma = info.ma;
+   all_a = info.all_a;
+
+   if (info.hsz == 12) {
+      if (info.bpp < 24)
+         psize = (info.offset - 14 - 24) / 3;
+   } else {
+      if (info.bpp < 16)
+         psize = (info.offset - 14 - info.hsz) >> 2;
+   }
+
+   s->img_n = ma ? 4 : 3;
+   if (req_comp && req_comp >= 3) // we can directly decode 3 or 4
+      target = req_comp;
+   else
+      target = s->img_n; // if they want monochrome, we'll post-convert
+
+   // sanity-check size
+   if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0))
+      return stbi__errpuc("too large", "Corrupt BMP");
+
+   out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0);
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   if (info.bpp < 16) {
+      int z=0;
+      if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); }
+      for (i=0; i < psize; ++i) {
+         pal[i][2] = stbi__get8(s);
+         pal[i][1] = stbi__get8(s);
+         pal[i][0] = stbi__get8(s);
+         if (info.hsz != 12) stbi__get8(s);
+         pal[i][3] = 255;
+      }
+      stbi__skip(s, info.offset - 14 - info.hsz - psize * (info.hsz == 12 ? 3 : 4));
+      if (info.bpp == 1) width = (s->img_x + 7) >> 3;
+      else if (info.bpp == 4) width = (s->img_x + 1) >> 1;
+      else if (info.bpp == 8) width = s->img_x;
+      else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); }
+      pad = (-width)&3;
+      if (info.bpp == 1) {
+         for (j=0; j < (int) s->img_y; ++j) {
+            int bit_offset = 7, v = stbi__get8(s);
+            for (i=0; i < (int) s->img_x; ++i) {
+               int color = (v>>bit_offset)&0x1;
+               out[z++] = pal[color][0];
+               out[z++] = pal[color][1];
+               out[z++] = pal[color][2];
+               if (target == 4) out[z++] = 255;
+               if (i+1 == (int) s->img_x) break;
+               if((--bit_offset) < 0) {
+                  bit_offset = 7;
+                  v = stbi__get8(s);
+               }
+            }
+            stbi__skip(s, pad);
+         }
+      } else {
+         for (j=0; j < (int) s->img_y; ++j) {
+            for (i=0; i < (int) s->img_x; i += 2) {
+               int v=stbi__get8(s),v2=0;
+               if (info.bpp == 4) {
+                  v2 = v & 15;
+                  v >>= 4;
+               }
+               out[z++] = pal[v][0];
+               out[z++] = pal[v][1];
+               out[z++] = pal[v][2];
+               if (target == 4) out[z++] = 255;
+               if (i+1 == (int) s->img_x) break;
+               v = (info.bpp == 8) ? stbi__get8(s) : v2;
+               out[z++] = pal[v][0];
+               out[z++] = pal[v][1];
+               out[z++] = pal[v][2];
+               if (target == 4) out[z++] = 255;
+            }
+            stbi__skip(s, pad);
+         }
+      }
+   } else {
+      int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0;
+      int z = 0;
+      int easy=0;
+      stbi__skip(s, info.offset - 14 - info.hsz);
+      if (info.bpp == 24) width = 3 * s->img_x;
+      else if (info.bpp == 16) width = 2*s->img_x;
+      else /* bpp = 32 and pad = 0 */ width=0;
+      pad = (-width) & 3;
+      if (info.bpp == 24) {
+         easy = 1;
+      } else if (info.bpp == 32) {
+         if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000)
+            easy = 2;
+      }
+      if (!easy) {
+         if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); }
+         // right shift amt to put high bit in position #7
+         rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr);
+         gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg);
+         bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb);
+         ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma);
+      }
+      for (j=0; j < (int) s->img_y; ++j) {
+         if (easy) {
+            for (i=0; i < (int) s->img_x; ++i) {
+               unsigned char a;
+               out[z+2] = stbi__get8(s);
+               out[z+1] = stbi__get8(s);
+               out[z+0] = stbi__get8(s);
+               z += 3;
+               a = (easy == 2 ? stbi__get8(s) : 255);
+               all_a |= a;
+               if (target == 4) out[z++] = a;
+            }
+         } else {
+            int bpp = info.bpp;
+            for (i=0; i < (int) s->img_x; ++i) {
+               stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s));
+               unsigned int a;
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount));
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount));
+               out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount));
+               a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255);
+               all_a |= a;
+               if (target == 4) out[z++] = STBI__BYTECAST(a);
+            }
+         }
+         stbi__skip(s, pad);
+      }
+   }
+
+   // if alpha channel is all 0s, replace with all 255s
+   if (target == 4 && all_a == 0)
+      for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4)
+         out[i] = 255;
+
+   if (flip_vertically) {
+      stbi_uc t;
+      for (j=0; j < (int) s->img_y>>1; ++j) {
+         stbi_uc *p1 = out +      j     *s->img_x*target;
+         stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target;
+         for (i=0; i < (int) s->img_x*target; ++i) {
+            t = p1[i]; p1[i] = p2[i]; p2[i] = t;
+         }
+      }
+   }
+
+   if (req_comp && req_comp != target) {
+      out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y);
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+
+   *x = s->img_x;
+   *y = s->img_y;
+   if (comp) *comp = s->img_n;
+   return out;
+}
+#endif
+
+// Targa Truevision - TGA
+// by Jonathan Dummer
+#ifndef STBI_NO_TGA
+// returns STBI_rgb or whatever, 0 on error
+static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16)
+{
+   // only RGB or RGBA (incl. 16bit) or grey allowed
+   if (is_rgb16) *is_rgb16 = 0;
+   switch(bits_per_pixel) {
+      case 8:  return STBI_grey;
+      case 16: if(is_grey) return STBI_grey_alpha;
+               // fallthrough
+      case 15: if(is_rgb16) *is_rgb16 = 1;
+               return STBI_rgb;
+      case 24: // fallthrough
+      case 32: return bits_per_pixel/8;
+      default: return 0;
+   }
+}
+
+static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp)
+{
+    int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp;
+    int sz, tga_colormap_type;
+    stbi__get8(s);                   // discard Offset
+    tga_colormap_type = stbi__get8(s); // colormap type
+    if( tga_colormap_type > 1 ) {
+        stbi__rewind(s);
+        return 0;      // only RGB or indexed allowed
+    }
+    tga_image_type = stbi__get8(s); // image type
+    if ( tga_colormap_type == 1 ) { // colormapped (paletted) image
+        if (tga_image_type != 1 && tga_image_type != 9) {
+            stbi__rewind(s);
+            return 0;
+        }
+        stbi__skip(s,4);       // skip index of first colormap entry and number of entries
+        sz = stbi__get8(s);    //   check bits per palette color entry
+        if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) {
+            stbi__rewind(s);
+            return 0;
+        }
+        stbi__skip(s,4);       // skip image x and y origin
+        tga_colormap_bpp = sz;
+    } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE
+        if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) {
+            stbi__rewind(s);
+            return 0; // only RGB or grey allowed, +/- RLE
+        }
+        stbi__skip(s,9); // skip colormap specification and image x/y origin
+        tga_colormap_bpp = 0;
+    }
+    tga_w = stbi__get16le(s);
+    if( tga_w < 1 ) {
+        stbi__rewind(s);
+        return 0;   // test width
+    }
+    tga_h = stbi__get16le(s);
+    if( tga_h < 1 ) {
+        stbi__rewind(s);
+        return 0;   // test height
+    }
+    tga_bits_per_pixel = stbi__get8(s); // bits per pixel
+    stbi__get8(s); // ignore alpha bits
+    if (tga_colormap_bpp != 0) {
+        if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) {
+            // when using a colormap, tga_bits_per_pixel is the size of the indexes
+            // I don't think anything but 8 or 16bit indexes makes sense
+            stbi__rewind(s);
+            return 0;
+        }
+        tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL);
+    } else {
+        tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL);
+    }
+    if(!tga_comp) {
+      stbi__rewind(s);
+      return 0;
+    }
+    if (x) *x = tga_w;
+    if (y) *y = tga_h;
+    if (comp) *comp = tga_comp;
+    return 1;                   // seems to have passed everything
+}
+
+static int stbi__tga_test(stbi__context *s)
+{
+   int res = 0;
+   int sz, tga_color_type;
+   stbi__get8(s);      //   discard Offset
+   tga_color_type = stbi__get8(s);   //   color type
+   if ( tga_color_type > 1 ) goto errorEnd;   //   only RGB or indexed allowed
+   sz = stbi__get8(s);   //   image type
+   if ( tga_color_type == 1 ) { // colormapped (paletted) image
+      if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9
+      stbi__skip(s,4);       // skip index of first colormap entry and number of entries
+      sz = stbi__get8(s);    //   check bits per palette color entry
+      if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+      stbi__skip(s,4);       // skip image x and y origin
+   } else { // "normal" image w/o colormap
+      if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE
+      stbi__skip(s,9); // skip colormap specification and image x/y origin
+   }
+   if ( stbi__get16le(s) < 1 ) goto errorEnd;      //   test width
+   if ( stbi__get16le(s) < 1 ) goto errorEnd;      //   test height
+   sz = stbi__get8(s);   //   bits per pixel
+   if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index
+   if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd;
+
+   res = 1; // if we got this far, everything's good and we can return 1 instead of 0
+
+errorEnd:
+   stbi__rewind(s);
+   return res;
+}
+
+// read 16bit value and convert to 24bit RGB
+static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out)
+{
+   stbi__uint16 px = (stbi__uint16)stbi__get16le(s);
+   stbi__uint16 fiveBitMask = 31;
+   // we have 3 channels with 5bits each
+   int r = (px >> 10) & fiveBitMask;
+   int g = (px >> 5) & fiveBitMask;
+   int b = px & fiveBitMask;
+   // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later
+   out[0] = (stbi_uc)((r * 255)/31);
+   out[1] = (stbi_uc)((g * 255)/31);
+   out[2] = (stbi_uc)((b * 255)/31);
+
+   // some people claim that the most significant bit might be used for alpha
+   // (possibly if an alpha-bit is set in the "image descriptor byte")
+   // but that only made 16bit test images completely translucent..
+   // so let's treat all 15 and 16bit TGAs as RGB with no alpha.
+}
+
+static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   //   read in the TGA header stuff
+   int tga_offset = stbi__get8(s);
+   int tga_indexed = stbi__get8(s);
+   int tga_image_type = stbi__get8(s);
+   int tga_is_RLE = 0;
+   int tga_palette_start = stbi__get16le(s);
+   int tga_palette_len = stbi__get16le(s);
+   int tga_palette_bits = stbi__get8(s);
+   int tga_x_origin = stbi__get16le(s);
+   int tga_y_origin = stbi__get16le(s);
+   int tga_width = stbi__get16le(s);
+   int tga_height = stbi__get16le(s);
+   int tga_bits_per_pixel = stbi__get8(s);
+   int tga_comp, tga_rgb16=0;
+   int tga_inverted = stbi__get8(s);
+   // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?)
+   //   image data
+   unsigned char *tga_data;
+   unsigned char *tga_palette = NULL;
+   int i, j;
+   unsigned char raw_data[4] = {0};
+   int RLE_count = 0;
+   int RLE_repeating = 0;
+   int read_next_pixel = 1;
+   STBI_NOTUSED(ri);
+
+   //   do a tiny bit of precessing
+   if ( tga_image_type >= 8 )
+   {
+      tga_image_type -= 8;
+      tga_is_RLE = 1;
+   }
+   tga_inverted = 1 - ((tga_inverted >> 5) & 1);
+
+   //   If I'm paletted, then I'll use the number of bits from the palette
+   if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16);
+   else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16);
+
+   if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency
+      return stbi__errpuc("bad format", "Can't find out TGA pixelformat");
+
+   //   tga info
+   *x = tga_width;
+   *y = tga_height;
+   if (comp) *comp = tga_comp;
+
+   if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0))
+      return stbi__errpuc("too large", "Corrupt TGA");
+
+   tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0);
+   if (!tga_data) return stbi__errpuc("outofmem", "Out of memory");
+
+   // skip to the data's starting position (offset usually = 0)
+   stbi__skip(s, tga_offset );
+
+   if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) {
+      for (i=0; i < tga_height; ++i) {
+         int row = tga_inverted ? tga_height -i - 1 : i;
+         stbi_uc *tga_row = tga_data + row*tga_width*tga_comp;
+         stbi__getn(s, tga_row, tga_width * tga_comp);
+      }
+   } else  {
+      //   do I need to load a palette?
+      if ( tga_indexed)
+      {
+         //   any data to skip? (offset usually = 0)
+         stbi__skip(s, tga_palette_start );
+         //   load the palette
+         tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0);
+         if (!tga_palette) {
+            STBI_FREE(tga_data);
+            return stbi__errpuc("outofmem", "Out of memory");
+         }
+         if (tga_rgb16) {
+            stbi_uc *pal_entry = tga_palette;
+            STBI_ASSERT(tga_comp == STBI_rgb);
+            for (i=0; i < tga_palette_len; ++i) {
+               stbi__tga_read_rgb16(s, pal_entry);
+               pal_entry += tga_comp;
+            }
+         } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) {
+               STBI_FREE(tga_data);
+               STBI_FREE(tga_palette);
+               return stbi__errpuc("bad palette", "Corrupt TGA");
+         }
+      }
+      //   load the data
+      for (i=0; i < tga_width * tga_height; ++i)
+      {
+         //   if I'm in RLE mode, do I need to get a RLE stbi__pngchunk?
+         if ( tga_is_RLE )
+         {
+            if ( RLE_count == 0 )
+            {
+               //   yep, get the next byte as a RLE command
+               int RLE_cmd = stbi__get8(s);
+               RLE_count = 1 + (RLE_cmd & 127);
+               RLE_repeating = RLE_cmd >> 7;
+               read_next_pixel = 1;
+            } else if ( !RLE_repeating )
+            {
+               read_next_pixel = 1;
+            }
+         } else
+         {
+            read_next_pixel = 1;
+         }
+         //   OK, if I need to read a pixel, do it now
+         if ( read_next_pixel )
+         {
+            //   load however much data we did have
+            if ( tga_indexed )
+            {
+               // read in index, then perform the lookup
+               int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s);
+               if ( pal_idx >= tga_palette_len ) {
+                  // invalid index
+                  pal_idx = 0;
+               }
+               pal_idx *= tga_comp;
+               for (j = 0; j < tga_comp; ++j) {
+                  raw_data[j] = tga_palette[pal_idx+j];
+               }
+            } else if(tga_rgb16) {
+               STBI_ASSERT(tga_comp == STBI_rgb);
+               stbi__tga_read_rgb16(s, raw_data);
+            } else {
+               //   read in the data raw
+               for (j = 0; j < tga_comp; ++j) {
+                  raw_data[j] = stbi__get8(s);
+               }
+            }
+            //   clear the reading flag for the next pixel
+            read_next_pixel = 0;
+         } // end of reading a pixel
+
+         // copy data
+         for (j = 0; j < tga_comp; ++j)
+           tga_data[i*tga_comp+j] = raw_data[j];
+
+         //   in case we're in RLE mode, keep counting down
+         --RLE_count;
+      }
+      //   do I need to invert the image?
+      if ( tga_inverted )
+      {
+         for (j = 0; j*2 < tga_height; ++j)
+         {
+            int index1 = j * tga_width * tga_comp;
+            int index2 = (tga_height - 1 - j) * tga_width * tga_comp;
+            for (i = tga_width * tga_comp; i > 0; --i)
+            {
+               unsigned char temp = tga_data[index1];
+               tga_data[index1] = tga_data[index2];
+               tga_data[index2] = temp;
+               ++index1;
+               ++index2;
+            }
+         }
+      }
+      //   clear my palette, if I had one
+      if ( tga_palette != NULL )
+      {
+         STBI_FREE( tga_palette );
+      }
+   }
+
+   // swap RGB - if the source data was RGB16, it already is in the right order
+   if (tga_comp >= 3 && !tga_rgb16)
+   {
+      unsigned char* tga_pixel = tga_data;
+      for (i=0; i < tga_width * tga_height; ++i)
+      {
+         unsigned char temp = tga_pixel[0];
+         tga_pixel[0] = tga_pixel[2];
+         tga_pixel[2] = temp;
+         tga_pixel += tga_comp;
+      }
+   }
+
+   // convert to target component count
+   if (req_comp && req_comp != tga_comp)
+      tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height);
+
+   //   the things I do to get rid of an error message, and yet keep
+   //   Microsoft's C compilers happy... [8^(
+   tga_palette_start = tga_palette_len = tga_palette_bits =
+         tga_x_origin = tga_y_origin = 0;
+   //   OK, done
+   return tga_data;
+}
+#endif
+
+// *************************************************************************************************
+// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_test(stbi__context *s)
+{
+   int r = (stbi__get32be(s) == 0x38425053);
+   stbi__rewind(s);
+   return r;
+}
+
+static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount)
+{
+   int count, nleft, len;
+
+   count = 0;
+   while ((nleft = pixelCount - count) > 0) {
+      len = stbi__get8(s);
+      if (len == 128) {
+         // No-op.
+      } else if (len < 128) {
+         // Copy next len+1 bytes literally.
+         len++;
+         if (len > nleft) return 0; // corrupt data
+         count += len;
+         while (len) {
+            *p = stbi__get8(s);
+            p += 4;
+            len--;
+         }
+      } else if (len > 128) {
+         stbi_uc   val;
+         // Next -len+1 bytes in the dest are replicated from next source byte.
+         // (Interpret len as a negative 8-bit int.)
+         len = 257 - len;
+         if (len > nleft) return 0; // corrupt data
+         val = stbi__get8(s);
+         count += len;
+         while (len) {
+            *p = val;
+            p += 4;
+            len--;
+         }
+      }
+   }
+
+   return 1;
+}
+
+static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc)
+{
+   int pixelCount;
+   int channelCount, compression;
+   int channel, i;
+   int bitdepth;
+   int w,h;
+   stbi_uc *out;
+   STBI_NOTUSED(ri);
+
+   // Check identifier
+   if (stbi__get32be(s) != 0x38425053)   // "8BPS"
+      return stbi__errpuc("not PSD", "Corrupt PSD image");
+
+   // Check file type version.
+   if (stbi__get16be(s) != 1)
+      return stbi__errpuc("wrong version", "Unsupported version of PSD image");
+
+   // Skip 6 reserved bytes.
+   stbi__skip(s, 6 );
+
+   // Read the number of channels (R, G, B, A, etc).
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16)
+      return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image");
+
+   // Read the rows and columns of the image.
+   h = stbi__get32be(s);
+   w = stbi__get32be(s);
+
+   // Make sure the depth is 8 bits.
+   bitdepth = stbi__get16be(s);
+   if (bitdepth != 8 && bitdepth != 16)
+      return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit");
+
+   // Make sure the color mode is RGB.
+   // Valid options are:
+   //   0: Bitmap
+   //   1: Grayscale
+   //   2: Indexed color
+   //   3: RGB color
+   //   4: CMYK color
+   //   7: Multichannel
+   //   8: Duotone
+   //   9: Lab color
+   if (stbi__get16be(s) != 3)
+      return stbi__errpuc("wrong color format", "PSD is not in RGB color format");
+
+   // Skip the Mode Data.  (It's the palette for indexed color; other info for other modes.)
+   stbi__skip(s,stbi__get32be(s) );
+
+   // Skip the image resources.  (resolution, pen tool paths, etc)
+   stbi__skip(s, stbi__get32be(s) );
+
+   // Skip the reserved data.
+   stbi__skip(s, stbi__get32be(s) );
+
+   // Find out if the data is compressed.
+   // Known values:
+   //   0: no compression
+   //   1: RLE compressed
+   compression = stbi__get16be(s);
+   if (compression > 1)
+      return stbi__errpuc("bad compression", "PSD has an unknown compression format");
+
+   // Check size
+   if (!stbi__mad3sizes_valid(4, w, h, 0))
+      return stbi__errpuc("too large", "Corrupt PSD");
+
+   // Create the destination image.
+
+   if (!compression && bitdepth == 16 && bpc == 16) {
+      out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0);
+      ri->bits_per_channel = 16;
+   } else
+      out = (stbi_uc *) stbi__malloc(4 * w*h);
+
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   pixelCount = w*h;
+
+   // Initialize the data to zero.
+   //memset( out, 0, pixelCount * 4 );
+
+   // Finally, the image data.
+   if (compression) {
+      // RLE as used by .PSD and .TIFF
+      // Loop until you get the number of unpacked bytes you are expecting:
+      //     Read the next source byte into n.
+      //     If n is between 0 and 127 inclusive, copy the next n+1 bytes literally.
+      //     Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times.
+      //     Else if n is 128, noop.
+      // Endloop
+
+      // The RLE-compressed data is preceded by a 2-byte data count for each row in the data,
+      // which we're going to just skip.
+      stbi__skip(s, h * channelCount * 2 );
+
+      // Read the RLE data by channel.
+      for (channel = 0; channel < 4; channel++) {
+         stbi_uc *p;
+
+         p = out+channel;
+         if (channel >= channelCount) {
+            // Fill this channel with default data.
+            for (i = 0; i < pixelCount; i++, p += 4)
+               *p = (channel == 3 ? 255 : 0);
+         } else {
+            // Read the RLE data.
+            if (!stbi__psd_decode_rle(s, p, pixelCount)) {
+               STBI_FREE(out);
+               return stbi__errpuc("corrupt", "bad RLE data");
+            }
+         }
+      }
+
+   } else {
+      // We're at the raw image data.  It's each channel in order (Red, Green, Blue, Alpha, ...)
+      // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image.
+
+      // Read the data by channel.
+      for (channel = 0; channel < 4; channel++) {
+         if (channel >= channelCount) {
+            // Fill this channel with default data.
+            if (bitdepth == 16 && bpc == 16) {
+               stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+               stbi__uint16 val = channel == 3 ? 65535 : 0;
+               for (i = 0; i < pixelCount; i++, q += 4)
+                  *q = val;
+            } else {
+               stbi_uc *p = out+channel;
+               stbi_uc val = channel == 3 ? 255 : 0;
+               for (i = 0; i < pixelCount; i++, p += 4)
+                  *p = val;
+            }
+         } else {
+            if (ri->bits_per_channel == 16) {    // output bpc
+               stbi__uint16 *q = ((stbi__uint16 *) out) + channel;
+               for (i = 0; i < pixelCount; i++, q += 4)
+                  *q = (stbi__uint16) stbi__get16be(s);
+            } else {
+               stbi_uc *p = out+channel;
+               if (bitdepth == 16) {  // input bpc
+                  for (i = 0; i < pixelCount; i++, p += 4)
+                     *p = (stbi_uc) (stbi__get16be(s) >> 8);
+               } else {
+                  for (i = 0; i < pixelCount; i++, p += 4)
+                     *p = stbi__get8(s);
+               }
+            }
+         }
+      }
+   }
+
+   // remove weird white matte from PSD
+   if (channelCount >= 4) {
+      if (ri->bits_per_channel == 16) {
+         for (i=0; i < w*h; ++i) {
+            stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i;
+            if (pixel[3] != 0 && pixel[3] != 65535) {
+               float a = pixel[3] / 65535.0f;
+               float ra = 1.0f / a;
+               float inv_a = 65535.0f * (1 - ra);
+               pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a);
+               pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a);
+               pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a);
+            }
+         }
+      } else {
+         for (i=0; i < w*h; ++i) {
+            unsigned char *pixel = out + 4*i;
+            if (pixel[3] != 0 && pixel[3] != 255) {
+               float a = pixel[3] / 255.0f;
+               float ra = 1.0f / a;
+               float inv_a = 255.0f * (1 - ra);
+               pixel[0] = (unsigned char) (pixel[0]*ra + inv_a);
+               pixel[1] = (unsigned char) (pixel[1]*ra + inv_a);
+               pixel[2] = (unsigned char) (pixel[2]*ra + inv_a);
+            }
+         }
+      }
+   }
+
+   // convert to desired output format
+   if (req_comp && req_comp != 4) {
+      if (ri->bits_per_channel == 16)
+         out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h);
+      else
+         out = stbi__convert_format(out, 4, req_comp, w, h);
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+
+   if (comp) *comp = 4;
+   *y = h;
+   *x = w;
+
+   return out;
+}
+#endif
+
+// *************************************************************************************************
+// Softimage PIC loader
+// by Tom Seddon
+//
+// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format
+// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_is4(stbi__context *s,const char *str)
+{
+   int i;
+   for (i=0; i<4; ++i)
+      if (stbi__get8(s) != (stbi_uc)str[i])
+         return 0;
+
+   return 1;
+}
+
+static int stbi__pic_test_core(stbi__context *s)
+{
+   int i;
+
+   if (!stbi__pic_is4(s,"\x53\x80\xF6\x34"))
+      return 0;
+
+   for(i=0;i<84;++i)
+      stbi__get8(s);
+
+   if (!stbi__pic_is4(s,"PICT"))
+      return 0;
+
+   return 1;
+}
+
+typedef struct
+{
+   stbi_uc size,type,channel;
+} stbi__pic_packet;
+
+static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest)
+{
+   int mask=0x80, i;
+
+   for (i=0; i<4; ++i, mask>>=1) {
+      if (channel & mask) {
+         if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short");
+         dest[i]=stbi__get8(s);
+      }
+   }
+
+   return dest;
+}
+
+static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src)
+{
+   int mask=0x80,i;
+
+   for (i=0;i<4; ++i, mask>>=1)
+      if (channel&mask)
+         dest[i]=src[i];
+}
+
+static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result)
+{
+   int act_comp=0,num_packets=0,y,chained;
+   stbi__pic_packet packets[10];
+
+   // this will (should...) cater for even some bizarre stuff like having data
+    // for the same channel in multiple packets.
+   do {
+      stbi__pic_packet *packet;
+
+      if (num_packets==sizeof(packets)/sizeof(packets[0]))
+         return stbi__errpuc("bad format","too many packets");
+
+      packet = &packets[num_packets++];
+
+      chained = stbi__get8(s);
+      packet->size    = stbi__get8(s);
+      packet->type    = stbi__get8(s);
+      packet->channel = stbi__get8(s);
+
+      act_comp |= packet->channel;
+
+      if (stbi__at_eof(s))          return stbi__errpuc("bad file","file too short (reading packets)");
+      if (packet->size != 8)  return stbi__errpuc("bad format","packet isn't 8bpp");
+   } while (chained);
+
+   *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel?
+
+   for(y=0; y<height; ++y) {
+      int packet_idx;
+
+      for(packet_idx=0; packet_idx < num_packets; ++packet_idx) {
+         stbi__pic_packet *packet = &packets[packet_idx];
+         stbi_uc *dest = result+y*width*4;
+
+         switch (packet->type) {
+            default:
+               return stbi__errpuc("bad format","packet has bad compression type");
+
+            case 0: {//uncompressed
+               int x;
+
+               for(x=0;x<width;++x, dest+=4)
+                  if (!stbi__readval(s,packet->channel,dest))
+                     return 0;
+               break;
+            }
+
+            case 1://Pure RLE
+               {
+                  int left=width, i;
+
+                  while (left>0) {
+                     stbi_uc count,value[4];
+
+                     count=stbi__get8(s);
+                     if (stbi__at_eof(s))   return stbi__errpuc("bad file","file too short (pure read count)");
+
+                     if (count > left)
+                        count = (stbi_uc) left;
+
+                     if (!stbi__readval(s,packet->channel,value))  return 0;
+
+                     for(i=0; i<count; ++i,dest+=4)
+                        stbi__copyval(packet->channel,dest,value);
+                     left -= count;
+                  }
+               }
+               break;
+
+            case 2: {//Mixed RLE
+               int left=width;
+               while (left>0) {
+                  int count = stbi__get8(s), i;
+                  if (stbi__at_eof(s))  return stbi__errpuc("bad file","file too short (mixed read count)");
+
+                  if (count >= 128) { // Repeated
+                     stbi_uc value[4];
+
+                     if (count==128)
+                        count = stbi__get16be(s);
+                     else
+                        count -= 127;
+                     if (count > left)
+                        return stbi__errpuc("bad file","scanline overrun");
+
+                     if (!stbi__readval(s,packet->channel,value))
+                        return 0;
+
+                     for(i=0;i<count;++i, dest += 4)
+                        stbi__copyval(packet->channel,dest,value);
+                  } else { // Raw
+                     ++count;
+                     if (count>left) return stbi__errpuc("bad file","scanline overrun");
+
+                     for(i=0;i<count;++i, dest+=4)
+                        if (!stbi__readval(s,packet->channel,dest))
+                           return 0;
+                  }
+                  left-=count;
+               }
+               break;
+            }
+         }
+      }
+   }
+
+   return result;
+}
+
+static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *result;
+   int i, x,y, internal_comp;
+   STBI_NOTUSED(ri);
+
+   if (!comp) comp = &internal_comp;
+
+   for (i=0; i<92; ++i)
+      stbi__get8(s);
+
+   x = stbi__get16be(s);
+   y = stbi__get16be(s);
+   if (stbi__at_eof(s))  return stbi__errpuc("bad file","file too short (pic header)");
+   if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode");
+
+   stbi__get32be(s); //skip `ratio'
+   stbi__get16be(s); //skip `fields'
+   stbi__get16be(s); //skip `pad'
+
+   // intermediate buffer is RGBA
+   result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0);
+   memset(result, 0xff, x*y*4);
+
+   if (!stbi__pic_load_core(s,x,y,comp, result)) {
+      STBI_FREE(result);
+      result=0;
+   }
+   *px = x;
+   *py = y;
+   if (req_comp == 0) req_comp = *comp;
+   result=stbi__convert_format(result,4,req_comp,x,y);
+
+   return result;
+}
+
+static int stbi__pic_test(stbi__context *s)
+{
+   int r = stbi__pic_test_core(s);
+   stbi__rewind(s);
+   return r;
+}
+#endif
+
+// *************************************************************************************************
+// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb
+
+#ifndef STBI_NO_GIF
+typedef struct
+{
+   stbi__int16 prefix;
+   stbi_uc first;
+   stbi_uc suffix;
+} stbi__gif_lzw;
+
+typedef struct
+{
+   int w,h;
+   stbi_uc *out;                 // output buffer (always 4 components)
+   stbi_uc *background;          // The current "background" as far as a gif is concerned
+   stbi_uc *history; 
+   int flags, bgindex, ratio, transparent, eflags;
+   stbi_uc  pal[256][4];
+   stbi_uc lpal[256][4];
+   stbi__gif_lzw codes[8192];
+   stbi_uc *color_table;
+   int parse, step;
+   int lflags;
+   int start_x, start_y;
+   int max_x, max_y;
+   int cur_x, cur_y;
+   int line_size;
+   int delay;
+} stbi__gif;
+
+static int stbi__gif_test_raw(stbi__context *s)
+{
+   int sz;
+   if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0;
+   sz = stbi__get8(s);
+   if (sz != '9' && sz != '7') return 0;
+   if (stbi__get8(s) != 'a') return 0;
+   return 1;
+}
+
+static int stbi__gif_test(stbi__context *s)
+{
+   int r = stbi__gif_test_raw(s);
+   stbi__rewind(s);
+   return r;
+}
+
+static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp)
+{
+   int i;
+   for (i=0; i < num_entries; ++i) {
+      pal[i][2] = stbi__get8(s);
+      pal[i][1] = stbi__get8(s);
+      pal[i][0] = stbi__get8(s);
+      pal[i][3] = transp == i ? 0 : 255;
+   }
+}
+
+static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info)
+{
+   stbi_uc version;
+   if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8')
+      return stbi__err("not GIF", "Corrupt GIF");
+
+   version = stbi__get8(s);
+   if (version != '7' && version != '9')    return stbi__err("not GIF", "Corrupt GIF");
+   if (stbi__get8(s) != 'a')                return stbi__err("not GIF", "Corrupt GIF");
+
+   stbi__g_failure_reason = "";
+   g->w = stbi__get16le(s);
+   g->h = stbi__get16le(s);
+   g->flags = stbi__get8(s);
+   g->bgindex = stbi__get8(s);
+   g->ratio = stbi__get8(s);
+   g->transparent = -1;
+
+   if (comp != 0) *comp = 4;  // can't actually tell whether it's 3 or 4 until we parse the comments
+
+   if (is_info) return 1;
+
+   if (g->flags & 0x80)
+      stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1);
+
+   return 1;
+}
+
+static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp)
+{
+   stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif));
+   if (!stbi__gif_header(s, g, comp, 1)) {
+      STBI_FREE(g);
+      stbi__rewind( s );
+      return 0;
+   }
+   if (x) *x = g->w;
+   if (y) *y = g->h;
+   STBI_FREE(g);
+   return 1;
+}
+
+static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code)
+{
+   stbi_uc *p, *c;
+   int idx; 
+
+   // recurse to decode the prefixes, since the linked-list is backwards,
+   // and working backwards through an interleaved image would be nasty
+   if (g->codes[code].prefix >= 0)
+      stbi__out_gif_code(g, g->codes[code].prefix);
+
+   if (g->cur_y >= g->max_y) return;
+
+   idx = g->cur_x + g->cur_y; 
+   p = &g->out[idx];
+   g->history[idx / 4] = 1;  
+
+   c = &g->color_table[g->codes[code].suffix * 4];
+   if (c[3] > 128) { // don't render transparent pixels; 
+      p[0] = c[2];
+      p[1] = c[1];
+      p[2] = c[0];
+      p[3] = c[3];
+   }
+   g->cur_x += 4;
+
+   if (g->cur_x >= g->max_x) {
+      g->cur_x = g->start_x;
+      g->cur_y += g->step;
+
+      while (g->cur_y >= g->max_y && g->parse > 0) {
+         g->step = (1 << g->parse) * g->line_size;
+         g->cur_y = g->start_y + (g->step >> 1);
+         --g->parse;
+      }
+   }
+}
+
+static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g)
+{
+   stbi_uc lzw_cs;
+   stbi__int32 len, init_code;
+   stbi__uint32 first;
+   stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear;
+   stbi__gif_lzw *p;
+
+   lzw_cs = stbi__get8(s);
+   if (lzw_cs > 12) return NULL;
+   clear = 1 << lzw_cs;
+   first = 1;
+   codesize = lzw_cs + 1;
+   codemask = (1 << codesize) - 1;
+   bits = 0;
+   valid_bits = 0;
+   for (init_code = 0; init_code < clear; init_code++) {
+      g->codes[init_code].prefix = -1;
+      g->codes[init_code].first = (stbi_uc) init_code;
+      g->codes[init_code].suffix = (stbi_uc) init_code;
+   }
+
+   // support no starting clear code
+   avail = clear+2;
+   oldcode = -1;
+
+   len = 0;
+   for(;;) {
+      if (valid_bits < codesize) {
+         if (len == 0) {
+            len = stbi__get8(s); // start new block
+            if (len == 0)
+               return g->out;
+         }
+         --len;
+         bits |= (stbi__int32) stbi__get8(s) << valid_bits;
+         valid_bits += 8;
+      } else {
+         stbi__int32 code = bits & codemask;
+         bits >>= codesize;
+         valid_bits -= codesize;
+         // @OPTIMIZE: is there some way we can accelerate the non-clear path?
+         if (code == clear) {  // clear code
+            codesize = lzw_cs + 1;
+            codemask = (1 << codesize) - 1;
+            avail = clear + 2;
+            oldcode = -1;
+            first = 0;
+         } else if (code == clear + 1) { // end of stream code
+            stbi__skip(s, len);
+            while ((len = stbi__get8(s)) > 0)
+               stbi__skip(s,len);
+            return g->out;
+         } else if (code <= avail) {
+            if (first) {
+               return stbi__errpuc("no clear code", "Corrupt GIF");
+            }
+
+            if (oldcode >= 0) {
+               p = &g->codes[avail++];
+               if (avail > 8192) {
+                  return stbi__errpuc("too many codes", "Corrupt GIF");
+               }
+
+               p->prefix = (stbi__int16) oldcode;
+               p->first = g->codes[oldcode].first;
+               p->suffix = (code == avail) ? p->first : g->codes[code].first;
+            } else if (code == avail)
+               return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+
+            stbi__out_gif_code(g, (stbi__uint16) code);
+
+            if ((avail & codemask) == 0 && avail <= 0x0FFF) {
+               codesize++;
+               codemask = (1 << codesize) - 1;
+            }
+
+            oldcode = code;
+         } else {
+            return stbi__errpuc("illegal code in raster", "Corrupt GIF");
+         }
+      }
+   }
+}
+
+// this function is designed to support animated gifs, although stb_image doesn't support it
+// two back is the image from two frames ago, used for a very specific disposal format
+static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back)
+{
+   int dispose; 
+   int first_frame; 
+   int pi; 
+   int pcount; 
+   STBI_NOTUSED(req_comp);
+
+   // on first frame, any non-written pixels get the background colour (non-transparent)
+   first_frame = 0; 
+   if (g->out == 0) {
+      if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header
+      if (!stbi__mad3sizes_valid(4, g->w, g->h, 0))
+         return stbi__errpuc("too large", "GIF image is too large");
+      pcount = g->w * g->h;
+      g->out = (stbi_uc *) stbi__malloc(4 * pcount);
+      g->background = (stbi_uc *) stbi__malloc(4 * pcount);
+      g->history = (stbi_uc *) stbi__malloc(pcount);
+      if (!g->out || !g->background || !g->history)
+         return stbi__errpuc("outofmem", "Out of memory");
+
+      // image is treated as "transparent" at the start - ie, nothing overwrites the current background; 
+      // background colour is only used for pixels that are not rendered first frame, after that "background"
+      // color refers to the color that was there the previous frame. 
+      memset(g->out, 0x00, 4 * pcount);
+      memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent)
+      memset(g->history, 0x00, pcount);        // pixels that were affected previous frame
+      first_frame = 1; 
+   } else {
+      // second frame - how do we dispoase of the previous one?
+      dispose = (g->eflags & 0x1C) >> 2; 
+      pcount = g->w * g->h; 
+
+      if ((dispose == 3) && (two_back == 0)) {
+         dispose = 2; // if I don't have an image to revert back to, default to the old background
+      }
+
+      if (dispose == 3) { // use previous graphic
+         for (pi = 0; pi < pcount; ++pi) {
+            if (g->history[pi]) {
+               memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); 
+            }
+         }
+      } else if (dispose == 2) { 
+         // restore what was changed last frame to background before that frame; 
+         for (pi = 0; pi < pcount; ++pi) {
+            if (g->history[pi]) {
+               memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); 
+            }
+         }
+      } else {
+         // This is a non-disposal case eithe way, so just 
+         // leave the pixels as is, and they will become the new background
+         // 1: do not dispose
+         // 0:  not specified.
+      }
+
+      // background is what out is after the undoing of the previou frame; 
+      memcpy( g->background, g->out, 4 * g->w * g->h ); 
+   }
+
+   // clear my history; 
+   memset( g->history, 0x00, g->w * g->h );        // pixels that were affected previous frame
+
+   for (;;) {
+      int tag = stbi__get8(s); 
+      switch (tag) {
+         case 0x2C: /* Image Descriptor */
+         {
+            stbi__int32 x, y, w, h;
+            stbi_uc *o;
+
+            x = stbi__get16le(s);
+            y = stbi__get16le(s);
+            w = stbi__get16le(s);
+            h = stbi__get16le(s);
+            if (((x + w) > (g->w)) || ((y + h) > (g->h)))
+               return stbi__errpuc("bad Image Descriptor", "Corrupt GIF");
+
+            g->line_size = g->w * 4;
+            g->start_x = x * 4;
+            g->start_y = y * g->line_size;
+            g->max_x   = g->start_x + w * 4;
+            g->max_y   = g->start_y + h * g->line_size;
+            g->cur_x   = g->start_x;
+            g->cur_y   = g->start_y;
+
+            // if the width of the specified rectangle is 0, that means
+            // we may not see *any* pixels or the image is malformed;
+            // to make sure this is caught, move the current y down to
+            // max_y (which is what out_gif_code checks).
+            if (w == 0)
+               g->cur_y = g->max_y;
+
+            g->lflags = stbi__get8(s);
+
+            if (g->lflags & 0x40) {
+               g->step = 8 * g->line_size; // first interlaced spacing
+               g->parse = 3;
+            } else {
+               g->step = g->line_size;
+               g->parse = 0;
+            }
+
+            if (g->lflags & 0x80) {
+               stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1);
+               g->color_table = (stbi_uc *) g->lpal;
+            } else if (g->flags & 0x80) {
+               g->color_table = (stbi_uc *) g->pal;
+            } else
+               return stbi__errpuc("missing color table", "Corrupt GIF");            
+            
+            o = stbi__process_gif_raster(s, g);
+            if (!o) return NULL;
+
+            // if this was the first frame, 
+            pcount = g->w * g->h; 
+            if (first_frame && (g->bgindex > 0)) {
+               // if first frame, any pixel not drawn to gets the background color
+               for (pi = 0; pi < pcount; ++pi) {
+                  if (g->history[pi] == 0) {
+                     g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; 
+                     memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); 
+                  }
+               }
+            }
+
+            return o;
+         }
+
+         case 0x21: // Comment Extension.
+         {
+            int len;
+            int ext = stbi__get8(s); 
+            if (ext == 0xF9) { // Graphic Control Extension.
+               len = stbi__get8(s);
+               if (len == 4) {
+                  g->eflags = stbi__get8(s);
+                  g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths.
+
+                  // unset old transparent
+                  if (g->transparent >= 0) {
+                     g->pal[g->transparent][3] = 255; 
+                  } 
+                  if (g->eflags & 0x01) {
+                     g->transparent = stbi__get8(s);
+                     if (g->transparent >= 0) {
+                        g->pal[g->transparent][3] = 0; 
+                     }
+                  } else {
+                     // don't need transparent
+                     stbi__skip(s, 1); 
+                     g->transparent = -1; 
+                  }
+               } else {
+                  stbi__skip(s, len);
+                  break;
+               }
+            } 
+            while ((len = stbi__get8(s)) != 0) {
+               stbi__skip(s, len);
+            }
+            break;
+         }
+
+         case 0x3B: // gif stream termination code
+            return (stbi_uc *) s; // using '1' causes warning on some compilers
+
+         default:
+            return stbi__errpuc("unknown code", "Corrupt GIF");
+      }
+   }
+}
+
+static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp)
+{
+   if (stbi__gif_test(s)) {
+      int layers = 0; 
+      stbi_uc *u = 0;
+      stbi_uc *out = 0;
+      stbi_uc *two_back = 0; 
+      stbi__gif g;
+      int stride; 
+      memset(&g, 0, sizeof(g));
+      if (delays) {
+         *delays = 0; 
+      }
+
+      do {
+         u = stbi__gif_load_next(s, &g, comp, req_comp, two_back);
+         if (u == (stbi_uc *) s) u = 0;  // end of animated gif marker
+
+         if (u) {
+            *x = g.w;
+            *y = g.h;
+            ++layers; 
+            stride = g.w * g.h * 4; 
+         
+            if (out) {
+               out = (stbi_uc*) STBI_REALLOC( out, layers * stride ); 
+               if (delays) {
+                  *delays = (int*) STBI_REALLOC( *delays, sizeof(int) * layers ); 
+               }
+            } else {
+               out = (stbi_uc*)stbi__malloc( layers * stride ); 
+               if (delays) {
+                  *delays = (int*) stbi__malloc( layers * sizeof(int) ); 
+               }
+            }
+            memcpy( out + ((layers - 1) * stride), u, stride ); 
+            if (layers >= 2) {
+               two_back = out - 2 * stride; 
+            }
+
+            if (delays) {
+               (*delays)[layers - 1U] = g.delay; 
+            }
+         }
+      } while (u != 0); 
+
+      // free temp buffer; 
+      STBI_FREE(g.out); 
+      STBI_FREE(g.history); 
+      STBI_FREE(g.background); 
+
+      // do the final conversion after loading everything; 
+      if (req_comp && req_comp != 4)
+         out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h);
+
+      *z = layers; 
+      return out;
+   } else {
+      return stbi__errpuc("not GIF", "Image was not as a gif type."); 
+   }
+}
+
+static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *u = 0;
+   stbi__gif g;
+   memset(&g, 0, sizeof(g));
+   STBI_NOTUSED(ri);
+
+   u = stbi__gif_load_next(s, &g, comp, req_comp, 0);
+   if (u == (stbi_uc *) s) u = 0;  // end of animated gif marker
+   if (u) {
+      *x = g.w;
+      *y = g.h;
+
+      // moved conversion to after successful load so that the same
+      // can be done for multiple frames. 
+      if (req_comp && req_comp != 4)
+         u = stbi__convert_format(u, 4, req_comp, g.w, g.h);
+   } else if (g.out) {
+      // if there was an error and we allocated an image buffer, free it!
+      STBI_FREE(g.out);
+   }
+
+   // free buffers needed for multiple frame loading; 
+   STBI_FREE(g.history);
+   STBI_FREE(g.background); 
+
+   return u;
+}
+
+static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   return stbi__gif_info_raw(s,x,y,comp);
+}
+#endif
+
+// *************************************************************************************************
+// Radiance RGBE HDR loader
+// originally by Nicolas Schulz
+#ifndef STBI_NO_HDR
+static int stbi__hdr_test_core(stbi__context *s, const char *signature)
+{
+   int i;
+   for (i=0; signature[i]; ++i)
+      if (stbi__get8(s) != signature[i])
+          return 0;
+   stbi__rewind(s);
+   return 1;
+}
+
+static int stbi__hdr_test(stbi__context* s)
+{
+   int r = stbi__hdr_test_core(s, "#?RADIANCE\n");
+   stbi__rewind(s);
+   if(!r) {
+       r = stbi__hdr_test_core(s, "#?RGBE\n");
+       stbi__rewind(s);
+   }
+   return r;
+}
+
+#define STBI__HDR_BUFLEN  1024
+static char *stbi__hdr_gettoken(stbi__context *z, char *buffer)
+{
+   int len=0;
+   char c = '\0';
+
+   c = (char) stbi__get8(z);
+
+   while (!stbi__at_eof(z) && c != '\n') {
+      buffer[len++] = c;
+      if (len == STBI__HDR_BUFLEN-1) {
+         // flush to end of line
+         while (!stbi__at_eof(z) && stbi__get8(z) != '\n')
+            ;
+         break;
+      }
+      c = (char) stbi__get8(z);
+   }
+
+   buffer[len] = 0;
+   return buffer;
+}
+
+static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp)
+{
+   if ( input[3] != 0 ) {
+      float f1;
+      // Exponent
+      f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8));
+      if (req_comp <= 2)
+         output[0] = (input[0] + input[1] + input[2]) * f1 / 3;
+      else {
+         output[0] = input[0] * f1;
+         output[1] = input[1] * f1;
+         output[2] = input[2] * f1;
+      }
+      if (req_comp == 2) output[1] = 1;
+      if (req_comp == 4) output[3] = 1;
+   } else {
+      switch (req_comp) {
+         case 4: output[3] = 1; /* fallthrough */
+         case 3: output[0] = output[1] = output[2] = 0;
+                 break;
+         case 2: output[1] = 1; /* fallthrough */
+         case 1: output[0] = 0;
+                 break;
+      }
+   }
+}
+
+static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   char buffer[STBI__HDR_BUFLEN];
+   char *token;
+   int valid = 0;
+   int width, height;
+   stbi_uc *scanline;
+   float *hdr_data;
+   int len;
+   unsigned char count, value;
+   int i, j, k, c1,c2, z;
+   const char *headerToken;
+   STBI_NOTUSED(ri);
+
+   // Check identifier
+   headerToken = stbi__hdr_gettoken(s,buffer);
+   if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0)
+      return stbi__errpf("not HDR", "Corrupt HDR image");
+
+   // Parse header
+   for(;;) {
+      token = stbi__hdr_gettoken(s,buffer);
+      if (token[0] == 0) break;
+      if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+   }
+
+   if (!valid)    return stbi__errpf("unsupported format", "Unsupported HDR format");
+
+   // Parse width and height
+   // can't use sscanf() if we're not using stdio!
+   token = stbi__hdr_gettoken(s,buffer);
+   if (strncmp(token, "-Y ", 3))  return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+   token += 3;
+   height = (int) strtol(token, &token, 10);
+   while (*token == ' ') ++token;
+   if (strncmp(token, "+X ", 3))  return stbi__errpf("unsupported data layout", "Unsupported HDR format");
+   token += 3;
+   width = (int) strtol(token, NULL, 10);
+
+   *x = width;
+   *y = height;
+
+   if (comp) *comp = 3;
+   if (req_comp == 0) req_comp = 3;
+
+   if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0))
+      return stbi__errpf("too large", "HDR image is too large");
+
+   // Read data
+   hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0);
+   if (!hdr_data)
+      return stbi__errpf("outofmem", "Out of memory");
+
+   // Load image data
+   // image data is stored as some number of sca
+   if ( width < 8 || width >= 32768) {
+      // Read flat data
+      for (j=0; j < height; ++j) {
+         for (i=0; i < width; ++i) {
+            stbi_uc rgbe[4];
+           main_decode_loop:
+            stbi__getn(s, rgbe, 4);
+            stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp);
+         }
+      }
+   } else {
+      // Read RLE-encoded data
+      scanline = NULL;
+
+      for (j = 0; j < height; ++j) {
+         c1 = stbi__get8(s);
+         c2 = stbi__get8(s);
+         len = stbi__get8(s);
+         if (c1 != 2 || c2 != 2 || (len & 0x80)) {
+            // not run-length encoded, so we have to actually use THIS data as a decoded
+            // pixel (note this can't be a valid pixel--one of RGB must be >= 128)
+            stbi_uc rgbe[4];
+            rgbe[0] = (stbi_uc) c1;
+            rgbe[1] = (stbi_uc) c2;
+            rgbe[2] = (stbi_uc) len;
+            rgbe[3] = (stbi_uc) stbi__get8(s);
+            stbi__hdr_convert(hdr_data, rgbe, req_comp);
+            i = 1;
+            j = 0;
+            STBI_FREE(scanline);
+            goto main_decode_loop; // yes, this makes no sense
+         }
+         len <<= 8;
+         len |= stbi__get8(s);
+         if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); }
+         if (scanline == NULL) {
+            scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0);
+            if (!scanline) {
+               STBI_FREE(hdr_data);
+               return stbi__errpf("outofmem", "Out of memory");
+            }
+         }
+
+         for (k = 0; k < 4; ++k) {
+            int nleft;
+            i = 0;
+            while ((nleft = width - i) > 0) {
+               count = stbi__get8(s);
+               if (count > 128) {
+                  // Run
+                  value = stbi__get8(s);
+                  count -= 128;
+                  if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+                  for (z = 0; z < count; ++z)
+                     scanline[i++ * 4 + k] = value;
+               } else {
+                  // Dump
+                  if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); }
+                  for (z = 0; z < count; ++z)
+                     scanline[i++ * 4 + k] = stbi__get8(s);
+               }
+            }
+         }
+         for (i=0; i < width; ++i)
+            stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp);
+      }
+      if (scanline)
+         STBI_FREE(scanline);
+   }
+
+   return hdr_data;
+}
+
+static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   char buffer[STBI__HDR_BUFLEN];
+   char *token;
+   int valid = 0;
+   int dummy;
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   if (stbi__hdr_test(s) == 0) {
+       stbi__rewind( s );
+       return 0;
+   }
+
+   for(;;) {
+      token = stbi__hdr_gettoken(s,buffer);
+      if (token[0] == 0) break;
+      if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1;
+   }
+
+   if (!valid) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token = stbi__hdr_gettoken(s,buffer);
+   if (strncmp(token, "-Y ", 3)) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token += 3;
+   *y = (int) strtol(token, &token, 10);
+   while (*token == ' ') ++token;
+   if (strncmp(token, "+X ", 3)) {
+       stbi__rewind( s );
+       return 0;
+   }
+   token += 3;
+   *x = (int) strtol(token, NULL, 10);
+   *comp = 3;
+   return 1;
+}
+#endif // STBI_NO_HDR
+
+#ifndef STBI_NO_BMP
+static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   void *p;
+   stbi__bmp_data info;
+
+   info.all_a = 255;
+   p = stbi__bmp_parse_header(s, &info);
+   stbi__rewind( s );
+   if (p == NULL)
+      return 0;
+   if (x) *x = s->img_x;
+   if (y) *y = s->img_y;
+   if (comp) *comp = info.ma ? 4 : 3;
+   return 1;
+}
+#endif
+
+#ifndef STBI_NO_PSD
+static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int channelCount, dummy, depth;
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+   if (stbi__get32be(s) != 0x38425053) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 1) {
+       stbi__rewind( s );
+       return 0;
+   }
+   stbi__skip(s, 6);
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   *y = stbi__get32be(s);
+   *x = stbi__get32be(s);
+   depth = stbi__get16be(s);
+   if (depth != 8 && depth != 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 3) {
+       stbi__rewind( s );
+       return 0;
+   }
+   *comp = 4;
+   return 1;
+}
+
+static int stbi__psd_is16(stbi__context *s)
+{
+   int channelCount, depth;
+   if (stbi__get32be(s) != 0x38425053) {
+       stbi__rewind( s );
+       return 0;
+   }
+   if (stbi__get16be(s) != 1) {
+       stbi__rewind( s );
+       return 0;
+   }
+   stbi__skip(s, 6);
+   channelCount = stbi__get16be(s);
+   if (channelCount < 0 || channelCount > 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   (void) stbi__get32be(s);
+   (void) stbi__get32be(s);
+   depth = stbi__get16be(s);
+   if (depth != 16) {
+       stbi__rewind( s );
+       return 0;
+   }
+   return 1;
+}
+#endif
+
+#ifndef STBI_NO_PIC
+static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int act_comp=0,num_packets=0,chained,dummy;
+   stbi__pic_packet packets[10];
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) {
+      stbi__rewind(s);
+      return 0;
+   }
+
+   stbi__skip(s, 88);
+
+   *x = stbi__get16be(s);
+   *y = stbi__get16be(s);
+   if (stbi__at_eof(s)) {
+      stbi__rewind( s);
+      return 0;
+   }
+   if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) {
+      stbi__rewind( s );
+      return 0;
+   }
+
+   stbi__skip(s, 8);
+
+   do {
+      stbi__pic_packet *packet;
+
+      if (num_packets==sizeof(packets)/sizeof(packets[0]))
+         return 0;
+
+      packet = &packets[num_packets++];
+      chained = stbi__get8(s);
+      packet->size    = stbi__get8(s);
+      packet->type    = stbi__get8(s);
+      packet->channel = stbi__get8(s);
+      act_comp |= packet->channel;
+
+      if (stbi__at_eof(s)) {
+          stbi__rewind( s );
+          return 0;
+      }
+      if (packet->size != 8) {
+          stbi__rewind( s );
+          return 0;
+      }
+   } while (chained);
+
+   *comp = (act_comp & 0x10 ? 4 : 3);
+
+   return 1;
+}
+#endif
+
+// *************************************************************************************************
+// Portable Gray Map and Portable Pixel Map loader
+// by Ken Miller
+//
+// PGM: http://netpbm.sourceforge.net/doc/pgm.html
+// PPM: http://netpbm.sourceforge.net/doc/ppm.html
+//
+// Known limitations:
+//    Does not support comments in the header section
+//    Does not support ASCII image data (formats P2 and P3)
+//    Does not support 16-bit-per-channel
+
+#ifndef STBI_NO_PNM
+
+static int      stbi__pnm_test(stbi__context *s)
+{
+   char p, t;
+   p = (char) stbi__get8(s);
+   t = (char) stbi__get8(s);
+   if (p != 'P' || (t != '5' && t != '6')) {
+       stbi__rewind( s );
+       return 0;
+   }
+   return 1;
+}
+
+static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri)
+{
+   stbi_uc *out;
+   STBI_NOTUSED(ri);
+
+   if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n))
+      return 0;
+
+   *x = s->img_x;
+   *y = s->img_y;
+   if (comp) *comp = s->img_n;
+
+   if (!stbi__mad3sizes_valid(s->img_n, s->img_x, s->img_y, 0))
+      return stbi__errpuc("too large", "PNM too large");
+
+   out = (stbi_uc *) stbi__malloc_mad3(s->img_n, s->img_x, s->img_y, 0);
+   if (!out) return stbi__errpuc("outofmem", "Out of memory");
+   stbi__getn(s, out, s->img_n * s->img_x * s->img_y);
+
+   if (req_comp && req_comp != s->img_n) {
+      out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y);
+      if (out == NULL) return out; // stbi__convert_format frees input on failure
+   }
+   return out;
+}
+
+static int      stbi__pnm_isspace(char c)
+{
+   return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r';
+}
+
+static void     stbi__pnm_skip_whitespace(stbi__context *s, char *c)
+{
+   for (;;) {
+      while (!stbi__at_eof(s) && stbi__pnm_isspace(*c))
+         *c = (char) stbi__get8(s);
+
+      if (stbi__at_eof(s) || *c != '#')
+         break;
+
+      while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' )
+         *c = (char) stbi__get8(s);
+   }
+}
+
+static int      stbi__pnm_isdigit(char c)
+{
+   return c >= '0' && c <= '9';
+}
+
+static int      stbi__pnm_getinteger(stbi__context *s, char *c)
+{
+   int value = 0;
+
+   while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) {
+      value = value*10 + (*c - '0');
+      *c = (char) stbi__get8(s);
+   }
+
+   return value;
+}
+
+static int      stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp)
+{
+   int maxv, dummy;
+   char c, p, t;
+
+   if (!x) x = &dummy;
+   if (!y) y = &dummy;
+   if (!comp) comp = &dummy;
+
+   stbi__rewind(s);
+
+   // Get identifier
+   p = (char) stbi__get8(s);
+   t = (char) stbi__get8(s);
+   if (p != 'P' || (t != '5' && t != '6')) {
+       stbi__rewind(s);
+       return 0;
+   }
+
+   *comp = (t == '6') ? 3 : 1;  // '5' is 1-component .pgm; '6' is 3-component .ppm
+
+   c = (char) stbi__get8(s);
+   stbi__pnm_skip_whitespace(s, &c);
+
+   *x = stbi__pnm_getinteger(s, &c); // read width
+   stbi__pnm_skip_whitespace(s, &c);
+
+   *y = stbi__pnm_getinteger(s, &c); // read height
+   stbi__pnm_skip_whitespace(s, &c);
+
+   maxv = stbi__pnm_getinteger(s, &c);  // read max value
+
+   if (maxv > 255)
+      return stbi__err("max value > 255", "PPM image not 8-bit");
+   else
+      return 1;
+}
+#endif
+
+static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)
+{
+   #ifndef STBI_NO_JPEG
+   if (stbi__jpeg_info(s, x, y, comp)) return 1;
+   #endif
+
+   #ifndef STBI_NO_PNG
+   if (stbi__png_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_GIF
+   if (stbi__gif_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_BMP
+   if (stbi__bmp_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PIC
+   if (stbi__pic_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PNM
+   if (stbi__pnm_info(s, x, y, comp))  return 1;
+   #endif
+
+   #ifndef STBI_NO_HDR
+   if (stbi__hdr_info(s, x, y, comp))  return 1;
+   #endif
+
+   // test tga last because it's a crappy test!
+   #ifndef STBI_NO_TGA
+   if (stbi__tga_info(s, x, y, comp))
+       return 1;
+   #endif
+   return stbi__err("unknown image type", "Image not of any known type, or corrupt");
+}
+
+static int stbi__is_16_main(stbi__context *s)
+{
+   #ifndef STBI_NO_PNG
+   if (stbi__png_is16(s))  return 1;
+   #endif
+
+   #ifndef STBI_NO_PSD
+   if (stbi__psd_is16(s))  return 1;
+   #endif
+
+   return 0;
+}
+
+#ifndef STBI_NO_STDIO
+STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp)
+{
+    FILE *f = stbi__fopen(filename, "rb");
+    int result;
+    if (!f) return stbi__err("can't fopen", "Unable to open file");
+    result = stbi_info_from_file(f, x, y, comp);
+    fclose(f);
+    return result;
+}
+
+STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp)
+{
+   int r;
+   stbi__context s;
+   long pos = ftell(f);
+   stbi__start_file(&s, f);
+   r = stbi__info_main(&s,x,y,comp);
+   fseek(f,pos,SEEK_SET);
+   return r;
+}
+
+STBIDEF int stbi_is_16_bit(char const *filename)
+{
+    FILE *f = stbi__fopen(filename, "rb");
+    int result;
+    if (!f) return stbi__err("can't fopen", "Unable to open file");
+    result = stbi_is_16_bit_from_file(f);
+    fclose(f);
+    return result;
+}
+
+STBIDEF int stbi_is_16_bit_from_file(FILE *f)
+{
+   int r;
+   stbi__context s;
+   long pos = ftell(f);
+   stbi__start_file(&s, f);
+   r = stbi__is_16_main(&s);
+   fseek(f,pos,SEEK_SET);
+   return r;
+}
+#endif // !STBI_NO_STDIO
+
+STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+   return stbi__info_main(&s,x,y,comp);
+}
+
+STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__is_16_main(&s);
+}
+
+STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);
+   return stbi__is_16_main(&s);
+}
+
+#endif // STB_IMAGE_IMPLEMENTATION
+
+/*
+   revision history:
+      2.20  (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs 
+      2.19  (2018-02-11) fix warning
+      2.18  (2018-01-30) fix warnings
+      2.17  (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug
+                         1-bit BMP
+                         *_is_16_bit api
+                         avoid warnings
+      2.16  (2017-07-23) all functions have 16-bit variants;
+                         STBI_NO_STDIO works again;
+                         compilation fixes;
+                         fix rounding in unpremultiply;
+                         optimize vertical flip;
+                         disable raw_len validation;
+                         documentation fixes
+      2.15  (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode;
+                         warning fixes; disable run-time SSE detection on gcc;
+                         uniform handling of optional "return" values;
+                         thread-safe initialization of zlib tables
+      2.14  (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
+      2.13  (2016-11-29) add 16-bit API, only supported for PNG right now
+      2.12  (2016-04-02) fix typo in 2.11 PSD fix that caused crashes
+      2.11  (2016-04-02) allocate large structures on the stack
+                         remove white matting for transparent PSD
+                         fix reported channel count for PNG & BMP
+                         re-enable SSE2 in non-gcc 64-bit
+                         support RGB-formatted JPEG
+                         read 16-bit PNGs (only as 8-bit)
+      2.10  (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED
+      2.09  (2016-01-16) allow comments in PNM files
+                         16-bit-per-pixel TGA (not bit-per-component)
+                         info() for TGA could break due to .hdr handling
+                         info() for BMP to shares code instead of sloppy parse
+                         can use STBI_REALLOC_SIZED if allocator doesn't support realloc
+                         code cleanup
+      2.08  (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA
+      2.07  (2015-09-13) fix compiler warnings
+                         partial animated GIF support
+                         limited 16-bpc PSD support
+                         #ifdef unused functions
+                         bug with < 92 byte PIC,PNM,HDR,TGA
+      2.06  (2015-04-19) fix bug where PSD returns wrong '*comp' value
+      2.05  (2015-04-19) fix bug in progressive JPEG handling, fix warning
+      2.04  (2015-04-15) try to re-enable SIMD on MinGW 64-bit
+      2.03  (2015-04-12) extra corruption checking (mmozeiko)
+                         stbi_set_flip_vertically_on_load (nguillemot)
+                         fix NEON support; fix mingw support
+      2.02  (2015-01-19) fix incorrect assert, fix warning
+      2.01  (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2
+      2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG
+      2.00  (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg)
+                         progressive JPEG (stb)
+                         PGM/PPM support (Ken Miller)
+                         STBI_MALLOC,STBI_REALLOC,STBI_FREE
+                         GIF bugfix -- seemingly never worked
+                         STBI_NO_*, STBI_ONLY_*
+      1.48  (2014-12-14) fix incorrectly-named assert()
+      1.47  (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb)
+                         optimize PNG (ryg)
+                         fix bug in interlaced PNG with user-specified channel count (stb)
+      1.46  (2014-08-26)
+              fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG
+      1.45  (2014-08-16)
+              fix MSVC-ARM internal compiler error by wrapping malloc
+      1.44  (2014-08-07)
+              various warning fixes from Ronny Chevalier
+      1.43  (2014-07-15)
+              fix MSVC-only compiler problem in code changed in 1.42
+      1.42  (2014-07-09)
+              don't define _CRT_SECURE_NO_WARNINGS (affects user code)
+              fixes to stbi__cleanup_jpeg path
+              added STBI_ASSERT to avoid requiring assert.h
+      1.41  (2014-06-25)
+              fix search&replace from 1.36 that messed up comments/error messages
+      1.40  (2014-06-22)
+              fix gcc struct-initialization warning
+      1.39  (2014-06-15)
+              fix to TGA optimization when req_comp != number of components in TGA;
+              fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite)
+              add support for BMP version 5 (more ignored fields)
+      1.38  (2014-06-06)
+              suppress MSVC warnings on integer casts truncating values
+              fix accidental rename of 'skip' field of I/O
+      1.37  (2014-06-04)
+              remove duplicate typedef
+      1.36  (2014-06-03)
+              convert to header file single-file library
+              if de-iphone isn't set, load iphone images color-swapped instead of returning NULL
+      1.35  (2014-05-27)
+              various warnings
+              fix broken STBI_SIMD path
+              fix bug where stbi_load_from_file no longer left file pointer in correct place
+              fix broken non-easy path for 32-bit BMP (possibly never used)
+              TGA optimization by Arseny Kapoulkine
+      1.34  (unknown)
+              use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case
+      1.33  (2011-07-14)
+              make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements
+      1.32  (2011-07-13)
+              support for "info" function for all supported filetypes (SpartanJ)
+      1.31  (2011-06-20)
+              a few more leak fixes, bug in PNG handling (SpartanJ)
+      1.30  (2011-06-11)
+              added ability to load files via callbacks to accomidate custom input streams (Ben Wenger)
+              removed deprecated format-specific test/load functions
+              removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway
+              error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha)
+              fix inefficiency in decoding 32-bit BMP (David Woo)
+      1.29  (2010-08-16)
+              various warning fixes from Aurelien Pocheville
+      1.28  (2010-08-01)
+              fix bug in GIF palette transparency (SpartanJ)
+      1.27  (2010-08-01)
+              cast-to-stbi_uc to fix warnings
+      1.26  (2010-07-24)
+              fix bug in file buffering for PNG reported by SpartanJ
+      1.25  (2010-07-17)
+              refix trans_data warning (Won Chun)
+      1.24  (2010-07-12)
+              perf improvements reading from files on platforms with lock-heavy fgetc()
+              minor perf improvements for jpeg
+              deprecated type-specific functions so we'll get feedback if they're needed
+              attempt to fix trans_data warning (Won Chun)
+      1.23    fixed bug in iPhone support
+      1.22  (2010-07-10)
+              removed image *writing* support
+              stbi_info support from Jetro Lauha
+              GIF support from Jean-Marc Lienher
+              iPhone PNG-extensions from James Brown
+              warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva)
+      1.21    fix use of 'stbi_uc' in header (reported by jon blow)
+      1.20    added support for Softimage PIC, by Tom Seddon
+      1.19    bug in interlaced PNG corruption check (found by ryg)
+      1.18  (2008-08-02)
+              fix a threading bug (local mutable static)
+      1.17    support interlaced PNG
+      1.16    major bugfix - stbi__convert_format converted one too many pixels
+      1.15    initialize some fields for thread safety
+      1.14    fix threadsafe conversion bug
+              header-file-only version (#define STBI_HEADER_FILE_ONLY before including)
+      1.13    threadsafe
+      1.12    const qualifiers in the API
+      1.11    Support installable IDCT, colorspace conversion routines
+      1.10    Fixes for 64-bit (don't use "unsigned long")
+              optimized upsampling by Fabian "ryg" Giesen
+      1.09    Fix format-conversion for PSD code (bad global variables!)
+      1.08    Thatcher Ulrich's PSD code integrated by Nicolas Schulz
+      1.07    attempt to fix C++ warning/errors again
+      1.06    attempt to fix C++ warning/errors again
+      1.05    fix TGA loading to return correct *comp and use good luminance calc
+      1.04    default float alpha is 1, not 255; use 'void *' for stbi_image_free
+      1.03    bugfixes to STBI_NO_STDIO, STBI_NO_HDR
+      1.02    support for (subset of) HDR files, float interface for preferred access to them
+      1.01    fix bug: possible bug in handling right-side up bmps... not sure
+              fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all
+      1.00    interface to zlib that skips zlib header
+      0.99    correct handling of alpha in palette
+      0.98    TGA loader by lonesock; dynamically add loaders (untested)
+      0.97    jpeg errors on too large a file; also catch another malloc failure
+      0.96    fix detection of invalid v value - particleman@mollyrocket forum
+      0.95    during header scan, seek to markers in case of padding
+      0.94    STBI_NO_STDIO to disable stdio usage; rename all #defines the same
+      0.93    handle jpegtran output; verbose errors
+      0.92    read 4,8,16,24,32-bit BMP files of several formats
+      0.91    output 24-bit Windows 3.0 BMP files
+      0.90    fix a few more warnings; bump version number to approach 1.0
+      0.61    bugfixes due to Marc LeBlanc, Christopher Lloyd
+      0.60    fix compiling as c++
+      0.59    fix warnings: merge Dave Moore's -Wall fixes
+      0.58    fix bug: zlib uncompressed mode len/nlen was wrong endian
+      0.57    fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available
+      0.56    fix bug: zlib uncompressed mode len vs. nlen
+      0.55    fix bug: restart_interval not initialized to 0
+      0.54    allow NULL for 'int *comp'
+      0.53    fix bug in png 3->4; speedup png decoding
+      0.52    png handles req_comp=3,4 directly; minor cleanup; jpeg comments
+      0.51    obey req_comp requests, 1-component jpegs return as 1-component,
+              on 'test' only check type, not whether we support this variant
+      0.50  (2006-11-19)
+              first released version
+*/
+
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this
+software, either in source code form or as a compiled binary, for any purpose,
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this
+software dedicate any and all copyright interest in the software to the public
+domain. We make this dedication for the benefit of the public at large and to
+the detriment of our heirs and successors. We intend this dedication to be an
+overt act of relinquishment in perpetuity of all present and future rights to
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/
diff --git a/src/external/stb_truetype.h b/src/external/stb_truetype.h
new file mode 100644
index 0000000..767f005
--- /dev/null
+++ b/src/external/stb_truetype.h
@@ -0,0 +1,4882 @@
+// stb_truetype.h - v1.21 - public domain
+// authored from 2009-2016 by Sean Barrett / RAD Game Tools
+//
+//   This library processes TrueType files:
+//        parse files
+//        extract glyph metrics
+//        extract glyph shapes
+//        render glyphs to one-channel bitmaps with antialiasing (box filter)
+//        render glyphs to one-channel SDF bitmaps (signed-distance field/function)
+//
+//   Todo:
+//        non-MS cmaps
+//        crashproof on bad data
+//        hinting? (no longer patented)
+//        cleartype-style AA?
+//        optimize: use simple memory allocator for intermediates
+//        optimize: build edge-list directly from curves
+//        optimize: rasterize directly from curves?
+//
+// ADDITIONAL CONTRIBUTORS
+//
+//   Mikko Mononen: compound shape support, more cmap formats
+//   Tor Andersson: kerning, subpixel rendering
+//   Dougall Johnson: OpenType / Type 2 font handling
+//   Daniel Ribeiro Maciel: basic GPOS-based kerning
+//
+//   Misc other:
+//       Ryan Gordon
+//       Simon Glass
+//       github:IntellectualKitty
+//       Imanol Celaya
+//       Daniel Ribeiro Maciel
+//
+//   Bug/warning reports/fixes:
+//       "Zer" on mollyrocket       Fabian "ryg" Giesen
+//       Cass Everitt               Martins Mozeiko
+//       stoiko (Haemimont Games)   Cap Petschulat
+//       Brian Hook                 Omar Cornut
+//       Walter van Niftrik         github:aloucks
+//       David Gow                  Peter LaValle
+//       David Given                Sergey Popov
+//       Ivan-Assen Ivanov          Giumo X. Clanjor
+//       Anthony Pesch              Higor Euripedes
+//       Johan Duparc               Thomas Fields
+//       Hou Qiming                 Derek Vinyard
+//       Rob Loach                  Cort Stratton
+//       Kenney Phillis Jr.         github:oyvindjam
+//       Brian Costabile            github:vassvik
+//       
+// VERSION HISTORY
+//
+//   1.21 (2019-02-25) fix warning
+//   1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics()
+//   1.19 (2018-02-11) GPOS kerning, STBTT_fmod
+//   1.18 (2018-01-29) add missing function
+//   1.17 (2017-07-23) make more arguments const; doc fix
+//   1.16 (2017-07-12) SDF support
+//   1.15 (2017-03-03) make more arguments const
+//   1.14 (2017-01-16) num-fonts-in-TTC function
+//   1.13 (2017-01-02) support OpenType fonts, certain Apple fonts
+//   1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual
+//   1.11 (2016-04-02) fix unused-variable warning
+//   1.10 (2016-04-02) user-defined fabs(); rare memory leak; remove duplicate typedef
+//   1.09 (2016-01-16) warning fix; avoid crash on outofmem; use allocation userdata properly
+//   1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges
+//   1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints;
+//                     variant PackFontRanges to pack and render in separate phases;
+//                     fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?);
+//                     fixed an assert() bug in the new rasterizer
+//                     replace assert() with STBTT_assert() in new rasterizer
+//
+//   Full history can be found at the end of this file.
+//
+// LICENSE
+//
+//   See end of file for license information.
+//
+// USAGE
+//
+//   Include this file in whatever places need to refer to it. In ONE C/C++
+//   file, write:
+//      #define STB_TRUETYPE_IMPLEMENTATION
+//   before the #include of this file. This expands out the actual
+//   implementation into that C/C++ file.
+//
+//   To make the implementation private to the file that generates the implementation,
+//      #define STBTT_STATIC
+//
+//   Simple 3D API (don't ship this, but it's fine for tools and quick start)
+//           stbtt_BakeFontBitmap()               -- bake a font to a bitmap for use as texture
+//           stbtt_GetBakedQuad()                 -- compute quad to draw for a given char
+//
+//   Improved 3D API (more shippable):
+//           #include "stb_rect_pack.h"           -- optional, but you really want it
+//           stbtt_PackBegin()
+//           stbtt_PackSetOversampling()          -- for improved quality on small fonts
+//           stbtt_PackFontRanges()               -- pack and renders
+//           stbtt_PackEnd()
+//           stbtt_GetPackedQuad()
+//
+//   "Load" a font file from a memory buffer (you have to keep the buffer loaded)
+//           stbtt_InitFont()
+//           stbtt_GetFontOffsetForIndex()        -- indexing for TTC font collections
+//           stbtt_GetNumberOfFonts()             -- number of fonts for TTC font collections
+//
+//   Render a unicode codepoint to a bitmap
+//           stbtt_GetCodepointBitmap()           -- allocates and returns a bitmap
+//           stbtt_MakeCodepointBitmap()          -- renders into bitmap you provide
+//           stbtt_GetCodepointBitmapBox()        -- how big the bitmap must be
+//
+//   Character advance/positioning
+//           stbtt_GetCodepointHMetrics()
+//           stbtt_GetFontVMetrics()
+//           stbtt_GetFontVMetricsOS2()
+//           stbtt_GetCodepointKernAdvance()
+//
+//   Starting with version 1.06, the rasterizer was replaced with a new,
+//   faster and generally-more-precise rasterizer. The new rasterizer more
+//   accurately measures pixel coverage for anti-aliasing, except in the case
+//   where multiple shapes overlap, in which case it overestimates the AA pixel
+//   coverage. Thus, anti-aliasing of intersecting shapes may look wrong. If
+//   this turns out to be a problem, you can re-enable the old rasterizer with
+//        #define STBTT_RASTERIZER_VERSION 1
+//   which will incur about a 15% speed hit.
+//
+// ADDITIONAL DOCUMENTATION
+//
+//   Immediately after this block comment are a series of sample programs.
+//
+//   After the sample programs is the "header file" section. This section
+//   includes documentation for each API function.
+//
+//   Some important concepts to understand to use this library:
+//
+//      Codepoint
+//         Characters are defined by unicode codepoints, e.g. 65 is
+//         uppercase A, 231 is lowercase c with a cedilla, 0x7e30 is
+//         the hiragana for "ma".
+//
+//      Glyph
+//         A visual character shape (every codepoint is rendered as
+//         some glyph)
+//
+//      Glyph index
+//         A font-specific integer ID representing a glyph
+//
+//      Baseline
+//         Glyph shapes are defined relative to a baseline, which is the
+//         bottom of uppercase characters. Characters extend both above
+//         and below the baseline.
+//
+//      Current Point
+//         As you draw text to the screen, you keep track of a "current point"
+//         which is the origin of each character. The current point's vertical
+//         position is the baseline. Even "baked fonts" use this model.
+//
+//      Vertical Font Metrics
+//         The vertical qualities of the font, used to vertically position
+//         and space the characters. See docs for stbtt_GetFontVMetrics.
+//
+//      Font Size in Pixels or Points
+//         The preferred interface for specifying font sizes in stb_truetype
+//         is to specify how tall the font's vertical extent should be in pixels.
+//         If that sounds good enough, skip the next paragraph.
+//
+//         Most font APIs instead use "points", which are a common typographic
+//         measurement for describing font size, defined as 72 points per inch.
+//         stb_truetype provides a point API for compatibility. However, true
+//         "per inch" conventions don't make much sense on computer displays
+//         since different monitors have different number of pixels per
+//         inch. For example, Windows traditionally uses a convention that
+//         there are 96 pixels per inch, thus making 'inch' measurements have
+//         nothing to do with inches, and thus effectively defining a point to
+//         be 1.333 pixels. Additionally, the TrueType font data provides
+//         an explicit scale factor to scale a given font's glyphs to points,
+//         but the author has observed that this scale factor is often wrong
+//         for non-commercial fonts, thus making fonts scaled in points
+//         according to the TrueType spec incoherently sized in practice.
+//
+// DETAILED USAGE:
+//
+//  Scale:
+//    Select how high you want the font to be, in points or pixels.
+//    Call ScaleForPixelHeight or ScaleForMappingEmToPixels to compute
+//    a scale factor SF that will be used by all other functions.
+//
+//  Baseline:
+//    You need to select a y-coordinate that is the baseline of where
+//    your text will appear. Call GetFontBoundingBox to get the baseline-relative
+//    bounding box for all characters. SF*-y0 will be the distance in pixels
+//    that the worst-case character could extend above the baseline, so if
+//    you want the top edge of characters to appear at the top of the
+//    screen where y=0, then you would set the baseline to SF*-y0.
+//
+//  Current point:
+//    Set the current point where the first character will appear. The
+//    first character could extend left of the current point; this is font
+//    dependent. You can either choose a current point that is the leftmost
+//    point and hope, or add some padding, or check the bounding box or
+//    left-side-bearing of the first character to be displayed and set
+//    the current point based on that.
+//
+//  Displaying a character:
+//    Compute the bounding box of the character. It will contain signed values
+//    relative to <current_point, baseline>. I.e. if it returns x0,y0,x1,y1,
+//    then the character should be displayed in the rectangle from
+//    <current_point+SF*x0, baseline+SF*y0> to <current_point+SF*x1,baseline+SF*y1).
+//
+//  Advancing for the next character:
+//    Call GlyphHMetrics, and compute 'current_point += SF * advance'.
+// 
+//
+// ADVANCED USAGE
+//
+//   Quality:
+//
+//    - Use the functions with Subpixel at the end to allow your characters
+//      to have subpixel positioning. Since the font is anti-aliased, not
+//      hinted, this is very import for quality. (This is not possible with
+//      baked fonts.)
+//
+//    - Kerning is now supported, and if you're supporting subpixel rendering
+//      then kerning is worth using to give your text a polished look.
+//
+//   Performance:
+//
+//    - Convert Unicode codepoints to glyph indexes and operate on the glyphs;
+//      if you don't do this, stb_truetype is forced to do the conversion on
+//      every call.
+//
+//    - There are a lot of memory allocations. We should modify it to take
+//      a temp buffer and allocate from the temp buffer (without freeing),
+//      should help performance a lot.
+//
+// NOTES
+//
+//   The system uses the raw data found in the .ttf file without changing it
+//   and without building auxiliary data structures. This is a bit inefficient
+//   on little-endian systems (the data is big-endian), but assuming you're
+//   caching the bitmaps or glyph shapes this shouldn't be a big deal.
+//
+//   It appears to be very hard to programmatically determine what font a
+//   given file is in a general way. I provide an API for this, but I don't
+//   recommend it.
+//
+//
+// PERFORMANCE MEASUREMENTS FOR 1.06:
+//
+//                      32-bit     64-bit
+//   Previous release:  8.83 s     7.68 s
+//   Pool allocations:  7.72 s     6.34 s
+//   Inline sort     :  6.54 s     5.65 s
+//   New rasterizer  :  5.63 s     5.00 s
+
+//////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////
+////
+////  SAMPLE PROGRAMS
+////
+//
+//  Incomplete text-in-3d-api example, which draws quads properly aligned to be lossless
+//
+#if 0
+#define STB_TRUETYPE_IMPLEMENTATION  // force following include to generate implementation
+#include "stb_truetype.h"
+
+unsigned char ttf_buffer[1<<20];
+unsigned char temp_bitmap[512*512];
+
+stbtt_bakedchar cdata[96]; // ASCII 32..126 is 95 glyphs
+GLuint ftex;
+
+void my_stbtt_initfont(void)
+{
+   fread(ttf_buffer, 1, 1<<20, fopen("c:/windows/fonts/times.ttf", "rb"));
+   stbtt_BakeFontBitmap(ttf_buffer,0, 32.0, temp_bitmap,512,512, 32,96, cdata); // no guarantee this fits!
+   // can free ttf_buffer at this point
+   glGenTextures(1, &ftex);
+   glBindTexture(GL_TEXTURE_2D, ftex);
+   glTexImage2D(GL_TEXTURE_2D, 0, GL_ALPHA, 512,512, 0, GL_ALPHA, GL_UNSIGNED_BYTE, temp_bitmap);
+   // can free temp_bitmap at this point
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+}
+
+void my_stbtt_print(float x, float y, char *text)
+{
+   // assume orthographic projection with units = screen pixels, origin at top left
+   glEnable(GL_TEXTURE_2D);
+   glBindTexture(GL_TEXTURE_2D, ftex);
+   glBegin(GL_QUADS);
+   while (*text) {
+      if (*text >= 32 && *text < 128) {
+         stbtt_aligned_quad q;
+         stbtt_GetBakedQuad(cdata, 512,512, *text-32, &x,&y,&q,1);//1=opengl & d3d10+,0=d3d9
+         glTexCoord2f(q.s0,q.t1); glVertex2f(q.x0,q.y0);
+         glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y0);
+         glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y1);
+         glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y1);
+      }
+      ++text;
+   }
+   glEnd();
+}
+#endif
+//
+//
+//////////////////////////////////////////////////////////////////////////////
+//
+// Complete program (this compiles): get a single bitmap, print as ASCII art
+//
+#if 0
+#include <stdio.h>
+#define STB_TRUETYPE_IMPLEMENTATION  // force following include to generate implementation
+#include "stb_truetype.h"
+
+char ttf_buffer[1<<25];
+
+int main(int argc, char **argv)
+{
+   stbtt_fontinfo font;
+   unsigned char *bitmap;
+   int w,h,i,j,c = (argc > 1 ? atoi(argv[1]) : 'a'), s = (argc > 2 ? atoi(argv[2]) : 20);
+
+   fread(ttf_buffer, 1, 1<<25, fopen(argc > 3 ? argv[3] : "c:/windows/fonts/arialbd.ttf", "rb"));
+
+   stbtt_InitFont(&font, ttf_buffer, stbtt_GetFontOffsetForIndex(ttf_buffer,0));
+   bitmap = stbtt_GetCodepointBitmap(&font, 0,stbtt_ScaleForPixelHeight(&font, s), c, &w, &h, 0,0);
+
+   for (j=0; j < h; ++j) {
+      for (i=0; i < w; ++i)
+         putchar(" .:ioVM@"[bitmap[j*w+i]>>5]);
+      putchar('\n');
+   }
+   return 0;
+}
+#endif 
+//
+// Output:
+//
+//     .ii.
+//    @@@@@@.
+//   V@Mio@@o
+//   :i.  V@V
+//     :oM@@M
+//   :@@@MM@M
+//   @@o  o@M
+//  :@@.  M@M
+//   @@@o@@@@
+//   :M@@V:@@.
+//  
+//////////////////////////////////////////////////////////////////////////////
+// 
+// Complete program: print "Hello World!" banner, with bugs
+//
+#if 0
+char buffer[24<<20];
+unsigned char screen[20][79];
+
+int main(int arg, char **argv)
+{
+   stbtt_fontinfo font;
+   int i,j,ascent,baseline,ch=0;
+   float scale, xpos=2; // leave a little padding in case the character extends left
+   char *text = "Heljo World!"; // intentionally misspelled to show 'lj' brokenness
+
+   fread(buffer, 1, 1000000, fopen("c:/windows/fonts/arialbd.ttf", "rb"));
+   stbtt_InitFont(&font, buffer, 0);
+
+   scale = stbtt_ScaleForPixelHeight(&font, 15);
+   stbtt_GetFontVMetrics(&font, &ascent,0,0);
+   baseline = (int) (ascent*scale);
+
+   while (text[ch]) {
+      int advance,lsb,x0,y0,x1,y1;
+      float x_shift = xpos - (float) floor(xpos);
+      stbtt_GetCodepointHMetrics(&font, text[ch], &advance, &lsb);
+      stbtt_GetCodepointBitmapBoxSubpixel(&font, text[ch], scale,scale,x_shift,0, &x0,&y0,&x1,&y1);
+      stbtt_MakeCodepointBitmapSubpixel(&font, &screen[baseline + y0][(int) xpos + x0], x1-x0,y1-y0, 79, scale,scale,x_shift,0, text[ch]);
+      // note that this stomps the old data, so where character boxes overlap (e.g. 'lj') it's wrong
+      // because this API is really for baking character bitmaps into textures. if you want to render
+      // a sequence of characters, you really need to render each bitmap to a temp buffer, then
+      // "alpha blend" that into the working buffer
+      xpos += (advance * scale);
+      if (text[ch+1])
+         xpos += scale*stbtt_GetCodepointKernAdvance(&font, text[ch],text[ch+1]);
+      ++ch;
+   }
+
+   for (j=0; j < 20; ++j) {
+      for (i=0; i < 78; ++i)
+         putchar(" .:ioVM@"[screen[j][i]>>5]);
+      putchar('\n');
+   }
+
+   return 0;
+}
+#endif
+
+
+//////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////
+////
+////   INTEGRATION WITH YOUR CODEBASE
+////
+////   The following sections allow you to supply alternate definitions
+////   of C library functions used by stb_truetype, e.g. if you don't
+////   link with the C runtime library.
+
+#ifdef STB_TRUETYPE_IMPLEMENTATION
+   // #define your own (u)stbtt_int8/16/32 before including to override this
+   #ifndef stbtt_uint8
+   typedef unsigned char   stbtt_uint8;
+   typedef signed   char   stbtt_int8;
+   typedef unsigned short  stbtt_uint16;
+   typedef signed   short  stbtt_int16;
+   typedef unsigned int    stbtt_uint32;
+   typedef signed   int    stbtt_int32;
+   #endif
+
+   typedef char stbtt__check_size32[sizeof(stbtt_int32)==4 ? 1 : -1];
+   typedef char stbtt__check_size16[sizeof(stbtt_int16)==2 ? 1 : -1];
+
+   // e.g. #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h
+   #ifndef STBTT_ifloor
+   #include <math.h>
+   #define STBTT_ifloor(x)   ((int) floor(x))
+   #define STBTT_iceil(x)    ((int) ceil(x))
+   #endif
+
+   #ifndef STBTT_sqrt
+   #include <math.h>
+   #define STBTT_sqrt(x)      sqrt(x)
+   #define STBTT_pow(x,y)     pow(x,y)
+   #endif
+
+   #ifndef STBTT_fmod
+   #include <math.h>
+   #define STBTT_fmod(x,y)    fmod(x,y)
+   #endif
+
+   #ifndef STBTT_cos
+   #include <math.h>
+   #define STBTT_cos(x)       cos(x)
+   #define STBTT_acos(x)      acos(x)
+   #endif
+
+   #ifndef STBTT_fabs
+   #include <math.h>
+   #define STBTT_fabs(x)      fabs(x)
+   #endif
+
+   // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h
+   #ifndef STBTT_malloc
+   #include <stdlib.h>
+   #define STBTT_malloc(x,u)  ((void)(u),malloc(x))
+   #define STBTT_free(x,u)    ((void)(u),free(x))
+   #endif
+
+   #ifndef STBTT_assert
+   #include <assert.h>
+   #define STBTT_assert(x)    assert(x)
+   #endif
+
+   #ifndef STBTT_strlen
+   #include <string.h>
+   #define STBTT_strlen(x)    strlen(x)
+   #endif
+
+   #ifndef STBTT_memcpy
+   #include <string.h>
+   #define STBTT_memcpy       memcpy
+   #define STBTT_memset       memset
+   #endif
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+////
+////   INTERFACE
+////
+////
+
+#ifndef __STB_INCLUDE_STB_TRUETYPE_H__
+#define __STB_INCLUDE_STB_TRUETYPE_H__
+
+#ifdef STBTT_STATIC
+#define STBTT_DEF static
+#else
+#define STBTT_DEF extern
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// private structure
+typedef struct
+{
+   unsigned char *data;
+   int cursor;
+   int size;
+} stbtt__buf;
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// TEXTURE BAKING API
+//
+// If you use this API, you only have to call two functions ever.
+//
+
+typedef struct
+{
+   unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap
+   float xoff,yoff,xadvance;
+} stbtt_bakedchar;
+
+STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset,  // font location (use offset=0 for plain .ttf)
+                                float pixel_height,                     // height of font in pixels
+                                unsigned char *pixels, int pw, int ph,  // bitmap to be filled in
+                                int first_char, int num_chars,          // characters to bake
+                                stbtt_bakedchar *chardata);             // you allocate this, it's num_chars long
+// if return is positive, the first unused row of the bitmap
+// if return is negative, returns the negative of the number of characters that fit
+// if return is 0, no characters fit and no rows were used
+// This uses a very crappy packing.
+
+typedef struct
+{
+   float x0,y0,s0,t0; // top-left
+   float x1,y1,s1,t1; // bottom-right
+} stbtt_aligned_quad;
+
+STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph,  // same data as above
+                               int char_index,             // character to display
+                               float *xpos, float *ypos,   // pointers to current position in screen pixel space
+                               stbtt_aligned_quad *q,      // output: quad to draw
+                               int opengl_fillrule);       // true if opengl fill rule; false if DX9 or earlier
+// Call GetBakedQuad with char_index = 'character - first_char', and it
+// creates the quad you need to draw and advances the current position.
+//
+// The coordinate system used assumes y increases downwards.
+//
+// Characters will extend both above and below the current position;
+// see discussion of "BASELINE" above.
+//
+// It's inefficient; you might want to c&p it and optimize it.
+
+STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap);
+// Query the font vertical metrics without having to create a font first.
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// NEW TEXTURE BAKING API
+//
+// This provides options for packing multiple fonts into one atlas, not
+// perfectly but better than nothing.
+
+typedef struct
+{
+   unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap
+   float xoff,yoff,xadvance;
+   float xoff2,yoff2;
+} stbtt_packedchar;
+
+typedef struct stbtt_pack_context stbtt_pack_context;
+typedef struct stbtt_fontinfo stbtt_fontinfo;
+#ifndef STB_RECT_PACK_VERSION
+typedef struct stbrp_rect stbrp_rect;
+#endif
+
+STBTT_DEF int  stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int width, int height, int stride_in_bytes, int padding, void *alloc_context);
+// Initializes a packing context stored in the passed-in stbtt_pack_context.
+// Future calls using this context will pack characters into the bitmap passed
+// in here: a 1-channel bitmap that is width * height. stride_in_bytes is
+// the distance from one row to the next (or 0 to mean they are packed tightly
+// together). "padding" is the amount of padding to leave between each
+// character (normally you want '1' for bitmaps you'll use as textures with
+// bilinear filtering).
+//
+// Returns 0 on failure, 1 on success.
+
+STBTT_DEF void stbtt_PackEnd  (stbtt_pack_context *spc);
+// Cleans up the packing context and frees all memory.
+
+#define STBTT_POINT_SIZE(x)   (-(x))
+
+STBTT_DEF int  stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size,
+                                int first_unicode_char_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range);
+// Creates character bitmaps from the font_index'th font found in fontdata (use
+// font_index=0 if you don't know what that is). It creates num_chars_in_range
+// bitmaps for characters with unicode values starting at first_unicode_char_in_range
+// and increasing. Data for how to render them is stored in chardata_for_range;
+// pass these to stbtt_GetPackedQuad to get back renderable quads.
+//
+// font_size is the full height of the character from ascender to descender,
+// as computed by stbtt_ScaleForPixelHeight. To use a point size as computed
+// by stbtt_ScaleForMappingEmToPixels, wrap the point size in STBTT_POINT_SIZE()
+// and pass that result as 'font_size':
+//       ...,                  20 , ... // font max minus min y is 20 pixels tall
+//       ..., STBTT_POINT_SIZE(20), ... // 'M' is 20 pixels tall
+
+typedef struct
+{
+   float font_size;
+   int first_unicode_codepoint_in_range;  // if non-zero, then the chars are continuous, and this is the first codepoint
+   int *array_of_unicode_codepoints;       // if non-zero, then this is an array of unicode codepoints
+   int num_chars;
+   stbtt_packedchar *chardata_for_range; // output
+   unsigned char h_oversample, v_oversample; // don't set these, they're used internally
+} stbtt_pack_range;
+
+STBTT_DEF int  stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges);
+// Creates character bitmaps from multiple ranges of characters stored in
+// ranges. This will usually create a better-packed bitmap than multiple
+// calls to stbtt_PackFontRange. Note that you can call this multiple
+// times within a single PackBegin/PackEnd.
+
+STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample);
+// Oversampling a font increases the quality by allowing higher-quality subpixel
+// positioning, and is especially valuable at smaller text sizes.
+//
+// This function sets the amount of oversampling for all following calls to
+// stbtt_PackFontRange(s) or stbtt_PackFontRangesGatherRects for a given
+// pack context. The default (no oversampling) is achieved by h_oversample=1
+// and v_oversample=1. The total number of pixels required is
+// h_oversample*v_oversample larger than the default; for example, 2x2
+// oversampling requires 4x the storage of 1x1. For best results, render
+// oversampled textures with bilinear filtering. Look at the readme in
+// stb/tests/oversample for information about oversampled fonts
+//
+// To use with PackFontRangesGather etc., you must set it before calls
+// call to PackFontRangesGatherRects.
+
+STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip);
+// If skip != 0, this tells stb_truetype to skip any codepoints for which
+// there is no corresponding glyph. If skip=0, which is the default, then
+// codepoints without a glyph recived the font's "missing character" glyph,
+// typically an empty box by convention.
+
+STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph,  // same data as above
+                               int char_index,             // character to display
+                               float *xpos, float *ypos,   // pointers to current position in screen pixel space
+                               stbtt_aligned_quad *q,      // output: quad to draw
+                               int align_to_integer);
+
+STBTT_DEF int  stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects);
+STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects);
+STBTT_DEF int  stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects);
+// Calling these functions in sequence is roughly equivalent to calling
+// stbtt_PackFontRanges(). If you more control over the packing of multiple
+// fonts, or if you want to pack custom data into a font texture, take a look
+// at the source to of stbtt_PackFontRanges() and create a custom version 
+// using these functions, e.g. call GatherRects multiple times,
+// building up a single array of rects, then call PackRects once,
+// then call RenderIntoRects repeatedly. This may result in a
+// better packing than calling PackFontRanges multiple times
+// (or it may not).
+
+// this is an opaque structure that you shouldn't mess with which holds
+// all the context needed from PackBegin to PackEnd.
+struct stbtt_pack_context {
+   void *user_allocator_context;
+   void *pack_info;
+   int   width;
+   int   height;
+   int   stride_in_bytes;
+   int   padding;
+   int   skip_missing;
+   unsigned int   h_oversample, v_oversample;
+   unsigned char *pixels;
+   void  *nodes;
+};
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// FONT LOADING
+//
+//
+
+STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data);
+// This function will determine the number of fonts in a font file.  TrueType
+// collection (.ttc) files may contain multiple fonts, while TrueType font
+// (.ttf) files only contain one font. The number of fonts can be used for
+// indexing with the previous function where the index is between zero and one
+// less than the total fonts. If an error occurs, -1 is returned.
+
+STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index);
+// Each .ttf/.ttc file may have more than one font. Each font has a sequential
+// index number starting from 0. Call this function to get the font offset for
+// a given index; it returns -1 if the index is out of range. A regular .ttf
+// file will only define one font and it always be at offset 0, so it will
+// return '0' for index 0, and -1 for all other indices.
+
+// The following structure is defined publicly so you can declare one on
+// the stack or as a global or etc, but you should treat it as opaque.
+struct stbtt_fontinfo
+{
+   void           * userdata;
+   unsigned char  * data;              // pointer to .ttf file
+   int              fontstart;         // offset of start of font
+
+   int numGlyphs;                     // number of glyphs, needed for range checking
+
+   int loca,head,glyf,hhea,hmtx,kern,gpos; // table locations as offset from start of .ttf
+   int index_map;                     // a cmap mapping for our chosen character encoding
+   int indexToLocFormat;              // format needed to map from glyph index to glyph
+
+   stbtt__buf cff;                    // cff font data
+   stbtt__buf charstrings;            // the charstring index
+   stbtt__buf gsubrs;                 // global charstring subroutines index
+   stbtt__buf subrs;                  // private charstring subroutines index
+   stbtt__buf fontdicts;              // array of font dicts
+   stbtt__buf fdselect;               // map from glyph to fontdict
+};
+
+STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset);
+// Given an offset into the file that defines a font, this function builds
+// the necessary cached info for the rest of the system. You must allocate
+// the stbtt_fontinfo yourself, and stbtt_InitFont will fill it out. You don't
+// need to do anything special to free it, because the contents are pure
+// value data with no additional data structures. Returns 0 on failure.
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// CHARACTER TO GLYPH-INDEX CONVERSIOn
+
+STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint);
+// If you're going to perform multiple operations on the same character
+// and you want a speed-up, call this function with the character you're
+// going to process, then use glyph-based functions instead of the
+// codepoint-based functions.
+// Returns 0 if the character codepoint is not defined in the font.
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// CHARACTER PROPERTIES
+//
+
+STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float pixels);
+// computes a scale factor to produce a font whose "height" is 'pixels' tall.
+// Height is measured as the distance from the highest ascender to the lowest
+// descender; in other words, it's equivalent to calling stbtt_GetFontVMetrics
+// and computing:
+//       scale = pixels / (ascent - descent)
+// so if you prefer to measure height by the ascent only, use a similar calculation.
+
+STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels);
+// computes a scale factor to produce a font whose EM size is mapped to
+// 'pixels' tall. This is probably what traditional APIs compute, but
+// I'm not positive.
+
+STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap);
+// ascent is the coordinate above the baseline the font extends; descent
+// is the coordinate below the baseline the font extends (i.e. it is typically negative)
+// lineGap is the spacing between one row's descent and the next row's ascent...
+// so you should advance the vertical position by "*ascent - *descent + *lineGap"
+//   these are expressed in unscaled coordinates, so you must multiply by
+//   the scale factor for a given size
+
+STBTT_DEF int  stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap);
+// analogous to GetFontVMetrics, but returns the "typographic" values from the OS/2
+// table (specific to MS/Windows TTF files).
+//
+// Returns 1 on success (table present), 0 on failure.
+
+STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1);
+// the bounding box around all possible characters
+
+STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing);
+// leftSideBearing is the offset from the current horizontal position to the left edge of the character
+// advanceWidth is the offset from the current horizontal position to the next horizontal position
+//   these are expressed in unscaled coordinates
+
+STBTT_DEF int  stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2);
+// an additional amount to add to the 'advance' value between ch1 and ch2
+
+STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1);
+// Gets the bounding box of the visible part of the glyph, in unscaled coordinates
+
+STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing);
+STBTT_DEF int  stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2);
+STBTT_DEF int  stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1);
+// as above, but takes one or more glyph indices for greater efficiency
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// GLYPH SHAPES (you probably don't need these, but they have to go before
+// the bitmaps for C declaration-order reasons)
+//
+
+#ifndef STBTT_vmove // you can predefine these to use different values (but why?)
+   enum {
+      STBTT_vmove=1,
+      STBTT_vline,
+      STBTT_vcurve,
+      STBTT_vcubic
+   };
+#endif
+
+#ifndef stbtt_vertex // you can predefine this to use different values
+                   // (we share this with other code at RAD)
+   #define stbtt_vertex_type short // can't use stbtt_int16 because that's not visible in the header file
+   typedef struct
+   {
+      stbtt_vertex_type x,y,cx,cy,cx1,cy1;
+      unsigned char type,padding;
+   } stbtt_vertex;
+#endif
+
+STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index);
+// returns non-zero if nothing is drawn for this glyph
+
+STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices);
+STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **vertices);
+// returns # of vertices and fills *vertices with the pointer to them
+//   these are expressed in "unscaled" coordinates
+//
+// The shape is a series of contours. Each one starts with
+// a STBTT_moveto, then consists of a series of mixed
+// STBTT_lineto and STBTT_curveto segments. A lineto
+// draws a line from previous endpoint to its x,y; a curveto
+// draws a quadratic bezier from previous endpoint to
+// its x,y, using cx,cy as the bezier control point.
+
+STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *vertices);
+// frees the data allocated above
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// BITMAP RENDERING
+//
+
+STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata);
+// frees the bitmap allocated below
+
+STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff);
+// allocates a large-enough single-channel 8bpp bitmap and renders the
+// specified character/glyph at the specified scale into it, with
+// antialiasing. 0 is no coverage (transparent), 255 is fully covered (opaque).
+// *width & *height are filled out with the width & height of the bitmap,
+// which is stored left-to-right, top-to-bottom.
+//
+// xoff/yoff are the offset it pixel space from the glyph origin to the top-left of the bitmap
+
+STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff);
+// the same as stbtt_GetCodepoitnBitmap, but you can specify a subpixel
+// shift for the character
+
+STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint);
+// the same as stbtt_GetCodepointBitmap, but you pass in storage for the bitmap
+// in the form of 'output', with row spacing of 'out_stride' bytes. the bitmap
+// is clipped to out_w/out_h bytes. Call stbtt_GetCodepointBitmapBox to get the
+// width and height and positioning info for it first.
+
+STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint);
+// same as stbtt_MakeCodepointBitmap, but you can specify a subpixel
+// shift for the character
+
+STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint);
+// same as stbtt_MakeCodepointBitmapSubpixel, but prefiltering
+// is performed (see stbtt_PackSetOversampling)
+
+STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1);
+// get the bbox of the bitmap centered around the glyph origin; so the
+// bitmap width is ix1-ix0, height is iy1-iy0, and location to place
+// the bitmap top left is (leftSideBearing*scale,iy0).
+// (Note that the bitmap uses y-increases-down, but the shape uses
+// y-increases-up, so CodepointBitmapBox and CodepointBox are inverted.)
+
+STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1);
+// same as stbtt_GetCodepointBitmapBox, but you can specify a subpixel
+// shift for the character
+
+// the following functions are equivalent to the above functions, but operate
+// on glyph indices instead of Unicode codepoints (for efficiency)
+STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff);
+STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff);
+STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph);
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph);
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int glyph);
+STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1);
+STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1);
+
+
+// @TODO: don't expose this structure
+typedef struct
+{
+   int w,h,stride;
+   unsigned char *pixels;
+} stbtt__bitmap;
+
+// rasterize a shape with quadratic beziers into a bitmap
+STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result,        // 1-channel bitmap to draw into
+                               float flatness_in_pixels,     // allowable error of curve in pixels
+                               stbtt_vertex *vertices,       // array of vertices defining shape
+                               int num_verts,                // number of vertices in above array
+                               float scale_x, float scale_y, // scale applied to input vertices
+                               float shift_x, float shift_y, // translation applied to input vertices
+                               int x_off, int y_off,         // another translation applied to input
+                               int invert,                   // if non-zero, vertically flip shape
+                               void *userdata);              // context for to STBTT_MALLOC
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Signed Distance Function (or Field) rendering
+
+STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata);
+// frees the SDF bitmap allocated below
+
+STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff);
+STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff);
+// These functions compute a discretized SDF field for a single character, suitable for storing
+// in a single-channel texture, sampling with bilinear filtering, and testing against
+// larger than some threshold to produce scalable fonts.
+//        info              --  the font
+//        scale             --  controls the size of the resulting SDF bitmap, same as it would be creating a regular bitmap
+//        glyph/codepoint   --  the character to generate the SDF for
+//        padding           --  extra "pixels" around the character which are filled with the distance to the character (not 0),
+//                                 which allows effects like bit outlines
+//        onedge_value      --  value 0-255 to test the SDF against to reconstruct the character (i.e. the isocontour of the character)
+//        pixel_dist_scale  --  what value the SDF should increase by when moving one SDF "pixel" away from the edge (on the 0..255 scale)
+//                                 if positive, > onedge_value is inside; if negative, < onedge_value is inside
+//        width,height      --  output height & width of the SDF bitmap (including padding)
+//        xoff,yoff         --  output origin of the character
+//        return value      --  a 2D array of bytes 0..255, width*height in size
+//
+// pixel_dist_scale & onedge_value are a scale & bias that allows you to make
+// optimal use of the limited 0..255 for your application, trading off precision
+// and special effects. SDF values outside the range 0..255 are clamped to 0..255.
+//
+// Example:
+//      scale = stbtt_ScaleForPixelHeight(22)
+//      padding = 5
+//      onedge_value = 180
+//      pixel_dist_scale = 180/5.0 = 36.0
+//
+//      This will create an SDF bitmap in which the character is about 22 pixels
+//      high but the whole bitmap is about 22+5+5=32 pixels high. To produce a filled
+//      shape, sample the SDF at each pixel and fill the pixel if the SDF value
+//      is greater than or equal to 180/255. (You'll actually want to antialias,
+//      which is beyond the scope of this example.) Additionally, you can compute
+//      offset outlines (e.g. to stroke the character border inside & outside,
+//      or only outside). For example, to fill outside the character up to 3 SDF
+//      pixels, you would compare against (180-36.0*3)/255 = 72/255. The above
+//      choice of variables maps a range from 5 pixels outside the shape to
+//      2 pixels inside the shape to 0..255; this is intended primarily for apply
+//      outside effects only (the interior range is needed to allow proper
+//      antialiasing of the font at *smaller* sizes)
+//
+// The function computes the SDF analytically at each SDF pixel, not by e.g.
+// building a higher-res bitmap and approximating it. In theory the quality
+// should be as high as possible for an SDF of this size & representation, but
+// unclear if this is true in practice (perhaps building a higher-res bitmap
+// and computing from that can allow drop-out prevention).
+//
+// The algorithm has not been optimized at all, so expect it to be slow
+// if computing lots of characters or very large sizes. 
+
+
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// Finding the right font...
+//
+// You should really just solve this offline, keep your own tables
+// of what font is what, and don't try to get it out of the .ttf file.
+// That's because getting it out of the .ttf file is really hard, because
+// the names in the file can appear in many possible encodings, in many
+// possible languages, and e.g. if you need a case-insensitive comparison,
+// the details of that depend on the encoding & language in a complex way
+// (actually underspecified in truetype, but also gigantic).
+//
+// But you can use the provided functions in two possible ways:
+//     stbtt_FindMatchingFont() will use *case-sensitive* comparisons on
+//             unicode-encoded names to try to find the font you want;
+//             you can run this before calling stbtt_InitFont()
+//
+//     stbtt_GetFontNameString() lets you get any of the various strings
+//             from the file yourself and do your own comparisons on them.
+//             You have to have called stbtt_InitFont() first.
+
+
+STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags);
+// returns the offset (not index) of the font that matches, or -1 if none
+//   if you use STBTT_MACSTYLE_DONTCARE, use a font name like "Arial Bold".
+//   if you use any other flag, use a font name like "Arial"; this checks
+//     the 'macStyle' header field; i don't know if fonts set this consistently
+#define STBTT_MACSTYLE_DONTCARE     0
+#define STBTT_MACSTYLE_BOLD         1
+#define STBTT_MACSTYLE_ITALIC       2
+#define STBTT_MACSTYLE_UNDERSCORE   4
+#define STBTT_MACSTYLE_NONE         8   // <= not same as 0, this makes us check the bitfield is 0
+
+STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2);
+// returns 1/0 whether the first string interpreted as utf8 is identical to
+// the second string interpreted as big-endian utf16... useful for strings from next func
+
+STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID);
+// returns the string (which may be big-endian double byte, e.g. for unicode)
+// and puts the length in bytes in *length.
+//
+// some of the values for the IDs are below; for more see the truetype spec:
+//     http://developer.apple.com/textfonts/TTRefMan/RM06/Chap6name.html
+//     http://www.microsoft.com/typography/otspec/name.htm
+
+enum { // platformID
+   STBTT_PLATFORM_ID_UNICODE   =0,
+   STBTT_PLATFORM_ID_MAC       =1,
+   STBTT_PLATFORM_ID_ISO       =2,
+   STBTT_PLATFORM_ID_MICROSOFT =3
+};
+
+enum { // encodingID for STBTT_PLATFORM_ID_UNICODE
+   STBTT_UNICODE_EID_UNICODE_1_0    =0,
+   STBTT_UNICODE_EID_UNICODE_1_1    =1,
+   STBTT_UNICODE_EID_ISO_10646      =2,
+   STBTT_UNICODE_EID_UNICODE_2_0_BMP=3,
+   STBTT_UNICODE_EID_UNICODE_2_0_FULL=4
+};
+
+enum { // encodingID for STBTT_PLATFORM_ID_MICROSOFT
+   STBTT_MS_EID_SYMBOL        =0,
+   STBTT_MS_EID_UNICODE_BMP   =1,
+   STBTT_MS_EID_SHIFTJIS      =2,
+   STBTT_MS_EID_UNICODE_FULL  =10
+};
+
+enum { // encodingID for STBTT_PLATFORM_ID_MAC; same as Script Manager codes
+   STBTT_MAC_EID_ROMAN        =0,   STBTT_MAC_EID_ARABIC       =4,
+   STBTT_MAC_EID_JAPANESE     =1,   STBTT_MAC_EID_HEBREW       =5,
+   STBTT_MAC_EID_CHINESE_TRAD =2,   STBTT_MAC_EID_GREEK        =6,
+   STBTT_MAC_EID_KOREAN       =3,   STBTT_MAC_EID_RUSSIAN      =7
+};
+
+enum { // languageID for STBTT_PLATFORM_ID_MICROSOFT; same as LCID...
+       // problematic because there are e.g. 16 english LCIDs and 16 arabic LCIDs
+   STBTT_MS_LANG_ENGLISH     =0x0409,   STBTT_MS_LANG_ITALIAN     =0x0410,
+   STBTT_MS_LANG_CHINESE     =0x0804,   STBTT_MS_LANG_JAPANESE    =0x0411,
+   STBTT_MS_LANG_DUTCH       =0x0413,   STBTT_MS_LANG_KOREAN      =0x0412,
+   STBTT_MS_LANG_FRENCH      =0x040c,   STBTT_MS_LANG_RUSSIAN     =0x0419,
+   STBTT_MS_LANG_GERMAN      =0x0407,   STBTT_MS_LANG_SPANISH     =0x0409,
+   STBTT_MS_LANG_HEBREW      =0x040d,   STBTT_MS_LANG_SWEDISH     =0x041D
+};
+
+enum { // languageID for STBTT_PLATFORM_ID_MAC
+   STBTT_MAC_LANG_ENGLISH      =0 ,   STBTT_MAC_LANG_JAPANESE     =11,
+   STBTT_MAC_LANG_ARABIC       =12,   STBTT_MAC_LANG_KOREAN       =23,
+   STBTT_MAC_LANG_DUTCH        =4 ,   STBTT_MAC_LANG_RUSSIAN      =32,
+   STBTT_MAC_LANG_FRENCH       =1 ,   STBTT_MAC_LANG_SPANISH      =6 ,
+   STBTT_MAC_LANG_GERMAN       =2 ,   STBTT_MAC_LANG_SWEDISH      =5 ,
+   STBTT_MAC_LANG_HEBREW       =10,   STBTT_MAC_LANG_CHINESE_SIMPLIFIED =33,
+   STBTT_MAC_LANG_ITALIAN      =3 ,   STBTT_MAC_LANG_CHINESE_TRAD =19
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // __STB_INCLUDE_STB_TRUETYPE_H__
+
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+////
+////   IMPLEMENTATION
+////
+////
+
+#ifdef STB_TRUETYPE_IMPLEMENTATION
+
+#ifndef STBTT_MAX_OVERSAMPLE
+#define STBTT_MAX_OVERSAMPLE   8
+#endif
+
+#if STBTT_MAX_OVERSAMPLE > 255
+#error "STBTT_MAX_OVERSAMPLE cannot be > 255"
+#endif
+
+typedef int stbtt__test_oversample_pow2[(STBTT_MAX_OVERSAMPLE & (STBTT_MAX_OVERSAMPLE-1)) == 0 ? 1 : -1];
+
+#ifndef STBTT_RASTERIZER_VERSION
+#define STBTT_RASTERIZER_VERSION 2
+#endif
+
+#ifdef _MSC_VER
+#define STBTT__NOTUSED(v)  (void)(v)
+#else
+#define STBTT__NOTUSED(v)  (void)sizeof(v)
+#endif
+
+//////////////////////////////////////////////////////////////////////////
+//
+// stbtt__buf helpers to parse data from file
+//
+
+static stbtt_uint8 stbtt__buf_get8(stbtt__buf *b)
+{
+   if (b->cursor >= b->size)
+      return 0;
+   return b->data[b->cursor++];
+}
+
+static stbtt_uint8 stbtt__buf_peek8(stbtt__buf *b)
+{
+   if (b->cursor >= b->size)
+      return 0;
+   return b->data[b->cursor];
+}
+
+static void stbtt__buf_seek(stbtt__buf *b, int o)
+{
+   STBTT_assert(!(o > b->size || o < 0));
+   b->cursor = (o > b->size || o < 0) ? b->size : o;
+}
+
+static void stbtt__buf_skip(stbtt__buf *b, int o)
+{
+   stbtt__buf_seek(b, b->cursor + o);
+}
+
+static stbtt_uint32 stbtt__buf_get(stbtt__buf *b, int n)
+{
+   stbtt_uint32 v = 0;
+   int i;
+   STBTT_assert(n >= 1 && n <= 4);
+   for (i = 0; i < n; i++)
+      v = (v << 8) | stbtt__buf_get8(b);
+   return v;
+}
+
+static stbtt__buf stbtt__new_buf(const void *p, size_t size)
+{
+   stbtt__buf r;
+   STBTT_assert(size < 0x40000000);
+   r.data = (stbtt_uint8*) p;
+   r.size = (int) size;
+   r.cursor = 0;
+   return r;
+}
+
+#define stbtt__buf_get16(b)  stbtt__buf_get((b), 2)
+#define stbtt__buf_get32(b)  stbtt__buf_get((b), 4)
+
+static stbtt__buf stbtt__buf_range(const stbtt__buf *b, int o, int s)
+{
+   stbtt__buf r = stbtt__new_buf(NULL, 0);
+   if (o < 0 || s < 0 || o > b->size || s > b->size - o) return r;
+   r.data = b->data + o;
+   r.size = s;
+   return r;
+}
+
+static stbtt__buf stbtt__cff_get_index(stbtt__buf *b)
+{
+   int count, start, offsize;
+   start = b->cursor;
+   count = stbtt__buf_get16(b);
+   if (count) {
+      offsize = stbtt__buf_get8(b);
+      STBTT_assert(offsize >= 1 && offsize <= 4);
+      stbtt__buf_skip(b, offsize * count);
+      stbtt__buf_skip(b, stbtt__buf_get(b, offsize) - 1);
+   }
+   return stbtt__buf_range(b, start, b->cursor - start);
+}
+
+static stbtt_uint32 stbtt__cff_int(stbtt__buf *b)
+{
+   int b0 = stbtt__buf_get8(b);
+   if (b0 >= 32 && b0 <= 246)       return b0 - 139;
+   else if (b0 >= 247 && b0 <= 250) return (b0 - 247)*256 + stbtt__buf_get8(b) + 108;
+   else if (b0 >= 251 && b0 <= 254) return -(b0 - 251)*256 - stbtt__buf_get8(b) - 108;
+   else if (b0 == 28)               return stbtt__buf_get16(b);
+   else if (b0 == 29)               return stbtt__buf_get32(b);
+   STBTT_assert(0);
+   return 0;
+}
+
+static void stbtt__cff_skip_operand(stbtt__buf *b) {
+   int v, b0 = stbtt__buf_peek8(b);
+   STBTT_assert(b0 >= 28);
+   if (b0 == 30) {
+      stbtt__buf_skip(b, 1);
+      while (b->cursor < b->size) {
+         v = stbtt__buf_get8(b);
+         if ((v & 0xF) == 0xF || (v >> 4) == 0xF)
+            break;
+      }
+   } else {
+      stbtt__cff_int(b);
+   }
+}
+
+static stbtt__buf stbtt__dict_get(stbtt__buf *b, int key)
+{
+   stbtt__buf_seek(b, 0);
+   while (b->cursor < b->size) {
+      int start = b->cursor, end, op;
+      while (stbtt__buf_peek8(b) >= 28)
+         stbtt__cff_skip_operand(b);
+      end = b->cursor;
+      op = stbtt__buf_get8(b);
+      if (op == 12)  op = stbtt__buf_get8(b) | 0x100;
+      if (op == key) return stbtt__buf_range(b, start, end-start);
+   }
+   return stbtt__buf_range(b, 0, 0);
+}
+
+static void stbtt__dict_get_ints(stbtt__buf *b, int key, int outcount, stbtt_uint32 *out)
+{
+   int i;
+   stbtt__buf operands = stbtt__dict_get(b, key);
+   for (i = 0; i < outcount && operands.cursor < operands.size; i++)
+      out[i] = stbtt__cff_int(&operands);
+}
+
+static int stbtt__cff_index_count(stbtt__buf *b)
+{
+   stbtt__buf_seek(b, 0);
+   return stbtt__buf_get16(b);
+}
+
+static stbtt__buf stbtt__cff_index_get(stbtt__buf b, int i)
+{
+   int count, offsize, start, end;
+   stbtt__buf_seek(&b, 0);
+   count = stbtt__buf_get16(&b);
+   offsize = stbtt__buf_get8(&b);
+   STBTT_assert(i >= 0 && i < count);
+   STBTT_assert(offsize >= 1 && offsize <= 4);
+   stbtt__buf_skip(&b, i*offsize);
+   start = stbtt__buf_get(&b, offsize);
+   end = stbtt__buf_get(&b, offsize);
+   return stbtt__buf_range(&b, 2+(count+1)*offsize+start, end - start);
+}
+
+//////////////////////////////////////////////////////////////////////////
+//
+// accessors to parse data from file
+//
+
+// on platforms that don't allow misaligned reads, if we want to allow
+// truetype fonts that aren't padded to alignment, define ALLOW_UNALIGNED_TRUETYPE
+
+#define ttBYTE(p)     (* (stbtt_uint8 *) (p))
+#define ttCHAR(p)     (* (stbtt_int8 *) (p))
+#define ttFixed(p)    ttLONG(p)
+
+static stbtt_uint16 ttUSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; }
+static stbtt_int16 ttSHORT(stbtt_uint8 *p)   { return p[0]*256 + p[1]; }
+static stbtt_uint32 ttULONG(stbtt_uint8 *p)  { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; }
+static stbtt_int32 ttLONG(stbtt_uint8 *p)    { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; }
+
+#define stbtt_tag4(p,c0,c1,c2,c3) ((p)[0] == (c0) && (p)[1] == (c1) && (p)[2] == (c2) && (p)[3] == (c3))
+#define stbtt_tag(p,str)           stbtt_tag4(p,str[0],str[1],str[2],str[3])
+
+static int stbtt__isfont(stbtt_uint8 *font)
+{
+   // check the version number
+   if (stbtt_tag4(font, '1',0,0,0))  return 1; // TrueType 1
+   if (stbtt_tag(font, "typ1"))   return 1; // TrueType with type 1 font -- we don't support this!
+   if (stbtt_tag(font, "OTTO"))   return 1; // OpenType with CFF
+   if (stbtt_tag4(font, 0,1,0,0)) return 1; // OpenType 1.0
+   if (stbtt_tag(font, "true"))   return 1; // Apple specification for TrueType fonts
+   return 0;
+}
+
+// @OPTIMIZE: binary search
+static stbtt_uint32 stbtt__find_table(stbtt_uint8 *data, stbtt_uint32 fontstart, const char *tag)
+{
+   stbtt_int32 num_tables = ttUSHORT(data+fontstart+4);
+   stbtt_uint32 tabledir = fontstart + 12;
+   stbtt_int32 i;
+   for (i=0; i < num_tables; ++i) {
+      stbtt_uint32 loc = tabledir + 16*i;
+      if (stbtt_tag(data+loc+0, tag))
+         return ttULONG(data+loc+8);
+   }
+   return 0;
+}
+
+static int stbtt_GetFontOffsetForIndex_internal(unsigned char *font_collection, int index)
+{
+   // if it's just a font, there's only one valid index
+   if (stbtt__isfont(font_collection))
+      return index == 0 ? 0 : -1;
+
+   // check if it's a TTC
+   if (stbtt_tag(font_collection, "ttcf")) {
+      // version 1?
+      if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) {
+         stbtt_int32 n = ttLONG(font_collection+8);
+         if (index >= n)
+            return -1;
+         return ttULONG(font_collection+12+index*4);
+      }
+   }
+   return -1;
+}
+
+static int stbtt_GetNumberOfFonts_internal(unsigned char *font_collection)
+{
+   // if it's just a font, there's only one valid font
+   if (stbtt__isfont(font_collection))
+      return 1;
+
+   // check if it's a TTC
+   if (stbtt_tag(font_collection, "ttcf")) {
+      // version 1?
+      if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) {
+         return ttLONG(font_collection+8);
+      }
+   }
+   return 0;
+}
+
+static stbtt__buf stbtt__get_subrs(stbtt__buf cff, stbtt__buf fontdict)
+{
+   stbtt_uint32 subrsoff = 0, private_loc[2] = { 0, 0 };
+   stbtt__buf pdict;
+   stbtt__dict_get_ints(&fontdict, 18, 2, private_loc);
+   if (!private_loc[1] || !private_loc[0]) return stbtt__new_buf(NULL, 0);
+   pdict = stbtt__buf_range(&cff, private_loc[1], private_loc[0]);
+   stbtt__dict_get_ints(&pdict, 19, 1, &subrsoff);
+   if (!subrsoff) return stbtt__new_buf(NULL, 0);
+   stbtt__buf_seek(&cff, private_loc[1]+subrsoff);
+   return stbtt__cff_get_index(&cff);
+}
+
+static int stbtt_InitFont_internal(stbtt_fontinfo *info, unsigned char *data, int fontstart)
+{
+   stbtt_uint32 cmap, t;
+   stbtt_int32 i,numTables;
+
+   info->data = data;
+   info->fontstart = fontstart;
+   info->cff = stbtt__new_buf(NULL, 0);
+
+   cmap = stbtt__find_table(data, fontstart, "cmap");       // required
+   info->loca = stbtt__find_table(data, fontstart, "loca"); // required
+   info->head = stbtt__find_table(data, fontstart, "head"); // required
+   info->glyf = stbtt__find_table(data, fontstart, "glyf"); // required
+   info->hhea = stbtt__find_table(data, fontstart, "hhea"); // required
+   info->hmtx = stbtt__find_table(data, fontstart, "hmtx"); // required
+   info->kern = stbtt__find_table(data, fontstart, "kern"); // not required
+   info->gpos = stbtt__find_table(data, fontstart, "GPOS"); // not required
+
+   if (!cmap || !info->head || !info->hhea || !info->hmtx)
+      return 0;
+   if (info->glyf) {
+      // required for truetype
+      if (!info->loca) return 0;
+   } else {
+      // initialization for CFF / Type2 fonts (OTF)
+      stbtt__buf b, topdict, topdictidx;
+      stbtt_uint32 cstype = 2, charstrings = 0, fdarrayoff = 0, fdselectoff = 0;
+      stbtt_uint32 cff;
+
+      cff = stbtt__find_table(data, fontstart, "CFF ");
+      if (!cff) return 0;
+
+      info->fontdicts = stbtt__new_buf(NULL, 0);
+      info->fdselect = stbtt__new_buf(NULL, 0);
+
+      // @TODO this should use size from table (not 512MB)
+      info->cff = stbtt__new_buf(data+cff, 512*1024*1024);
+      b = info->cff;
+
+      // read the header
+      stbtt__buf_skip(&b, 2);
+      stbtt__buf_seek(&b, stbtt__buf_get8(&b)); // hdrsize
+
+      // @TODO the name INDEX could list multiple fonts,
+      // but we just use the first one.
+      stbtt__cff_get_index(&b);  // name INDEX
+      topdictidx = stbtt__cff_get_index(&b);
+      topdict = stbtt__cff_index_get(topdictidx, 0);
+      stbtt__cff_get_index(&b);  // string INDEX
+      info->gsubrs = stbtt__cff_get_index(&b);
+
+      stbtt__dict_get_ints(&topdict, 17, 1, &charstrings);
+      stbtt__dict_get_ints(&topdict, 0x100 | 6, 1, &cstype);
+      stbtt__dict_get_ints(&topdict, 0x100 | 36, 1, &fdarrayoff);
+      stbtt__dict_get_ints(&topdict, 0x100 | 37, 1, &fdselectoff);
+      info->subrs = stbtt__get_subrs(b, topdict);
+
+      // we only support Type 2 charstrings
+      if (cstype != 2) return 0;
+      if (charstrings == 0) return 0;
+
+      if (fdarrayoff) {
+         // looks like a CID font
+         if (!fdselectoff) return 0;
+         stbtt__buf_seek(&b, fdarrayoff);
+         info->fontdicts = stbtt__cff_get_index(&b);
+         info->fdselect = stbtt__buf_range(&b, fdselectoff, b.size-fdselectoff);
+      }
+
+      stbtt__buf_seek(&b, charstrings);
+      info->charstrings = stbtt__cff_get_index(&b);
+   }
+
+   t = stbtt__find_table(data, fontstart, "maxp");
+   if (t)
+      info->numGlyphs = ttUSHORT(data+t+4);
+   else
+      info->numGlyphs = 0xffff;
+
+   // find a cmap encoding table we understand *now* to avoid searching
+   // later. (todo: could make this installable)
+   // the same regardless of glyph.
+   numTables = ttUSHORT(data + cmap + 2);
+   info->index_map = 0;
+   for (i=0; i < numTables; ++i) {
+      stbtt_uint32 encoding_record = cmap + 4 + 8 * i;
+      // find an encoding we understand:
+      switch(ttUSHORT(data+encoding_record)) {
+         case STBTT_PLATFORM_ID_MICROSOFT:
+            switch (ttUSHORT(data+encoding_record+2)) {
+               case STBTT_MS_EID_UNICODE_BMP:
+               case STBTT_MS_EID_UNICODE_FULL:
+                  // MS/Unicode
+                  info->index_map = cmap + ttULONG(data+encoding_record+4);
+                  break;
+            }
+            break;
+        case STBTT_PLATFORM_ID_UNICODE:
+            // Mac/iOS has these
+            // all the encodingIDs are unicode, so we don't bother to check it
+            info->index_map = cmap + ttULONG(data+encoding_record+4);
+            break;
+      }
+   }
+   if (info->index_map == 0)
+      return 0;
+
+   info->indexToLocFormat = ttUSHORT(data+info->head + 50);
+   return 1;
+}
+
+STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint)
+{
+   stbtt_uint8 *data = info->data;
+   stbtt_uint32 index_map = info->index_map;
+
+   stbtt_uint16 format = ttUSHORT(data + index_map + 0);
+   if (format == 0) { // apple byte encoding
+      stbtt_int32 bytes = ttUSHORT(data + index_map + 2);
+      if (unicode_codepoint < bytes-6)
+         return ttBYTE(data + index_map + 6 + unicode_codepoint);
+      return 0;
+   } else if (format == 6) {
+      stbtt_uint32 first = ttUSHORT(data + index_map + 6);
+      stbtt_uint32 count = ttUSHORT(data + index_map + 8);
+      if ((stbtt_uint32) unicode_codepoint >= first && (stbtt_uint32) unicode_codepoint < first+count)
+         return ttUSHORT(data + index_map + 10 + (unicode_codepoint - first)*2);
+      return 0;
+   } else if (format == 2) {
+      STBTT_assert(0); // @TODO: high-byte mapping for japanese/chinese/korean
+      return 0;
+   } else if (format == 4) { // standard mapping for windows fonts: binary search collection of ranges
+      stbtt_uint16 segcount = ttUSHORT(data+index_map+6) >> 1;
+      stbtt_uint16 searchRange = ttUSHORT(data+index_map+8) >> 1;
+      stbtt_uint16 entrySelector = ttUSHORT(data+index_map+10);
+      stbtt_uint16 rangeShift = ttUSHORT(data+index_map+12) >> 1;
+
+      // do a binary search of the segments
+      stbtt_uint32 endCount = index_map + 14;
+      stbtt_uint32 search = endCount;
+
+      if (unicode_codepoint > 0xffff)
+         return 0;
+
+      // they lie from endCount .. endCount + segCount
+      // but searchRange is the nearest power of two, so...
+      if (unicode_codepoint >= ttUSHORT(data + search + rangeShift*2))
+         search += rangeShift*2;
+
+      // now decrement to bias correctly to find smallest
+      search -= 2;
+      while (entrySelector) {
+         stbtt_uint16 end;
+         searchRange >>= 1;
+         end = ttUSHORT(data + search + searchRange*2);
+         if (unicode_codepoint > end)
+            search += searchRange*2;
+         --entrySelector;
+      }
+      search += 2;
+
+      {
+         stbtt_uint16 offset, start;
+         stbtt_uint16 item = (stbtt_uint16) ((search - endCount) >> 1);
+
+         STBTT_assert(unicode_codepoint <= ttUSHORT(data + endCount + 2*item));
+         start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item);
+         if (unicode_codepoint < start)
+            return 0;
+
+         offset = ttUSHORT(data + index_map + 14 + segcount*6 + 2 + 2*item);
+         if (offset == 0)
+            return (stbtt_uint16) (unicode_codepoint + ttSHORT(data + index_map + 14 + segcount*4 + 2 + 2*item));
+
+         return ttUSHORT(data + offset + (unicode_codepoint-start)*2 + index_map + 14 + segcount*6 + 2 + 2*item);
+      }
+   } else if (format == 12 || format == 13) {
+      stbtt_uint32 ngroups = ttULONG(data+index_map+12);
+      stbtt_int32 low,high;
+      low = 0; high = (stbtt_int32)ngroups;
+      // Binary search the right group.
+      while (low < high) {
+         stbtt_int32 mid = low + ((high-low) >> 1); // rounds down, so low <= mid < high
+         stbtt_uint32 start_char = ttULONG(data+index_map+16+mid*12);
+         stbtt_uint32 end_char = ttULONG(data+index_map+16+mid*12+4);
+         if ((stbtt_uint32) unicode_codepoint < start_char)
+            high = mid;
+         else if ((stbtt_uint32) unicode_codepoint > end_char)
+            low = mid+1;
+         else {
+            stbtt_uint32 start_glyph = ttULONG(data+index_map+16+mid*12+8);
+            if (format == 12)
+               return start_glyph + unicode_codepoint-start_char;
+            else // format == 13
+               return start_glyph;
+         }
+      }
+      return 0; // not found
+   }
+   // @TODO
+   STBTT_assert(0);
+   return 0;
+}
+
+STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices)
+{
+   return stbtt_GetGlyphShape(info, stbtt_FindGlyphIndex(info, unicode_codepoint), vertices);
+}
+
+static void stbtt_setvertex(stbtt_vertex *v, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy)
+{
+   v->type = type;
+   v->x = (stbtt_int16) x;
+   v->y = (stbtt_int16) y;
+   v->cx = (stbtt_int16) cx;
+   v->cy = (stbtt_int16) cy;
+}
+
+static int stbtt__GetGlyfOffset(const stbtt_fontinfo *info, int glyph_index)
+{
+   int g1,g2;
+
+   STBTT_assert(!info->cff.size);
+
+   if (glyph_index >= info->numGlyphs) return -1; // glyph index out of range
+   if (info->indexToLocFormat >= 2)    return -1; // unknown index->glyph map format
+
+   if (info->indexToLocFormat == 0) {
+      g1 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2) * 2;
+      g2 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2 + 2) * 2;
+   } else {
+      g1 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4);
+      g2 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4 + 4);
+   }
+
+   return g1==g2 ? -1 : g1; // if length is 0, return -1
+}
+
+static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1);
+
+STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1)
+{
+   if (info->cff.size) {
+      stbtt__GetGlyphInfoT2(info, glyph_index, x0, y0, x1, y1);
+   } else {
+      int g = stbtt__GetGlyfOffset(info, glyph_index);
+      if (g < 0) return 0;
+
+      if (x0) *x0 = ttSHORT(info->data + g + 2);
+      if (y0) *y0 = ttSHORT(info->data + g + 4);
+      if (x1) *x1 = ttSHORT(info->data + g + 6);
+      if (y1) *y1 = ttSHORT(info->data + g + 8);
+   }
+   return 1;
+}
+
+STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1)
+{
+   return stbtt_GetGlyphBox(info, stbtt_FindGlyphIndex(info,codepoint), x0,y0,x1,y1);
+}
+
+STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index)
+{
+   stbtt_int16 numberOfContours;
+   int g;
+   if (info->cff.size)
+      return stbtt__GetGlyphInfoT2(info, glyph_index, NULL, NULL, NULL, NULL) == 0;
+   g = stbtt__GetGlyfOffset(info, glyph_index);
+   if (g < 0) return 1;
+   numberOfContours = ttSHORT(info->data + g);
+   return numberOfContours == 0;
+}
+
+static int stbtt__close_shape(stbtt_vertex *vertices, int num_vertices, int was_off, int start_off,
+    stbtt_int32 sx, stbtt_int32 sy, stbtt_int32 scx, stbtt_int32 scy, stbtt_int32 cx, stbtt_int32 cy)
+{
+   if (start_off) {
+      if (was_off)
+         stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+scx)>>1, (cy+scy)>>1, cx,cy);
+      stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, sx,sy,scx,scy);
+   } else {
+      if (was_off)
+         stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve,sx,sy,cx,cy);
+      else
+         stbtt_setvertex(&vertices[num_vertices++], STBTT_vline,sx,sy,0,0);
+   }
+   return num_vertices;
+}
+
+static int stbtt__GetGlyphShapeTT(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices)
+{
+   stbtt_int16 numberOfContours;
+   stbtt_uint8 *endPtsOfContours;
+   stbtt_uint8 *data = info->data;
+   stbtt_vertex *vertices=0;
+   int num_vertices=0;
+   int g = stbtt__GetGlyfOffset(info, glyph_index);
+
+   *pvertices = NULL;
+
+   if (g < 0) return 0;
+
+   numberOfContours = ttSHORT(data + g);
+
+   if (numberOfContours > 0) {
+      stbtt_uint8 flags=0,flagcount;
+      stbtt_int32 ins, i,j=0,m,n, next_move, was_off=0, off, start_off=0;
+      stbtt_int32 x,y,cx,cy,sx,sy, scx,scy;
+      stbtt_uint8 *points;
+      endPtsOfContours = (data + g + 10);
+      ins = ttUSHORT(data + g + 10 + numberOfContours * 2);
+      points = data + g + 10 + numberOfContours * 2 + 2 + ins;
+
+      n = 1+ttUSHORT(endPtsOfContours + numberOfContours*2-2);
+
+      m = n + 2*numberOfContours;  // a loose bound on how many vertices we might need
+      vertices = (stbtt_vertex *) STBTT_malloc(m * sizeof(vertices[0]), info->userdata);
+      if (vertices == 0)
+         return 0;
+
+      next_move = 0;
+      flagcount=0;
+
+      // in first pass, we load uninterpreted data into the allocated array
+      // above, shifted to the end of the array so we won't overwrite it when
+      // we create our final data starting from the front
+
+      off = m - n; // starting offset for uninterpreted data, regardless of how m ends up being calculated
+
+      // first load flags
+
+      for (i=0; i < n; ++i) {
+         if (flagcount == 0) {
+            flags = *points++;
+            if (flags & 8)
+               flagcount = *points++;
+         } else
+            --flagcount;
+         vertices[off+i].type = flags;
+      }
+
+      // now load x coordinates
+      x=0;
+      for (i=0; i < n; ++i) {
+         flags = vertices[off+i].type;
+         if (flags & 2) {
+            stbtt_int16 dx = *points++;
+            x += (flags & 16) ? dx : -dx; // ???
+         } else {
+            if (!(flags & 16)) {
+               x = x + (stbtt_int16) (points[0]*256 + points[1]);
+               points += 2;
+            }
+         }
+         vertices[off+i].x = (stbtt_int16) x;
+      }
+
+      // now load y coordinates
+      y=0;
+      for (i=0; i < n; ++i) {
+         flags = vertices[off+i].type;
+         if (flags & 4) {
+            stbtt_int16 dy = *points++;
+            y += (flags & 32) ? dy : -dy; // ???
+         } else {
+            if (!(flags & 32)) {
+               y = y + (stbtt_int16) (points[0]*256 + points[1]);
+               points += 2;
+            }
+         }
+         vertices[off+i].y = (stbtt_int16) y;
+      }
+
+      // now convert them to our format
+      num_vertices=0;
+      sx = sy = cx = cy = scx = scy = 0;
+      for (i=0; i < n; ++i) {
+         flags = vertices[off+i].type;
+         x     = (stbtt_int16) vertices[off+i].x;
+         y     = (stbtt_int16) vertices[off+i].y;
+
+         if (next_move == i) {
+            if (i != 0)
+               num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy);
+
+            // now start the new one               
+            start_off = !(flags & 1);
+            if (start_off) {
+               // if we start off with an off-curve point, then when we need to find a point on the curve
+               // where we can start, and we need to save some state for when we wraparound.
+               scx = x;
+               scy = y;
+               if (!(vertices[off+i+1].type & 1)) {
+                  // next point is also a curve point, so interpolate an on-point curve
+                  sx = (x + (stbtt_int32) vertices[off+i+1].x) >> 1;
+                  sy = (y + (stbtt_int32) vertices[off+i+1].y) >> 1;
+               } else {
+                  // otherwise just use the next point as our start point
+                  sx = (stbtt_int32) vertices[off+i+1].x;
+                  sy = (stbtt_int32) vertices[off+i+1].y;
+                  ++i; // we're using point i+1 as the starting point, so skip it
+               }
+            } else {
+               sx = x;
+               sy = y;
+            }
+            stbtt_setvertex(&vertices[num_vertices++], STBTT_vmove,sx,sy,0,0);
+            was_off = 0;
+            next_move = 1 + ttUSHORT(endPtsOfContours+j*2);
+            ++j;
+         } else {
+            if (!(flags & 1)) { // if it's a curve
+               if (was_off) // two off-curve control points in a row means interpolate an on-curve midpoint
+                  stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+x)>>1, (cy+y)>>1, cx, cy);
+               cx = x;
+               cy = y;
+               was_off = 1;
+            } else {
+               if (was_off)
+                  stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, x,y, cx, cy);
+               else
+                  stbtt_setvertex(&vertices[num_vertices++], STBTT_vline, x,y,0,0);
+               was_off = 0;
+            }
+         }
+      }
+      num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy);
+   } else if (numberOfContours == -1) {
+      // Compound shapes.
+      int more = 1;
+      stbtt_uint8 *comp = data + g + 10;
+      num_vertices = 0;
+      vertices = 0;
+      while (more) {
+         stbtt_uint16 flags, gidx;
+         int comp_num_verts = 0, i;
+         stbtt_vertex *comp_verts = 0, *tmp = 0;
+         float mtx[6] = {1,0,0,1,0,0}, m, n;
+         
+         flags = ttSHORT(comp); comp+=2;
+         gidx = ttSHORT(comp); comp+=2;
+
+         if (flags & 2) { // XY values
+            if (flags & 1) { // shorts
+               mtx[4] = ttSHORT(comp); comp+=2;
+               mtx[5] = ttSHORT(comp); comp+=2;
+            } else {
+               mtx[4] = ttCHAR(comp); comp+=1;
+               mtx[5] = ttCHAR(comp); comp+=1;
+            }
+         }
+         else {
+            // @TODO handle matching point
+            STBTT_assert(0);
+         }
+         if (flags & (1<<3)) { // WE_HAVE_A_SCALE
+            mtx[0] = mtx[3] = ttSHORT(comp)/16384.0f; comp+=2;
+            mtx[1] = mtx[2] = 0;
+         } else if (flags & (1<<6)) { // WE_HAVE_AN_X_AND_YSCALE
+            mtx[0] = ttSHORT(comp)/16384.0f; comp+=2;
+            mtx[1] = mtx[2] = 0;
+            mtx[3] = ttSHORT(comp)/16384.0f; comp+=2;
+         } else if (flags & (1<<7)) { // WE_HAVE_A_TWO_BY_TWO
+            mtx[0] = ttSHORT(comp)/16384.0f; comp+=2;
+            mtx[1] = ttSHORT(comp)/16384.0f; comp+=2;
+            mtx[2] = ttSHORT(comp)/16384.0f; comp+=2;
+            mtx[3] = ttSHORT(comp)/16384.0f; comp+=2;
+         }
+         
+         // Find transformation scales.
+         m = (float) STBTT_sqrt(mtx[0]*mtx[0] + mtx[1]*mtx[1]);
+         n = (float) STBTT_sqrt(mtx[2]*mtx[2] + mtx[3]*mtx[3]);
+
+         // Get indexed glyph.
+         comp_num_verts = stbtt_GetGlyphShape(info, gidx, &comp_verts);
+         if (comp_num_verts > 0) {
+            // Transform vertices.
+            for (i = 0; i < comp_num_verts; ++i) {
+               stbtt_vertex* v = &comp_verts[i];
+               stbtt_vertex_type x,y;
+               x=v->x; y=v->y;
+               v->x = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4]));
+               v->y = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5]));
+               x=v->cx; y=v->cy;
+               v->cx = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4]));
+               v->cy = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5]));
+            }
+            // Append vertices.
+            tmp = (stbtt_vertex*)STBTT_malloc((num_vertices+comp_num_verts)*sizeof(stbtt_vertex), info->userdata);
+            if (!tmp) {
+               if (vertices) STBTT_free(vertices, info->userdata);
+               if (comp_verts) STBTT_free(comp_verts, info->userdata);
+               return 0;
+            }
+            if (num_vertices > 0) STBTT_memcpy(tmp, vertices, num_vertices*sizeof(stbtt_vertex));
+            STBTT_memcpy(tmp+num_vertices, comp_verts, comp_num_verts*sizeof(stbtt_vertex));
+            if (vertices) STBTT_free(vertices, info->userdata);
+            vertices = tmp;
+            STBTT_free(comp_verts, info->userdata);
+            num_vertices += comp_num_verts;
+         }
+         // More components ?
+         more = flags & (1<<5);
+      }
+   } else if (numberOfContours < 0) {
+      // @TODO other compound variations?
+      STBTT_assert(0);
+   } else {
+      // numberOfCounters == 0, do nothing
+   }
+
+   *pvertices = vertices;
+   return num_vertices;
+}
+
+typedef struct
+{
+   int bounds;
+   int started;
+   float first_x, first_y;
+   float x, y;
+   stbtt_int32 min_x, max_x, min_y, max_y;
+
+   stbtt_vertex *pvertices;
+   int num_vertices;
+} stbtt__csctx;
+
+#define STBTT__CSCTX_INIT(bounds) {bounds,0, 0,0, 0,0, 0,0,0,0, NULL, 0}
+
+static void stbtt__track_vertex(stbtt__csctx *c, stbtt_int32 x, stbtt_int32 y)
+{
+   if (x > c->max_x || !c->started) c->max_x = x;
+   if (y > c->max_y || !c->started) c->max_y = y;
+   if (x < c->min_x || !c->started) c->min_x = x;
+   if (y < c->min_y || !c->started) c->min_y = y;
+   c->started = 1;
+}
+
+static void stbtt__csctx_v(stbtt__csctx *c, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy, stbtt_int32 cx1, stbtt_int32 cy1)
+{
+   if (c->bounds) {
+      stbtt__track_vertex(c, x, y);
+      if (type == STBTT_vcubic) {
+         stbtt__track_vertex(c, cx, cy);
+         stbtt__track_vertex(c, cx1, cy1);
+      }
+   } else {
+      stbtt_setvertex(&c->pvertices[c->num_vertices], type, x, y, cx, cy);
+      c->pvertices[c->num_vertices].cx1 = (stbtt_int16) cx1;
+      c->pvertices[c->num_vertices].cy1 = (stbtt_int16) cy1;
+   }
+   c->num_vertices++;
+}
+
+static void stbtt__csctx_close_shape(stbtt__csctx *ctx)
+{
+   if (ctx->first_x != ctx->x || ctx->first_y != ctx->y)
+      stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->first_x, (int)ctx->first_y, 0, 0, 0, 0);
+}
+
+static void stbtt__csctx_rmove_to(stbtt__csctx *ctx, float dx, float dy)
+{
+   stbtt__csctx_close_shape(ctx);
+   ctx->first_x = ctx->x = ctx->x + dx;
+   ctx->first_y = ctx->y = ctx->y + dy;
+   stbtt__csctx_v(ctx, STBTT_vmove, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0);
+}
+
+static void stbtt__csctx_rline_to(stbtt__csctx *ctx, float dx, float dy)
+{
+   ctx->x += dx;
+   ctx->y += dy;
+   stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0);
+}
+
+static void stbtt__csctx_rccurve_to(stbtt__csctx *ctx, float dx1, float dy1, float dx2, float dy2, float dx3, float dy3)
+{
+   float cx1 = ctx->x + dx1;
+   float cy1 = ctx->y + dy1;
+   float cx2 = cx1 + dx2;
+   float cy2 = cy1 + dy2;
+   ctx->x = cx2 + dx3;
+   ctx->y = cy2 + dy3;
+   stbtt__csctx_v(ctx, STBTT_vcubic, (int)ctx->x, (int)ctx->y, (int)cx1, (int)cy1, (int)cx2, (int)cy2);
+}
+
+static stbtt__buf stbtt__get_subr(stbtt__buf idx, int n)
+{
+   int count = stbtt__cff_index_count(&idx);
+   int bias = 107;
+   if (count >= 33900)
+      bias = 32768;
+   else if (count >= 1240)
+      bias = 1131;
+   n += bias;
+   if (n < 0 || n >= count)
+      return stbtt__new_buf(NULL, 0);
+   return stbtt__cff_index_get(idx, n);
+}
+
+static stbtt__buf stbtt__cid_get_glyph_subrs(const stbtt_fontinfo *info, int glyph_index)
+{
+   stbtt__buf fdselect = info->fdselect;
+   int nranges, start, end, v, fmt, fdselector = -1, i;
+
+   stbtt__buf_seek(&fdselect, 0);
+   fmt = stbtt__buf_get8(&fdselect);
+   if (fmt == 0) {
+      // untested
+      stbtt__buf_skip(&fdselect, glyph_index);
+      fdselector = stbtt__buf_get8(&fdselect);
+   } else if (fmt == 3) {
+      nranges = stbtt__buf_get16(&fdselect);
+      start = stbtt__buf_get16(&fdselect);
+      for (i = 0; i < nranges; i++) {
+         v = stbtt__buf_get8(&fdselect);
+         end = stbtt__buf_get16(&fdselect);
+         if (glyph_index >= start && glyph_index < end) {
+            fdselector = v;
+            break;
+         }
+         start = end;
+      }
+   }
+   if (fdselector == -1) stbtt__new_buf(NULL, 0);
+   return stbtt__get_subrs(info->cff, stbtt__cff_index_get(info->fontdicts, fdselector));
+}
+
+static int stbtt__run_charstring(const stbtt_fontinfo *info, int glyph_index, stbtt__csctx *c)
+{
+   int in_header = 1, maskbits = 0, subr_stack_height = 0, sp = 0, v, i, b0;
+   int has_subrs = 0, clear_stack;
+   float s[48];
+   stbtt__buf subr_stack[10], subrs = info->subrs, b;
+   float f;
+
+#define STBTT__CSERR(s) (0)
+
+   // this currently ignores the initial width value, which isn't needed if we have hmtx
+   b = stbtt__cff_index_get(info->charstrings, glyph_index);
+   while (b.cursor < b.size) {
+      i = 0;
+      clear_stack = 1;
+      b0 = stbtt__buf_get8(&b);
+      switch (b0) {
+      // @TODO implement hinting
+      case 0x13: // hintmask
+      case 0x14: // cntrmask
+         if (in_header)
+            maskbits += (sp / 2); // implicit "vstem"
+         in_header = 0;
+         stbtt__buf_skip(&b, (maskbits + 7) / 8);
+         break;
+
+      case 0x01: // hstem
+      case 0x03: // vstem
+      case 0x12: // hstemhm
+      case 0x17: // vstemhm
+         maskbits += (sp / 2);
+         break;
+
+      case 0x15: // rmoveto
+         in_header = 0;
+         if (sp < 2) return STBTT__CSERR("rmoveto stack");
+         stbtt__csctx_rmove_to(c, s[sp-2], s[sp-1]);
+         break;
+      case 0x04: // vmoveto
+         in_header = 0;
+         if (sp < 1) return STBTT__CSERR("vmoveto stack");
+         stbtt__csctx_rmove_to(c, 0, s[sp-1]);
+         break;
+      case 0x16: // hmoveto
+         in_header = 0;
+         if (sp < 1) return STBTT__CSERR("hmoveto stack");
+         stbtt__csctx_rmove_to(c, s[sp-1], 0);
+         break;
+
+      case 0x05: // rlineto
+         if (sp < 2) return STBTT__CSERR("rlineto stack");
+         for (; i + 1 < sp; i += 2)
+            stbtt__csctx_rline_to(c, s[i], s[i+1]);
+         break;
+
+      // hlineto/vlineto and vhcurveto/hvcurveto alternate horizontal and vertical
+      // starting from a different place.
+
+      case 0x07: // vlineto
+         if (sp < 1) return STBTT__CSERR("vlineto stack");
+         goto vlineto;
+      case 0x06: // hlineto
+         if (sp < 1) return STBTT__CSERR("hlineto stack");
+         for (;;) {
+            if (i >= sp) break;
+            stbtt__csctx_rline_to(c, s[i], 0);
+            i++;
+      vlineto:
+            if (i >= sp) break;
+            stbtt__csctx_rline_to(c, 0, s[i]);
+            i++;
+         }
+         break;
+
+      case 0x1F: // hvcurveto
+         if (sp < 4) return STBTT__CSERR("hvcurveto stack");
+         goto hvcurveto;
+      case 0x1E: // vhcurveto
+         if (sp < 4) return STBTT__CSERR("vhcurveto stack");
+         for (;;) {
+            if (i + 3 >= sp) break;
+            stbtt__csctx_rccurve_to(c, 0, s[i], s[i+1], s[i+2], s[i+3], (sp - i == 5) ? s[i + 4] : 0.0f);
+            i += 4;
+      hvcurveto:
+            if (i + 3 >= sp) break;
+            stbtt__csctx_rccurve_to(c, s[i], 0, s[i+1], s[i+2], (sp - i == 5) ? s[i+4] : 0.0f, s[i+3]);
+            i += 4;
+         }
+         break;
+
+      case 0x08: // rrcurveto
+         if (sp < 6) return STBTT__CSERR("rcurveline stack");
+         for (; i + 5 < sp; i += 6)
+            stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]);
+         break;
+
+      case 0x18: // rcurveline
+         if (sp < 8) return STBTT__CSERR("rcurveline stack");
+         for (; i + 5 < sp - 2; i += 6)
+            stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]);
+         if (i + 1 >= sp) return STBTT__CSERR("rcurveline stack");
+         stbtt__csctx_rline_to(c, s[i], s[i+1]);
+         break;
+
+      case 0x19: // rlinecurve
+         if (sp < 8) return STBTT__CSERR("rlinecurve stack");
+         for (; i + 1 < sp - 6; i += 2)
+            stbtt__csctx_rline_to(c, s[i], s[i+1]);
+         if (i + 5 >= sp) return STBTT__CSERR("rlinecurve stack");
+         stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]);
+         break;
+
+      case 0x1A: // vvcurveto
+      case 0x1B: // hhcurveto
+         if (sp < 4) return STBTT__CSERR("(vv|hh)curveto stack");
+         f = 0.0;
+         if (sp & 1) { f = s[i]; i++; }
+         for (; i + 3 < sp; i += 4) {
+            if (b0 == 0x1B)
+               stbtt__csctx_rccurve_to(c, s[i], f, s[i+1], s[i+2], s[i+3], 0.0);
+            else
+               stbtt__csctx_rccurve_to(c, f, s[i], s[i+1], s[i+2], 0.0, s[i+3]);
+            f = 0.0;
+         }
+         break;
+
+      case 0x0A: // callsubr
+         if (!has_subrs) {
+            if (info->fdselect.size)
+               subrs = stbtt__cid_get_glyph_subrs(info, glyph_index);
+            has_subrs = 1;
+         }
+         // fallthrough
+      case 0x1D: // callgsubr
+         if (sp < 1) return STBTT__CSERR("call(g|)subr stack");
+         v = (int) s[--sp];
+         if (subr_stack_height >= 10) return STBTT__CSERR("recursion limit");
+         subr_stack[subr_stack_height++] = b;
+         b = stbtt__get_subr(b0 == 0x0A ? subrs : info->gsubrs, v);
+         if (b.size == 0) return STBTT__CSERR("subr not found");
+         b.cursor = 0;
+         clear_stack = 0;
+         break;
+
+      case 0x0B: // return
+         if (subr_stack_height <= 0) return STBTT__CSERR("return outside subr");
+         b = subr_stack[--subr_stack_height];
+         clear_stack = 0;
+         break;
+
+      case 0x0E: // endchar
+         stbtt__csctx_close_shape(c);
+         return 1;
+
+      case 0x0C: { // two-byte escape
+         float dx1, dx2, dx3, dx4, dx5, dx6, dy1, dy2, dy3, dy4, dy5, dy6;
+         float dx, dy;
+         int b1 = stbtt__buf_get8(&b);
+         switch (b1) {
+         // @TODO These "flex" implementations ignore the flex-depth and resolution,
+         // and always draw beziers.
+         case 0x22: // hflex
+            if (sp < 7) return STBTT__CSERR("hflex stack");
+            dx1 = s[0];
+            dx2 = s[1];
+            dy2 = s[2];
+            dx3 = s[3];
+            dx4 = s[4];
+            dx5 = s[5];
+            dx6 = s[6];
+            stbtt__csctx_rccurve_to(c, dx1, 0, dx2, dy2, dx3, 0);
+            stbtt__csctx_rccurve_to(c, dx4, 0, dx5, -dy2, dx6, 0);
+            break;
+
+         case 0x23: // flex
+            if (sp < 13) return STBTT__CSERR("flex stack");
+            dx1 = s[0];
+            dy1 = s[1];
+            dx2 = s[2];
+            dy2 = s[3];
+            dx3 = s[4];
+            dy3 = s[5];
+            dx4 = s[6];
+            dy4 = s[7];
+            dx5 = s[8];
+            dy5 = s[9];
+            dx6 = s[10];
+            dy6 = s[11];
+            //fd is s[12]
+            stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3);
+            stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6);
+            break;
+
+         case 0x24: // hflex1
+            if (sp < 9) return STBTT__CSERR("hflex1 stack");
+            dx1 = s[0];
+            dy1 = s[1];
+            dx2 = s[2];
+            dy2 = s[3];
+            dx3 = s[4];
+            dx4 = s[5];
+            dx5 = s[6];
+            dy5 = s[7];
+            dx6 = s[8];
+            stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, 0);
+            stbtt__csctx_rccurve_to(c, dx4, 0, dx5, dy5, dx6, -(dy1+dy2+dy5));
+            break;
+
+         case 0x25: // flex1
+            if (sp < 11) return STBTT__CSERR("flex1 stack");
+            dx1 = s[0];
+            dy1 = s[1];
+            dx2 = s[2];
+            dy2 = s[3];
+            dx3 = s[4];
+            dy3 = s[5];
+            dx4 = s[6];
+            dy4 = s[7];
+            dx5 = s[8];
+            dy5 = s[9];
+            dx6 = dy6 = s[10];
+            dx = dx1+dx2+dx3+dx4+dx5;
+            dy = dy1+dy2+dy3+dy4+dy5;
+            if (STBTT_fabs(dx) > STBTT_fabs(dy))
+               dy6 = -dy;
+            else
+               dx6 = -dx;
+            stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3);
+            stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6);
+            break;
+
+         default:
+            return STBTT__CSERR("unimplemented");
+         }
+      } break;
+
+      default:
+         if (b0 != 255 && b0 != 28 && (b0 < 32 || b0 > 254))
+            return STBTT__CSERR("reserved operator");
+
+         // push immediate
+         if (b0 == 255) {
+            f = (float)(stbtt_int32)stbtt__buf_get32(&b) / 0x10000;
+         } else {
+            stbtt__buf_skip(&b, -1);
+            f = (float)(stbtt_int16)stbtt__cff_int(&b);
+         }
+         if (sp >= 48) return STBTT__CSERR("push stack overflow");
+         s[sp++] = f;
+         clear_stack = 0;
+         break;
+      }
+      if (clear_stack) sp = 0;
+   }
+   return STBTT__CSERR("no endchar");
+
+#undef STBTT__CSERR
+}
+
+static int stbtt__GetGlyphShapeT2(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices)
+{
+   // runs the charstring twice, once to count and once to output (to avoid realloc)
+   stbtt__csctx count_ctx = STBTT__CSCTX_INIT(1);
+   stbtt__csctx output_ctx = STBTT__CSCTX_INIT(0);
+   if (stbtt__run_charstring(info, glyph_index, &count_ctx)) {
+      *pvertices = (stbtt_vertex*)STBTT_malloc(count_ctx.num_vertices*sizeof(stbtt_vertex), info->userdata);
+      output_ctx.pvertices = *pvertices;
+      if (stbtt__run_charstring(info, glyph_index, &output_ctx)) {
+         STBTT_assert(output_ctx.num_vertices == count_ctx.num_vertices);
+         return output_ctx.num_vertices;
+      }
+   }
+   *pvertices = NULL;
+   return 0;
+}
+
+static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1)
+{
+   stbtt__csctx c = STBTT__CSCTX_INIT(1);
+   int r = stbtt__run_charstring(info, glyph_index, &c);
+   if (x0)  *x0 = r ? c.min_x : 0;
+   if (y0)  *y0 = r ? c.min_y : 0;
+   if (x1)  *x1 = r ? c.max_x : 0;
+   if (y1)  *y1 = r ? c.max_y : 0;
+   return r ? c.num_vertices : 0;
+}
+
+STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices)
+{
+   if (!info->cff.size)
+      return stbtt__GetGlyphShapeTT(info, glyph_index, pvertices);
+   else
+      return stbtt__GetGlyphShapeT2(info, glyph_index, pvertices);
+}
+
+STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing)
+{
+   stbtt_uint16 numOfLongHorMetrics = ttUSHORT(info->data+info->hhea + 34);
+   if (glyph_index < numOfLongHorMetrics) {
+      if (advanceWidth)     *advanceWidth    = ttSHORT(info->data + info->hmtx + 4*glyph_index);
+      if (leftSideBearing)  *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*glyph_index + 2);
+   } else {
+      if (advanceWidth)     *advanceWidth    = ttSHORT(info->data + info->hmtx + 4*(numOfLongHorMetrics-1));
+      if (leftSideBearing)  *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*numOfLongHorMetrics + 2*(glyph_index - numOfLongHorMetrics));
+   }
+}
+
+static int  stbtt__GetGlyphKernInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2)
+{
+   stbtt_uint8 *data = info->data + info->kern;
+   stbtt_uint32 needle, straw;
+   int l, r, m;
+
+   // we only look at the first table. it must be 'horizontal' and format 0.
+   if (!info->kern)
+      return 0;
+   if (ttUSHORT(data+2) < 1) // number of tables, need at least 1
+      return 0;
+   if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format
+      return 0;
+
+   l = 0;
+   r = ttUSHORT(data+10) - 1;
+   needle = glyph1 << 16 | glyph2;
+   while (l <= r) {
+      m = (l + r) >> 1;
+      straw = ttULONG(data+18+(m*6)); // note: unaligned read
+      if (needle < straw)
+         r = m - 1;
+      else if (needle > straw)
+         l = m + 1;
+      else
+         return ttSHORT(data+22+(m*6));
+   }
+   return 0;
+}
+
+static stbtt_int32  stbtt__GetCoverageIndex(stbtt_uint8 *coverageTable, int glyph)
+{
+    stbtt_uint16 coverageFormat = ttUSHORT(coverageTable);
+    switch(coverageFormat) {
+        case 1: {
+            stbtt_uint16 glyphCount = ttUSHORT(coverageTable + 2);
+
+            // Binary search.
+            stbtt_int32 l=0, r=glyphCount-1, m;
+            int straw, needle=glyph;
+            while (l <= r) {
+                stbtt_uint8 *glyphArray = coverageTable + 4;
+                stbtt_uint16 glyphID;
+                m = (l + r) >> 1;
+                glyphID = ttUSHORT(glyphArray + 2 * m);
+                straw = glyphID;
+                if (needle < straw)
+                    r = m - 1;
+                else if (needle > straw)
+                    l = m + 1;
+                else {
+                     return m;
+                }
+            }
+        } break;
+
+        case 2: {
+            stbtt_uint16 rangeCount = ttUSHORT(coverageTable + 2);
+            stbtt_uint8 *rangeArray = coverageTable + 4;
+
+            // Binary search.
+            stbtt_int32 l=0, r=rangeCount-1, m;
+            int strawStart, strawEnd, needle=glyph;
+            while (l <= r) {
+                stbtt_uint8 *rangeRecord;
+                m = (l + r) >> 1;
+                rangeRecord = rangeArray + 6 * m;
+                strawStart = ttUSHORT(rangeRecord);
+                strawEnd = ttUSHORT(rangeRecord + 2);
+                if (needle < strawStart)
+                    r = m - 1;
+                else if (needle > strawEnd)
+                    l = m + 1;
+                else {
+                    stbtt_uint16 startCoverageIndex = ttUSHORT(rangeRecord + 4);
+                    return startCoverageIndex + glyph - strawStart;
+                }
+            }
+        } break;
+
+        default: {
+            // There are no other cases.
+            STBTT_assert(0);
+        } break;
+    }
+
+    return -1;
+}
+
+static stbtt_int32  stbtt__GetGlyphClass(stbtt_uint8 *classDefTable, int glyph)
+{
+    stbtt_uint16 classDefFormat = ttUSHORT(classDefTable);
+    switch(classDefFormat)
+    {
+        case 1: {
+            stbtt_uint16 startGlyphID = ttUSHORT(classDefTable + 2);
+            stbtt_uint16 glyphCount = ttUSHORT(classDefTable + 4);
+            stbtt_uint8 *classDef1ValueArray = classDefTable + 6;
+
+            if (glyph >= startGlyphID && glyph < startGlyphID + glyphCount)
+                return (stbtt_int32)ttUSHORT(classDef1ValueArray + 2 * (glyph - startGlyphID));
+
+            classDefTable = classDef1ValueArray + 2 * glyphCount;
+        } break;
+
+        case 2: {
+            stbtt_uint16 classRangeCount = ttUSHORT(classDefTable + 2);
+            stbtt_uint8 *classRangeRecords = classDefTable + 4;
+
+            // Binary search.
+            stbtt_int32 l=0, r=classRangeCount-1, m;
+            int strawStart, strawEnd, needle=glyph;
+            while (l <= r) {
+                stbtt_uint8 *classRangeRecord;
+                m = (l + r) >> 1;
+                classRangeRecord = classRangeRecords + 6 * m;
+                strawStart = ttUSHORT(classRangeRecord);
+                strawEnd = ttUSHORT(classRangeRecord + 2);
+                if (needle < strawStart)
+                    r = m - 1;
+                else if (needle > strawEnd)
+                    l = m + 1;
+                else
+                    return (stbtt_int32)ttUSHORT(classRangeRecord + 4);
+            }
+
+            classDefTable = classRangeRecords + 6 * classRangeCount;
+        } break;
+
+        default: {
+            // There are no other cases.
+            STBTT_assert(0);
+        } break;
+    }
+
+    return -1;
+}
+
+// Define to STBTT_assert(x) if you want to break on unimplemented formats.
+#define STBTT_GPOS_TODO_assert(x)
+
+static stbtt_int32  stbtt__GetGlyphGPOSInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2)
+{
+    stbtt_uint16 lookupListOffset;
+    stbtt_uint8 *lookupList;
+    stbtt_uint16 lookupCount;
+    stbtt_uint8 *data;
+    stbtt_int32 i;
+
+    if (!info->gpos) return 0;
+
+    data = info->data + info->gpos;
+
+    if (ttUSHORT(data+0) != 1) return 0; // Major version 1
+    if (ttUSHORT(data+2) != 0) return 0; // Minor version 0
+
+    lookupListOffset = ttUSHORT(data+8);
+    lookupList = data + lookupListOffset;
+    lookupCount = ttUSHORT(lookupList);
+
+    for (i=0; i<lookupCount; ++i) {
+        stbtt_uint16 lookupOffset = ttUSHORT(lookupList + 2 + 2 * i);
+        stbtt_uint8 *lookupTable = lookupList + lookupOffset;
+
+        stbtt_uint16 lookupType = ttUSHORT(lookupTable);
+        stbtt_uint16 subTableCount = ttUSHORT(lookupTable + 4);
+        stbtt_uint8 *subTableOffsets = lookupTable + 6;
+        switch(lookupType) {
+            case 2: { // Pair Adjustment Positioning Subtable
+                stbtt_int32 sti;
+                for (sti=0; sti<subTableCount; sti++) {
+                    stbtt_uint16 subtableOffset = ttUSHORT(subTableOffsets + 2 * sti);
+                    stbtt_uint8 *table = lookupTable + subtableOffset;
+                    stbtt_uint16 posFormat = ttUSHORT(table);
+                    stbtt_uint16 coverageOffset = ttUSHORT(table + 2);
+                    stbtt_int32 coverageIndex = stbtt__GetCoverageIndex(table + coverageOffset, glyph1);
+                    if (coverageIndex == -1) continue;
+
+                    switch (posFormat) {
+                        case 1: {
+                            stbtt_int32 l, r, m;
+                            int straw, needle;
+                            stbtt_uint16 valueFormat1 = ttUSHORT(table + 4);
+                            stbtt_uint16 valueFormat2 = ttUSHORT(table + 6);
+                            stbtt_int32 valueRecordPairSizeInBytes = 2;
+                            stbtt_uint16 pairSetCount = ttUSHORT(table + 8);
+                            stbtt_uint16 pairPosOffset = ttUSHORT(table + 10 + 2 * coverageIndex);
+                            stbtt_uint8 *pairValueTable = table + pairPosOffset;
+                            stbtt_uint16 pairValueCount = ttUSHORT(pairValueTable);
+                            stbtt_uint8 *pairValueArray = pairValueTable + 2;
+                            // TODO: Support more formats.
+                            STBTT_GPOS_TODO_assert(valueFormat1 == 4);
+                            if (valueFormat1 != 4) return 0;
+                            STBTT_GPOS_TODO_assert(valueFormat2 == 0);
+                            if (valueFormat2 != 0) return 0;
+
+                            STBTT_assert(coverageIndex < pairSetCount);
+                            STBTT__NOTUSED(pairSetCount);
+
+                            needle=glyph2;
+                            r=pairValueCount-1;
+                            l=0;
+
+                            // Binary search.
+                            while (l <= r) {
+                                stbtt_uint16 secondGlyph;
+                                stbtt_uint8 *pairValue;
+                                m = (l + r) >> 1;
+                                pairValue = pairValueArray + (2 + valueRecordPairSizeInBytes) * m;
+                                secondGlyph = ttUSHORT(pairValue);
+                                straw = secondGlyph;
+                                if (needle < straw)
+                                    r = m - 1;
+                                else if (needle > straw)
+                                    l = m + 1;
+                                else {
+                                    stbtt_int16 xAdvance = ttSHORT(pairValue + 2);
+                                    return xAdvance;
+                                }
+                            }
+                        } break;
+
+                        case 2: {
+                            stbtt_uint16 valueFormat1 = ttUSHORT(table + 4);
+                            stbtt_uint16 valueFormat2 = ttUSHORT(table + 6);
+
+                            stbtt_uint16 classDef1Offset = ttUSHORT(table + 8);
+                            stbtt_uint16 classDef2Offset = ttUSHORT(table + 10);
+                            int glyph1class = stbtt__GetGlyphClass(table + classDef1Offset, glyph1);
+                            int glyph2class = stbtt__GetGlyphClass(table + classDef2Offset, glyph2);
+
+                            stbtt_uint16 class1Count = ttUSHORT(table + 12);
+                            stbtt_uint16 class2Count = ttUSHORT(table + 14);
+                            STBTT_assert(glyph1class < class1Count);
+                            STBTT_assert(glyph2class < class2Count);
+
+                            // TODO: Support more formats.
+                            STBTT_GPOS_TODO_assert(valueFormat1 == 4);
+                            if (valueFormat1 != 4) return 0;
+                            STBTT_GPOS_TODO_assert(valueFormat2 == 0);
+                            if (valueFormat2 != 0) return 0;
+
+                            if (glyph1class >= 0 && glyph1class < class1Count && glyph2class >= 0 && glyph2class < class2Count) {
+                                stbtt_uint8 *class1Records = table + 16;
+                                stbtt_uint8 *class2Records = class1Records + 2 * (glyph1class * class2Count);
+                                stbtt_int16 xAdvance = ttSHORT(class2Records + 2 * glyph2class);
+                                return xAdvance;
+                            }
+                        } break;
+
+                        default: {
+                            // There are no other cases.
+                            STBTT_assert(0);
+                            break;
+                        };
+                    }
+                }
+                break;
+            };
+
+            default:
+                // TODO: Implement other stuff.
+                break;
+        }
+    }
+
+    return 0;
+}
+
+STBTT_DEF int  stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int g1, int g2)
+{
+   int xAdvance = 0;
+
+   if (info->gpos)
+      xAdvance += stbtt__GetGlyphGPOSInfoAdvance(info, g1, g2);
+
+   if (info->kern)
+      xAdvance += stbtt__GetGlyphKernInfoAdvance(info, g1, g2);
+
+   return xAdvance;
+}
+
+STBTT_DEF int  stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2)
+{
+   if (!info->kern && !info->gpos) // if no kerning table, don't waste time looking up both codepoint->glyphs
+      return 0;
+   return stbtt_GetGlyphKernAdvance(info, stbtt_FindGlyphIndex(info,ch1), stbtt_FindGlyphIndex(info,ch2));
+}
+
+STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing)
+{
+   stbtt_GetGlyphHMetrics(info, stbtt_FindGlyphIndex(info,codepoint), advanceWidth, leftSideBearing);
+}
+
+STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap)
+{
+   if (ascent ) *ascent  = ttSHORT(info->data+info->hhea + 4);
+   if (descent) *descent = ttSHORT(info->data+info->hhea + 6);
+   if (lineGap) *lineGap = ttSHORT(info->data+info->hhea + 8);
+}
+
+STBTT_DEF int  stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap)
+{
+   int tab = stbtt__find_table(info->data, info->fontstart, "OS/2");
+   if (!tab)
+      return 0;
+   if (typoAscent ) *typoAscent  = ttSHORT(info->data+tab + 68);
+   if (typoDescent) *typoDescent = ttSHORT(info->data+tab + 70);
+   if (typoLineGap) *typoLineGap = ttSHORT(info->data+tab + 72);
+   return 1;
+}
+
+STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1)
+{
+   *x0 = ttSHORT(info->data + info->head + 36);
+   *y0 = ttSHORT(info->data + info->head + 38);
+   *x1 = ttSHORT(info->data + info->head + 40);
+   *y1 = ttSHORT(info->data + info->head + 42);
+}
+
+STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float height)
+{
+   int fheight = ttSHORT(info->data + info->hhea + 4) - ttSHORT(info->data + info->hhea + 6);
+   return (float) height / fheight;
+}
+
+STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels)
+{
+   int unitsPerEm = ttUSHORT(info->data + info->head + 18);
+   return pixels / unitsPerEm;
+}
+
+STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *v)
+{
+   STBTT_free(v, info->userdata);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// antialiasing software rasterizer
+//
+
+STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1)
+{
+   int x0=0,y0=0,x1,y1; // =0 suppresses compiler warning
+   if (!stbtt_GetGlyphBox(font, glyph, &x0,&y0,&x1,&y1)) {
+      // e.g. space character
+      if (ix0) *ix0 = 0;
+      if (iy0) *iy0 = 0;
+      if (ix1) *ix1 = 0;
+      if (iy1) *iy1 = 0;
+   } else {
+      // move to integral bboxes (treating pixels as little squares, what pixels get touched)?
+      if (ix0) *ix0 = STBTT_ifloor( x0 * scale_x + shift_x);
+      if (iy0) *iy0 = STBTT_ifloor(-y1 * scale_y + shift_y);
+      if (ix1) *ix1 = STBTT_iceil ( x1 * scale_x + shift_x);
+      if (iy1) *iy1 = STBTT_iceil (-y0 * scale_y + shift_y);
+   }
+}
+
+STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1)
+{
+   stbtt_GetGlyphBitmapBoxSubpixel(font, glyph, scale_x, scale_y,0.0f,0.0f, ix0, iy0, ix1, iy1);
+}
+
+STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1)
+{
+   stbtt_GetGlyphBitmapBoxSubpixel(font, stbtt_FindGlyphIndex(font,codepoint), scale_x, scale_y,shift_x,shift_y, ix0,iy0,ix1,iy1);
+}
+
+STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1)
+{
+   stbtt_GetCodepointBitmapBoxSubpixel(font, codepoint, scale_x, scale_y,0.0f,0.0f, ix0,iy0,ix1,iy1);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+//  Rasterizer
+
+typedef struct stbtt__hheap_chunk
+{
+   struct stbtt__hheap_chunk *next;
+} stbtt__hheap_chunk;
+
+typedef struct stbtt__hheap
+{
+   struct stbtt__hheap_chunk *head;
+   void   *first_free;
+   int    num_remaining_in_head_chunk;
+} stbtt__hheap;
+
+static void *stbtt__hheap_alloc(stbtt__hheap *hh, size_t size, void *userdata)
+{
+   if (hh->first_free) {
+      void *p = hh->first_free;
+      hh->first_free = * (void **) p;
+      return p;
+   } else {
+      if (hh->num_remaining_in_head_chunk == 0) {
+         int count = (size < 32 ? 2000 : size < 128 ? 800 : 100);
+         stbtt__hheap_chunk *c = (stbtt__hheap_chunk *) STBTT_malloc(sizeof(stbtt__hheap_chunk) + size * count, userdata);
+         if (c == NULL)
+            return NULL;
+         c->next = hh->head;
+         hh->head = c;
+         hh->num_remaining_in_head_chunk = count;
+      }
+      --hh->num_remaining_in_head_chunk;
+      return (char *) (hh->head) + sizeof(stbtt__hheap_chunk) + size * hh->num_remaining_in_head_chunk;
+   }
+}
+
+static void stbtt__hheap_free(stbtt__hheap *hh, void *p)
+{
+   *(void **) p = hh->first_free;
+   hh->first_free = p;
+}
+
+static void stbtt__hheap_cleanup(stbtt__hheap *hh, void *userdata)
+{
+   stbtt__hheap_chunk *c = hh->head;
+   while (c) {
+      stbtt__hheap_chunk *n = c->next;
+      STBTT_free(c, userdata);
+      c = n;
+   }
+}
+
+typedef struct stbtt__edge {
+   float x0,y0, x1,y1;
+   int invert;
+} stbtt__edge;
+
+
+typedef struct stbtt__active_edge
+{
+   struct stbtt__active_edge *next;
+   #if STBTT_RASTERIZER_VERSION==1
+   int x,dx;
+   float ey;
+   int direction;
+   #elif STBTT_RASTERIZER_VERSION==2
+   float fx,fdx,fdy;
+   float direction;
+   float sy;
+   float ey;
+   #else
+   #error "Unrecognized value of STBTT_RASTERIZER_VERSION"
+   #endif
+} stbtt__active_edge;
+
+#if STBTT_RASTERIZER_VERSION == 1
+#define STBTT_FIXSHIFT   10
+#define STBTT_FIX        (1 << STBTT_FIXSHIFT)
+#define STBTT_FIXMASK    (STBTT_FIX-1)
+
+static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata)
+{
+   stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata);
+   float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0);
+   STBTT_assert(z != NULL);
+   if (!z) return z;
+   
+   // round dx down to avoid overshooting
+   if (dxdy < 0)
+      z->dx = -STBTT_ifloor(STBTT_FIX * -dxdy);
+   else
+      z->dx = STBTT_ifloor(STBTT_FIX * dxdy);
+
+   z->x = STBTT_ifloor(STBTT_FIX * e->x0 + z->dx * (start_point - e->y0)); // use z->dx so when we offset later it's by the same amount
+   z->x -= off_x * STBTT_FIX;
+
+   z->ey = e->y1;
+   z->next = 0;
+   z->direction = e->invert ? 1 : -1;
+   return z;
+}
+#elif STBTT_RASTERIZER_VERSION == 2
+static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata)
+{
+   stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata);
+   float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0);
+   STBTT_assert(z != NULL);
+   //STBTT_assert(e->y0 <= start_point);
+   if (!z) return z;
+   z->fdx = dxdy;
+   z->fdy = dxdy != 0.0f ? (1.0f/dxdy) : 0.0f;
+   z->fx = e->x0 + dxdy * (start_point - e->y0);
+   z->fx -= off_x;
+   z->direction = e->invert ? 1.0f : -1.0f;
+   z->sy = e->y0;
+   z->ey = e->y1;
+   z->next = 0;
+   return z;
+}
+#else
+#error "Unrecognized value of STBTT_RASTERIZER_VERSION"
+#endif
+
+#if STBTT_RASTERIZER_VERSION == 1
+// note: this routine clips fills that extend off the edges... ideally this
+// wouldn't happen, but it could happen if the truetype glyph bounding boxes
+// are wrong, or if the user supplies a too-small bitmap
+static void stbtt__fill_active_edges(unsigned char *scanline, int len, stbtt__active_edge *e, int max_weight)
+{
+   // non-zero winding fill
+   int x0=0, w=0;
+
+   while (e) {
+      if (w == 0) {
+         // if we're currently at zero, we need to record the edge start point
+         x0 = e->x; w += e->direction;
+      } else {
+         int x1 = e->x; w += e->direction;
+         // if we went to zero, we need to draw
+         if (w == 0) {
+            int i = x0 >> STBTT_FIXSHIFT;
+            int j = x1 >> STBTT_FIXSHIFT;
+
+            if (i < len && j >= 0) {
+               if (i == j) {
+                  // x0,x1 are the same pixel, so compute combined coverage
+                  scanline[i] = scanline[i] + (stbtt_uint8) ((x1 - x0) * max_weight >> STBTT_FIXSHIFT);
+               } else {
+                  if (i >= 0) // add antialiasing for x0
+                     scanline[i] = scanline[i] + (stbtt_uint8) (((STBTT_FIX - (x0 & STBTT_FIXMASK)) * max_weight) >> STBTT_FIXSHIFT);
+                  else
+                     i = -1; // clip
+
+                  if (j < len) // add antialiasing for x1
+                     scanline[j] = scanline[j] + (stbtt_uint8) (((x1 & STBTT_FIXMASK) * max_weight) >> STBTT_FIXSHIFT);
+                  else
+                     j = len; // clip
+
+                  for (++i; i < j; ++i) // fill pixels between x0 and x1
+                     scanline[i] = scanline[i] + (stbtt_uint8) max_weight;
+               }
+            }
+         }
+      }
+      
+      e = e->next;
+   }
+}
+
+static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata)
+{
+   stbtt__hheap hh = { 0, 0, 0 };
+   stbtt__active_edge *active = NULL;
+   int y,j=0;
+   int max_weight = (255 / vsubsample);  // weight per vertical scanline
+   int s; // vertical subsample index
+   unsigned char scanline_data[512], *scanline;
+
+   if (result->w > 512)
+      scanline = (unsigned char *) STBTT_malloc(result->w, userdata);
+   else
+      scanline = scanline_data;
+
+   y = off_y * vsubsample;
+   e[n].y0 = (off_y + result->h) * (float) vsubsample + 1;
+
+   while (j < result->h) {
+      STBTT_memset(scanline, 0, result->w);
+      for (s=0; s < vsubsample; ++s) {
+         // find center of pixel for this scanline
+         float scan_y = y + 0.5f;
+         stbtt__active_edge **step = &active;
+
+         // update all active edges;
+         // remove all active edges that terminate before the center of this scanline
+         while (*step) {
+            stbtt__active_edge * z = *step;
+            if (z->ey <= scan_y) {
+               *step = z->next; // delete from list
+               STBTT_assert(z->direction);
+               z->direction = 0;
+               stbtt__hheap_free(&hh, z);
+            } else {
+               z->x += z->dx; // advance to position for current scanline
+               step = &((*step)->next); // advance through list
+            }
+         }
+
+         // resort the list if needed
+         for(;;) {
+            int changed=0;
+            step = &active;
+            while (*step && (*step)->next) {
+               if ((*step)->x > (*step)->next->x) {
+                  stbtt__active_edge *t = *step;
+                  stbtt__active_edge *q = t->next;
+
+                  t->next = q->next;
+                  q->next = t;
+                  *step = q;
+                  changed = 1;
+               }
+               step = &(*step)->next;
+            }
+            if (!changed) break;
+         }
+
+         // insert all edges that start before the center of this scanline -- omit ones that also end on this scanline
+         while (e->y0 <= scan_y) {
+            if (e->y1 > scan_y) {
+               stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y, userdata);
+               if (z != NULL) {
+                  // find insertion point
+                  if (active == NULL)
+                     active = z;
+                  else if (z->x < active->x) {
+                     // insert at front
+                     z->next = active;
+                     active = z;
+                  } else {
+                     // find thing to insert AFTER
+                     stbtt__active_edge *p = active;
+                     while (p->next && p->next->x < z->x)
+                        p = p->next;
+                     // at this point, p->next->x is NOT < z->x
+                     z->next = p->next;
+                     p->next = z;
+                  }
+               }
+            }
+            ++e;
+         }
+
+         // now process all active edges in XOR fashion
+         if (active)
+            stbtt__fill_active_edges(scanline, result->w, active, max_weight);
+
+         ++y;
+      }
+      STBTT_memcpy(result->pixels + j * result->stride, scanline, result->w);
+      ++j;
+   }
+
+   stbtt__hheap_cleanup(&hh, userdata);
+
+   if (scanline != scanline_data)
+      STBTT_free(scanline, userdata);
+}
+
+#elif STBTT_RASTERIZER_VERSION == 2
+
+// the edge passed in here does not cross the vertical line at x or the vertical line at x+1
+// (i.e. it has already been clipped to those)
+static void stbtt__handle_clipped_edge(float *scanline, int x, stbtt__active_edge *e, float x0, float y0, float x1, float y1)
+{
+   if (y0 == y1) return;
+   STBTT_assert(y0 < y1);
+   STBTT_assert(e->sy <= e->ey);
+   if (y0 > e->ey) return;
+   if (y1 < e->sy) return;
+   if (y0 < e->sy) {
+      x0 += (x1-x0) * (e->sy - y0) / (y1-y0);
+      y0 = e->sy;
+   }
+   if (y1 > e->ey) {
+      x1 += (x1-x0) * (e->ey - y1) / (y1-y0);
+      y1 = e->ey;
+   }
+
+   if (x0 == x)
+      STBTT_assert(x1 <= x+1);
+   else if (x0 == x+1)
+      STBTT_assert(x1 >= x);
+   else if (x0 <= x)
+      STBTT_assert(x1 <= x);
+   else if (x0 >= x+1)
+      STBTT_assert(x1 >= x+1);
+   else
+      STBTT_assert(x1 >= x && x1 <= x+1);
+
+   if (x0 <= x && x1 <= x)
+      scanline[x] += e->direction * (y1-y0);
+   else if (x0 >= x+1 && x1 >= x+1)
+      ;
+   else {
+      STBTT_assert(x0 >= x && x0 <= x+1 && x1 >= x && x1 <= x+1);
+      scanline[x] += e->direction * (y1-y0) * (1-((x0-x)+(x1-x))/2); // coverage = 1 - average x position
+   }
+}
+
+static void stbtt__fill_active_edges_new(float *scanline, float *scanline_fill, int len, stbtt__active_edge *e, float y_top)
+{
+   float y_bottom = y_top+1;
+
+   while (e) {
+      // brute force every pixel
+
+      // compute intersection points with top & bottom
+      STBTT_assert(e->ey >= y_top);
+
+      if (e->fdx == 0) {
+         float x0 = e->fx;
+         if (x0 < len) {
+            if (x0 >= 0) {
+               stbtt__handle_clipped_edge(scanline,(int) x0,e, x0,y_top, x0,y_bottom);
+               stbtt__handle_clipped_edge(scanline_fill-1,(int) x0+1,e, x0,y_top, x0,y_bottom);
+            } else {
+               stbtt__handle_clipped_edge(scanline_fill-1,0,e, x0,y_top, x0,y_bottom);
+            }
+         }
+      } else {
+         float x0 = e->fx;
+         float dx = e->fdx;
+         float xb = x0 + dx;
+         float x_top, x_bottom;
+         float sy0,sy1;
+         float dy = e->fdy;
+         STBTT_assert(e->sy <= y_bottom && e->ey >= y_top);
+
+         // compute endpoints of line segment clipped to this scanline (if the
+         // line segment starts on this scanline. x0 is the intersection of the
+         // line with y_top, but that may be off the line segment.
+         if (e->sy > y_top) {
+            x_top = x0 + dx * (e->sy - y_top);
+            sy0 = e->sy;
+         } else {
+            x_top = x0;
+            sy0 = y_top;
+         }
+         if (e->ey < y_bottom) {
+            x_bottom = x0 + dx * (e->ey - y_top);
+            sy1 = e->ey;
+         } else {
+            x_bottom = xb;
+            sy1 = y_bottom;
+         }
+
+         if (x_top >= 0 && x_bottom >= 0 && x_top < len && x_bottom < len) {
+            // from here on, we don't have to range check x values
+
+            if ((int) x_top == (int) x_bottom) {
+               float height;
+               // simple case, only spans one pixel
+               int x = (int) x_top;
+               height = sy1 - sy0;
+               STBTT_assert(x >= 0 && x < len);
+               scanline[x] += e->direction * (1-((x_top - x) + (x_bottom-x))/2)  * height;
+               scanline_fill[x] += e->direction * height; // everything right of this pixel is filled
+            } else {
+               int x,x1,x2;
+               float y_crossing, step, sign, area;
+               // covers 2+ pixels
+               if (x_top > x_bottom) {
+                  // flip scanline vertically; signed area is the same
+                  float t;
+                  sy0 = y_bottom - (sy0 - y_top);
+                  sy1 = y_bottom - (sy1 - y_top);
+                  t = sy0, sy0 = sy1, sy1 = t;
+                  t = x_bottom, x_bottom = x_top, x_top = t;
+                  dx = -dx;
+                  dy = -dy;
+                  t = x0, x0 = xb, xb = t;
+               }
+
+               x1 = (int) x_top;
+               x2 = (int) x_bottom;
+               // compute intersection with y axis at x1+1
+               y_crossing = (x1+1 - x0) * dy + y_top;
+
+               sign = e->direction;
+               // area of the rectangle covered from y0..y_crossing
+               area = sign * (y_crossing-sy0);
+               // area of the triangle (x_top,y0), (x+1,y0), (x+1,y_crossing)
+               scanline[x1] += area * (1-((x_top - x1)+(x1+1-x1))/2);
+
+               step = sign * dy;
+               for (x = x1+1; x < x2; ++x) {
+                  scanline[x] += area + step/2;
+                  area += step;
+               }
+               y_crossing += dy * (x2 - (x1+1));
+
+               STBTT_assert(STBTT_fabs(area) <= 1.01f);
+
+               scanline[x2] += area + sign * (1-((x2-x2)+(x_bottom-x2))/2) * (sy1-y_crossing);
+
+               scanline_fill[x2] += sign * (sy1-sy0);
+            }
+         } else {
+            // if edge goes outside of box we're drawing, we require
+            // clipping logic. since this does not match the intended use
+            // of this library, we use a different, very slow brute
+            // force implementation
+            int x;
+            for (x=0; x < len; ++x) {
+               // cases:
+               //
+               // there can be up to two intersections with the pixel. any intersection
+               // with left or right edges can be handled by splitting into two (or three)
+               // regions. intersections with top & bottom do not necessitate case-wise logic.
+               //
+               // the old way of doing this found the intersections with the left & right edges,
+               // then used some simple logic to produce up to three segments in sorted order
+               // from top-to-bottom. however, this had a problem: if an x edge was epsilon
+               // across the x border, then the corresponding y position might not be distinct
+               // from the other y segment, and it might ignored as an empty segment. to avoid
+               // that, we need to explicitly produce segments based on x positions.
+
+               // rename variables to clearly-defined pairs
+               float y0 = y_top;
+               float x1 = (float) (x);
+               float x2 = (float) (x+1);
+               float x3 = xb;
+               float y3 = y_bottom;
+
+               // x = e->x + e->dx * (y-y_top)
+               // (y-y_top) = (x - e->x) / e->dx
+               // y = (x - e->x) / e->dx + y_top
+               float y1 = (x - x0) / dx + y_top;
+               float y2 = (x+1 - x0) / dx + y_top;
+
+               if (x0 < x1 && x3 > x2) {         // three segments descending down-right
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1);
+                  stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x2,y2);
+                  stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3);
+               } else if (x3 < x1 && x0 > x2) {  // three segments descending down-left
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2);
+                  stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x1,y1);
+                  stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3);
+               } else if (x0 < x1 && x3 > x1) {  // two segments across x, down-right
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1);
+                  stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3);
+               } else if (x3 < x1 && x0 > x1) {  // two segments across x, down-left
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1);
+                  stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3);
+               } else if (x0 < x2 && x3 > x2) {  // two segments across x+1, down-right
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2);
+                  stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3);
+               } else if (x3 < x2 && x0 > x2) {  // two segments across x+1, down-left
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2);
+                  stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3);
+               } else {  // one segment
+                  stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x3,y3);
+               }
+            }
+         }
+      }
+      e = e->next;
+   }
+}
+
+// directly AA rasterize edges w/o supersampling
+static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata)
+{
+   stbtt__hheap hh = { 0, 0, 0 };
+   stbtt__active_edge *active = NULL;
+   int y,j=0, i;
+   float scanline_data[129], *scanline, *scanline2;
+
+   STBTT__NOTUSED(vsubsample);
+
+   if (result->w > 64)
+      scanline = (float *) STBTT_malloc((result->w*2+1) * sizeof(float), userdata);
+   else
+      scanline = scanline_data;
+
+   scanline2 = scanline + result->w;
+
+   y = off_y;
+   e[n].y0 = (float) (off_y + result->h) + 1;
+
+   while (j < result->h) {
+      // find center of pixel for this scanline
+      float scan_y_top    = y + 0.0f;
+      float scan_y_bottom = y + 1.0f;
+      stbtt__active_edge **step = &active;
+
+      STBTT_memset(scanline , 0, result->w*sizeof(scanline[0]));
+      STBTT_memset(scanline2, 0, (result->w+1)*sizeof(scanline[0]));
+
+      // update all active edges;
+      // remove all active edges that terminate before the top of this scanline
+      while (*step) {
+         stbtt__active_edge * z = *step;
+         if (z->ey <= scan_y_top) {
+            *step = z->next; // delete from list
+            STBTT_assert(z->direction);
+            z->direction = 0;
+            stbtt__hheap_free(&hh, z);
+         } else {
+            step = &((*step)->next); // advance through list
+         }
+      }
+
+      // insert all edges that start before the bottom of this scanline
+      while (e->y0 <= scan_y_bottom) {
+         if (e->y0 != e->y1) {
+            stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y_top, userdata);
+            if (z != NULL) {
+               if (j == 0 && off_y != 0) {
+                  if (z->ey < scan_y_top) {
+                     // this can happen due to subpixel positioning and some kind of fp rounding error i think
+                     z->ey = scan_y_top;
+                  }
+               }
+               STBTT_assert(z->ey >= scan_y_top); // if we get really unlucky a tiny bit of an edge can be out of bounds
+               // insert at front
+               z->next = active;
+               active = z;
+            }
+         }
+         ++e;
+      }
+
+      // now process all active edges
+      if (active)
+         stbtt__fill_active_edges_new(scanline, scanline2+1, result->w, active, scan_y_top);
+
+      {
+         float sum = 0;
+         for (i=0; i < result->w; ++i) {
+            float k;
+            int m;
+            sum += scanline2[i];
+            k = scanline[i] + sum;
+            k = (float) STBTT_fabs(k)*255 + 0.5f;
+            m = (int) k;
+            if (m > 255) m = 255;
+            result->pixels[j*result->stride + i] = (unsigned char) m;
+         }
+      }
+      // advance all the edges
+      step = &active;
+      while (*step) {
+         stbtt__active_edge *z = *step;
+         z->fx += z->fdx; // advance to position for current scanline
+         step = &((*step)->next); // advance through list
+      }
+
+      ++y;
+      ++j;
+   }
+
+   stbtt__hheap_cleanup(&hh, userdata);
+
+   if (scanline != scanline_data)
+      STBTT_free(scanline, userdata);
+}
+#else
+#error "Unrecognized value of STBTT_RASTERIZER_VERSION"
+#endif
+
+#define STBTT__COMPARE(a,b)  ((a)->y0 < (b)->y0)
+
+static void stbtt__sort_edges_ins_sort(stbtt__edge *p, int n)
+{
+   int i,j;
+   for (i=1; i < n; ++i) {
+      stbtt__edge t = p[i], *a = &t;
+      j = i;
+      while (j > 0) {
+         stbtt__edge *b = &p[j-1];
+         int c = STBTT__COMPARE(a,b);
+         if (!c) break;
+         p[j] = p[j-1];
+         --j;
+      }
+      if (i != j)
+         p[j] = t;
+   }
+}
+
+static void stbtt__sort_edges_quicksort(stbtt__edge *p, int n)
+{
+   /* threshold for transitioning to insertion sort */
+   while (n > 12) {
+      stbtt__edge t;
+      int c01,c12,c,m,i,j;
+
+      /* compute median of three */
+      m = n >> 1;
+      c01 = STBTT__COMPARE(&p[0],&p[m]);
+      c12 = STBTT__COMPARE(&p[m],&p[n-1]);
+      /* if 0 >= mid >= end, or 0 < mid < end, then use mid */
+      if (c01 != c12) {
+         /* otherwise, we'll need to swap something else to middle */
+         int z;
+         c = STBTT__COMPARE(&p[0],&p[n-1]);
+         /* 0>mid && mid<n:  0>n => n; 0<n => 0 */
+         /* 0<mid && mid>n:  0>n => 0; 0<n => n */
+         z = (c == c12) ? 0 : n-1;
+         t = p[z];
+         p[z] = p[m];
+         p[m] = t;
+      }
+      /* now p[m] is the median-of-three */
+      /* swap it to the beginning so it won't move around */
+      t = p[0];
+      p[0] = p[m];
+      p[m] = t;
+
+      /* partition loop */
+      i=1;
+      j=n-1;
+      for(;;) {
+         /* handling of equality is crucial here */
+         /* for sentinels & efficiency with duplicates */
+         for (;;++i) {
+            if (!STBTT__COMPARE(&p[i], &p[0])) break;
+         }
+         for (;;--j) {
+            if (!STBTT__COMPARE(&p[0], &p[j])) break;
+         }
+         /* make sure we haven't crossed */
+         if (i >= j) break;
+         t = p[i];
+         p[i] = p[j];
+         p[j] = t;
+
+         ++i;
+         --j;
+      }
+      /* recurse on smaller side, iterate on larger */
+      if (j < (n-i)) {
+         stbtt__sort_edges_quicksort(p,j);
+         p = p+i;
+         n = n-i;
+      } else {
+         stbtt__sort_edges_quicksort(p+i, n-i);
+         n = j;
+      }
+   }
+}
+
+static void stbtt__sort_edges(stbtt__edge *p, int n)
+{
+   stbtt__sort_edges_quicksort(p, n);
+   stbtt__sort_edges_ins_sort(p, n);
+}
+
+typedef struct
+{
+   float x,y;
+} stbtt__point;
+
+static void stbtt__rasterize(stbtt__bitmap *result, stbtt__point *pts, int *wcount, int windings, float scale_x, float scale_y, float shift_x, float shift_y, int off_x, int off_y, int invert, void *userdata)
+{
+   float y_scale_inv = invert ? -scale_y : scale_y;
+   stbtt__edge *e;
+   int n,i,j,k,m;
+#if STBTT_RASTERIZER_VERSION == 1
+   int vsubsample = result->h < 8 ? 15 : 5;
+#elif STBTT_RASTERIZER_VERSION == 2
+   int vsubsample = 1;
+#else
+   #error "Unrecognized value of STBTT_RASTERIZER_VERSION"
+#endif
+   // vsubsample should divide 255 evenly; otherwise we won't reach full opacity
+
+   // now we have to blow out the windings into explicit edge lists
+   n = 0;
+   for (i=0; i < windings; ++i)
+      n += wcount[i];
+
+   e = (stbtt__edge *) STBTT_malloc(sizeof(*e) * (n+1), userdata); // add an extra one as a sentinel
+   if (e == 0) return;
+   n = 0;
+
+   m=0;
+   for (i=0; i < windings; ++i) {
+      stbtt__point *p = pts + m;
+      m += wcount[i];
+      j = wcount[i]-1;
+      for (k=0; k < wcount[i]; j=k++) {
+         int a=k,b=j;
+         // skip the edge if horizontal
+         if (p[j].y == p[k].y)
+            continue;
+         // add edge from j to k to the list
+         e[n].invert = 0;
+         if (invert ? p[j].y > p[k].y : p[j].y < p[k].y) {
+            e[n].invert = 1;
+            a=j,b=k;
+         }
+         e[n].x0 = p[a].x * scale_x + shift_x;
+         e[n].y0 = (p[a].y * y_scale_inv + shift_y) * vsubsample;
+         e[n].x1 = p[b].x * scale_x + shift_x;
+         e[n].y1 = (p[b].y * y_scale_inv + shift_y) * vsubsample;
+         ++n;
+      }
+   }
+
+   // now sort the edges by their highest point (should snap to integer, and then by x)
+   //STBTT_sort(e, n, sizeof(e[0]), stbtt__edge_compare);
+   stbtt__sort_edges(e, n);
+
+   // now, traverse the scanlines and find the intersections on each scanline, use xor winding rule
+   stbtt__rasterize_sorted_edges(result, e, n, vsubsample, off_x, off_y, userdata);
+
+   STBTT_free(e, userdata);
+}
+
+static void stbtt__add_point(stbtt__point *points, int n, float x, float y)
+{
+   if (!points) return; // during first pass, it's unallocated
+   points[n].x = x;
+   points[n].y = y;
+}
+
+// tessellate until threshold p is happy... @TODO warped to compensate for non-linear stretching
+static int stbtt__tesselate_curve(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float objspace_flatness_squared, int n)
+{
+   // midpoint
+   float mx = (x0 + 2*x1 + x2)/4;
+   float my = (y0 + 2*y1 + y2)/4;
+   // versus directly drawn line
+   float dx = (x0+x2)/2 - mx;
+   float dy = (y0+y2)/2 - my;
+   if (n > 16) // 65536 segments on one curve better be enough!
+      return 1;
+   if (dx*dx+dy*dy > objspace_flatness_squared) { // half-pixel error allowed... need to be smaller if AA
+      stbtt__tesselate_curve(points, num_points, x0,y0, (x0+x1)/2.0f,(y0+y1)/2.0f, mx,my, objspace_flatness_squared,n+1);
+      stbtt__tesselate_curve(points, num_points, mx,my, (x1+x2)/2.0f,(y1+y2)/2.0f, x2,y2, objspace_flatness_squared,n+1);
+   } else {
+      stbtt__add_point(points, *num_points,x2,y2);
+      *num_points = *num_points+1;
+   }
+   return 1;
+}
+
+static void stbtt__tesselate_cubic(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float x3, float y3, float objspace_flatness_squared, int n)
+{
+   // @TODO this "flatness" calculation is just made-up nonsense that seems to work well enough
+   float dx0 = x1-x0;
+   float dy0 = y1-y0;
+   float dx1 = x2-x1;
+   float dy1 = y2-y1;
+   float dx2 = x3-x2;
+   float dy2 = y3-y2;
+   float dx = x3-x0;
+   float dy = y3-y0;
+   float longlen = (float) (STBTT_sqrt(dx0*dx0+dy0*dy0)+STBTT_sqrt(dx1*dx1+dy1*dy1)+STBTT_sqrt(dx2*dx2+dy2*dy2));
+   float shortlen = (float) STBTT_sqrt(dx*dx+dy*dy);
+   float flatness_squared = longlen*longlen-shortlen*shortlen;
+
+   if (n > 16) // 65536 segments on one curve better be enough!
+      return;
+
+   if (flatness_squared > objspace_flatness_squared) {
+      float x01 = (x0+x1)/2;
+      float y01 = (y0+y1)/2;
+      float x12 = (x1+x2)/2;
+      float y12 = (y1+y2)/2;
+      float x23 = (x2+x3)/2;
+      float y23 = (y2+y3)/2;
+
+      float xa = (x01+x12)/2;
+      float ya = (y01+y12)/2;
+      float xb = (x12+x23)/2;
+      float yb = (y12+y23)/2;
+
+      float mx = (xa+xb)/2;
+      float my = (ya+yb)/2;
+
+      stbtt__tesselate_cubic(points, num_points, x0,y0, x01,y01, xa,ya, mx,my, objspace_flatness_squared,n+1);
+      stbtt__tesselate_cubic(points, num_points, mx,my, xb,yb, x23,y23, x3,y3, objspace_flatness_squared,n+1);
+   } else {
+      stbtt__add_point(points, *num_points,x3,y3);
+      *num_points = *num_points+1;
+   }
+}
+
+// returns number of contours
+static stbtt__point *stbtt_FlattenCurves(stbtt_vertex *vertices, int num_verts, float objspace_flatness, int **contour_lengths, int *num_contours, void *userdata)
+{
+   stbtt__point *points=0;
+   int num_points=0;
+
+   float objspace_flatness_squared = objspace_flatness * objspace_flatness;
+   int i,n=0,start=0, pass;
+
+   // count how many "moves" there are to get the contour count
+   for (i=0; i < num_verts; ++i)
+      if (vertices[i].type == STBTT_vmove)
+         ++n;
+
+   *num_contours = n;
+   if (n == 0) return 0;
+
+   *contour_lengths = (int *) STBTT_malloc(sizeof(**contour_lengths) * n, userdata);
+
+   if (*contour_lengths == 0) {
+      *num_contours = 0;
+      return 0;
+   }
+
+   // make two passes through the points so we don't need to realloc
+   for (pass=0; pass < 2; ++pass) {
+      float x=0,y=0;
+      if (pass == 1) {
+         points = (stbtt__point *) STBTT_malloc(num_points * sizeof(points[0]), userdata);
+         if (points == NULL) goto error;
+      }
+      num_points = 0;
+      n= -1;
+      for (i=0; i < num_verts; ++i) {
+         switch (vertices[i].type) {
+            case STBTT_vmove:
+               // start the next contour
+               if (n >= 0)
+                  (*contour_lengths)[n] = num_points - start;
+               ++n;
+               start = num_points;
+
+               x = vertices[i].x, y = vertices[i].y;
+               stbtt__add_point(points, num_points++, x,y);
+               break;
+            case STBTT_vline:
+               x = vertices[i].x, y = vertices[i].y;
+               stbtt__add_point(points, num_points++, x, y);
+               break;
+            case STBTT_vcurve:
+               stbtt__tesselate_curve(points, &num_points, x,y,
+                                        vertices[i].cx, vertices[i].cy,
+                                        vertices[i].x,  vertices[i].y,
+                                        objspace_flatness_squared, 0);
+               x = vertices[i].x, y = vertices[i].y;
+               break;
+            case STBTT_vcubic:
+               stbtt__tesselate_cubic(points, &num_points, x,y,
+                                        vertices[i].cx, vertices[i].cy,
+                                        vertices[i].cx1, vertices[i].cy1,
+                                        vertices[i].x,  vertices[i].y,
+                                        objspace_flatness_squared, 0);
+               x = vertices[i].x, y = vertices[i].y;
+               break;
+         }
+      }
+      (*contour_lengths)[n] = num_points - start;
+   }
+
+   return points;
+error:
+   STBTT_free(points, userdata);
+   STBTT_free(*contour_lengths, userdata);
+   *contour_lengths = 0;
+   *num_contours = 0;
+   return NULL;
+}
+
+STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, float flatness_in_pixels, stbtt_vertex *vertices, int num_verts, float scale_x, float scale_y, float shift_x, float shift_y, int x_off, int y_off, int invert, void *userdata)
+{
+   float scale            = scale_x > scale_y ? scale_y : scale_x;
+   int winding_count      = 0;
+   int *winding_lengths   = NULL;
+   stbtt__point *windings = stbtt_FlattenCurves(vertices, num_verts, flatness_in_pixels / scale, &winding_lengths, &winding_count, userdata);
+   if (windings) {
+      stbtt__rasterize(result, windings, winding_lengths, winding_count, scale_x, scale_y, shift_x, shift_y, x_off, y_off, invert, userdata);
+      STBTT_free(winding_lengths, userdata);
+      STBTT_free(windings, userdata);
+   }
+}
+
+STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata)
+{
+   STBTT_free(bitmap, userdata);
+}
+
+STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff)
+{
+   int ix0,iy0,ix1,iy1;
+   stbtt__bitmap gbm;
+   stbtt_vertex *vertices;   
+   int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices);
+
+   if (scale_x == 0) scale_x = scale_y;
+   if (scale_y == 0) {
+      if (scale_x == 0) {
+         STBTT_free(vertices, info->userdata);
+         return NULL;
+      }
+      scale_y = scale_x;
+   }
+
+   stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,&ix1,&iy1);
+
+   // now we get the size
+   gbm.w = (ix1 - ix0);
+   gbm.h = (iy1 - iy0);
+   gbm.pixels = NULL; // in case we error
+
+   if (width ) *width  = gbm.w;
+   if (height) *height = gbm.h;
+   if (xoff  ) *xoff   = ix0;
+   if (yoff  ) *yoff   = iy0;
+   
+   if (gbm.w && gbm.h) {
+      gbm.pixels = (unsigned char *) STBTT_malloc(gbm.w * gbm.h, info->userdata);
+      if (gbm.pixels) {
+         gbm.stride = gbm.w;
+
+         stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0, iy0, 1, info->userdata);
+      }
+   }
+   STBTT_free(vertices, info->userdata);
+   return gbm.pixels;
+}   
+
+STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff)
+{
+   return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y, 0.0f, 0.0f, glyph, width, height, xoff, yoff);
+}
+
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph)
+{
+   int ix0,iy0;
+   stbtt_vertex *vertices;
+   int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices);
+   stbtt__bitmap gbm;   
+
+   stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,0,0);
+   gbm.pixels = output;
+   gbm.w = out_w;
+   gbm.h = out_h;
+   gbm.stride = out_stride;
+
+   if (gbm.w && gbm.h)
+      stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0,iy0, 1, info->userdata);
+
+   STBTT_free(vertices, info->userdata);
+}
+
+STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph)
+{
+   stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, glyph);
+}
+
+STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff)
+{
+   return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y,shift_x,shift_y, stbtt_FindGlyphIndex(info,codepoint), width,height,xoff,yoff);
+}   
+
+STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint)
+{
+   stbtt_MakeGlyphBitmapSubpixelPrefilter(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, oversample_x, oversample_y, sub_x, sub_y, stbtt_FindGlyphIndex(info,codepoint));
+}
+
+STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint)
+{
+   stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, stbtt_FindGlyphIndex(info,codepoint));
+}
+
+STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff)
+{
+   return stbtt_GetCodepointBitmapSubpixel(info, scale_x, scale_y, 0.0f,0.0f, codepoint, width,height,xoff,yoff);
+}   
+
+STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint)
+{
+   stbtt_MakeCodepointBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, codepoint);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// bitmap baking
+//
+// This is SUPER-CRAPPY packing to keep source code small
+
+static int stbtt_BakeFontBitmap_internal(unsigned char *data, int offset,  // font location (use offset=0 for plain .ttf)
+                                float pixel_height,                     // height of font in pixels
+                                unsigned char *pixels, int pw, int ph,  // bitmap to be filled in
+                                int first_char, int num_chars,          // characters to bake
+                                stbtt_bakedchar *chardata)
+{
+   float scale;
+   int x,y,bottom_y, i;
+   stbtt_fontinfo f;
+   f.userdata = NULL;
+   if (!stbtt_InitFont(&f, data, offset))
+      return -1;
+   STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels
+   x=y=1;
+   bottom_y = 1;
+
+   scale = stbtt_ScaleForPixelHeight(&f, pixel_height);
+
+   for (i=0; i < num_chars; ++i) {
+      int advance, lsb, x0,y0,x1,y1,gw,gh;
+      int g = stbtt_FindGlyphIndex(&f, first_char + i);
+      stbtt_GetGlyphHMetrics(&f, g, &advance, &lsb);
+      stbtt_GetGlyphBitmapBox(&f, g, scale,scale, &x0,&y0,&x1,&y1);
+      gw = x1-x0;
+      gh = y1-y0;
+      if (x + gw + 1 >= pw)
+         y = bottom_y, x = 1; // advance to next row
+      if (y + gh + 1 >= ph) // check if it fits vertically AFTER potentially moving to next row
+         return -i;
+      STBTT_assert(x+gw < pw);
+      STBTT_assert(y+gh < ph);
+      stbtt_MakeGlyphBitmap(&f, pixels+x+y*pw, gw,gh,pw, scale,scale, g);
+      chardata[i].x0 = (stbtt_int16) x;
+      chardata[i].y0 = (stbtt_int16) y;
+      chardata[i].x1 = (stbtt_int16) (x + gw);
+      chardata[i].y1 = (stbtt_int16) (y + gh);
+      chardata[i].xadvance = scale * advance;
+      chardata[i].xoff     = (float) x0;
+      chardata[i].yoff     = (float) y0;
+      x = x + gw + 1;
+      if (y+gh+1 > bottom_y)
+         bottom_y = y+gh+1;
+   }
+   return bottom_y;
+}
+
+STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int opengl_fillrule)
+{
+   float d3d_bias = opengl_fillrule ? 0 : -0.5f;
+   float ipw = 1.0f / pw, iph = 1.0f / ph;
+   const stbtt_bakedchar *b = chardata + char_index;
+   int round_x = STBTT_ifloor((*xpos + b->xoff) + 0.5f);
+   int round_y = STBTT_ifloor((*ypos + b->yoff) + 0.5f);
+
+   q->x0 = round_x + d3d_bias;
+   q->y0 = round_y + d3d_bias;
+   q->x1 = round_x + b->x1 - b->x0 + d3d_bias;
+   q->y1 = round_y + b->y1 - b->y0 + d3d_bias;
+
+   q->s0 = b->x0 * ipw;
+   q->t0 = b->y0 * iph;
+   q->s1 = b->x1 * ipw;
+   q->t1 = b->y1 * iph;
+
+   *xpos += b->xadvance;
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// rectangle packing replacement routines if you don't have stb_rect_pack.h
+//
+
+#ifndef STB_RECT_PACK_VERSION
+
+typedef int stbrp_coord;
+
+////////////////////////////////////////////////////////////////////////////////////
+//                                                                                //
+//                                                                                //
+// COMPILER WARNING ?!?!?                                                         //
+//                                                                                //
+//                                                                                //
+// if you get a compile warning due to these symbols being defined more than      //
+// once, move #include "stb_rect_pack.h" before #include "stb_truetype.h"         //
+//                                                                                //
+////////////////////////////////////////////////////////////////////////////////////
+
+typedef struct
+{
+   int width,height;
+   int x,y,bottom_y;
+} stbrp_context;
+
+typedef struct
+{
+   unsigned char x;
+} stbrp_node;
+
+struct stbrp_rect
+{
+   stbrp_coord x,y;
+   int id,w,h,was_packed;
+};
+
+static void stbrp_init_target(stbrp_context *con, int pw, int ph, stbrp_node *nodes, int num_nodes)
+{
+   con->width  = pw;
+   con->height = ph;
+   con->x = 0;
+   con->y = 0;
+   con->bottom_y = 0;
+   STBTT__NOTUSED(nodes);
+   STBTT__NOTUSED(num_nodes);   
+}
+
+static void stbrp_pack_rects(stbrp_context *con, stbrp_rect *rects, int num_rects)
+{
+   int i;
+   for (i=0; i < num_rects; ++i) {
+      if (con->x + rects[i].w > con->width) {
+         con->x = 0;
+         con->y = con->bottom_y;
+      }
+      if (con->y + rects[i].h > con->height)
+         break;
+      rects[i].x = con->x;
+      rects[i].y = con->y;
+      rects[i].was_packed = 1;
+      con->x += rects[i].w;
+      if (con->y + rects[i].h > con->bottom_y)
+         con->bottom_y = con->y + rects[i].h;
+   }
+   for (   ; i < num_rects; ++i)
+      rects[i].was_packed = 0;
+}
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// bitmap baking
+//
+// This is SUPER-AWESOME (tm Ryan Gordon) packing using stb_rect_pack.h. If
+// stb_rect_pack.h isn't available, it uses the BakeFontBitmap strategy.
+
+STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int pw, int ph, int stride_in_bytes, int padding, void *alloc_context)
+{
+   stbrp_context *context = (stbrp_context *) STBTT_malloc(sizeof(*context)            ,alloc_context);
+   int            num_nodes = pw - padding;
+   stbrp_node    *nodes   = (stbrp_node    *) STBTT_malloc(sizeof(*nodes  ) * num_nodes,alloc_context);
+
+   if (context == NULL || nodes == NULL) {
+      if (context != NULL) STBTT_free(context, alloc_context);
+      if (nodes   != NULL) STBTT_free(nodes  , alloc_context);
+      return 0;
+   }
+
+   spc->user_allocator_context = alloc_context;
+   spc->width = pw;
+   spc->height = ph;
+   spc->pixels = pixels;
+   spc->pack_info = context;
+   spc->nodes = nodes;
+   spc->padding = padding;
+   spc->stride_in_bytes = stride_in_bytes != 0 ? stride_in_bytes : pw;
+   spc->h_oversample = 1;
+   spc->v_oversample = 1;
+   spc->skip_missing = 0;
+
+   stbrp_init_target(context, pw-padding, ph-padding, nodes, num_nodes);
+
+   if (pixels)
+      STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels
+
+   return 1;
+}
+
+STBTT_DEF void stbtt_PackEnd  (stbtt_pack_context *spc)
+{
+   STBTT_free(spc->nodes    , spc->user_allocator_context);
+   STBTT_free(spc->pack_info, spc->user_allocator_context);
+}
+
+STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample)
+{
+   STBTT_assert(h_oversample <= STBTT_MAX_OVERSAMPLE);
+   STBTT_assert(v_oversample <= STBTT_MAX_OVERSAMPLE);
+   if (h_oversample <= STBTT_MAX_OVERSAMPLE)
+      spc->h_oversample = h_oversample;
+   if (v_oversample <= STBTT_MAX_OVERSAMPLE)
+      spc->v_oversample = v_oversample;
+}
+
+STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip)
+{
+   spc->skip_missing = skip;
+}
+
+#define STBTT__OVER_MASK  (STBTT_MAX_OVERSAMPLE-1)
+
+static void stbtt__h_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width)
+{
+   unsigned char buffer[STBTT_MAX_OVERSAMPLE];
+   int safe_w = w - kernel_width;
+   int j;
+   STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze
+   for (j=0; j < h; ++j) {
+      int i;
+      unsigned int total;
+      STBTT_memset(buffer, 0, kernel_width);
+
+      total = 0;
+
+      // make kernel_width a constant in common cases so compiler can optimize out the divide
+      switch (kernel_width) {
+         case 2:
+            for (i=0; i <= safe_w; ++i) {
+               total += pixels[i] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i];
+               pixels[i] = (unsigned char) (total / 2);
+            }
+            break;
+         case 3:
+            for (i=0; i <= safe_w; ++i) {
+               total += pixels[i] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i];
+               pixels[i] = (unsigned char) (total / 3);
+            }
+            break;
+         case 4:
+            for (i=0; i <= safe_w; ++i) {
+               total += pixels[i] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i];
+               pixels[i] = (unsigned char) (total / 4);
+            }
+            break;
+         case 5:
+            for (i=0; i <= safe_w; ++i) {
+               total += pixels[i] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i];
+               pixels[i] = (unsigned char) (total / 5);
+            }
+            break;
+         default:
+            for (i=0; i <= safe_w; ++i) {
+               total += pixels[i] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i];
+               pixels[i] = (unsigned char) (total / kernel_width);
+            }
+            break;
+      }
+
+      for (; i < w; ++i) {
+         STBTT_assert(pixels[i] == 0);
+         total -= buffer[i & STBTT__OVER_MASK];
+         pixels[i] = (unsigned char) (total / kernel_width);
+      }
+
+      pixels += stride_in_bytes;
+   }
+}
+
+static void stbtt__v_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width)
+{
+   unsigned char buffer[STBTT_MAX_OVERSAMPLE];
+   int safe_h = h - kernel_width;
+   int j;
+   STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze
+   for (j=0; j < w; ++j) {
+      int i;
+      unsigned int total;
+      STBTT_memset(buffer, 0, kernel_width);
+
+      total = 0;
+
+      // make kernel_width a constant in common cases so compiler can optimize out the divide
+      switch (kernel_width) {
+         case 2:
+            for (i=0; i <= safe_h; ++i) {
+               total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes];
+               pixels[i*stride_in_bytes] = (unsigned char) (total / 2);
+            }
+            break;
+         case 3:
+            for (i=0; i <= safe_h; ++i) {
+               total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes];
+               pixels[i*stride_in_bytes] = (unsigned char) (total / 3);
+            }
+            break;
+         case 4:
+            for (i=0; i <= safe_h; ++i) {
+               total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes];
+               pixels[i*stride_in_bytes] = (unsigned char) (total / 4);
+            }
+            break;
+         case 5:
+            for (i=0; i <= safe_h; ++i) {
+               total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes];
+               pixels[i*stride_in_bytes] = (unsigned char) (total / 5);
+            }
+            break;
+         default:
+            for (i=0; i <= safe_h; ++i) {
+               total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK];
+               buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes];
+               pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width);
+            }
+            break;
+      }
+
+      for (; i < h; ++i) {
+         STBTT_assert(pixels[i*stride_in_bytes] == 0);
+         total -= buffer[i & STBTT__OVER_MASK];
+         pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width);
+      }
+
+      pixels += 1;
+   }
+}
+
+static float stbtt__oversample_shift(int oversample)
+{
+   if (!oversample)
+      return 0.0f;
+
+   // The prefilter is a box filter of width "oversample",
+   // which shifts phase by (oversample - 1)/2 pixels in
+   // oversampled space. We want to shift in the opposite
+   // direction to counter this.
+   return (float)-(oversample - 1) / (2.0f * (float)oversample);
+}
+
+// rects array must be big enough to accommodate all characters in the given ranges
+STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects)
+{
+   int i,j,k;
+
+   k=0;
+   for (i=0; i < num_ranges; ++i) {
+      float fh = ranges[i].font_size;
+      float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh);
+      ranges[i].h_oversample = (unsigned char) spc->h_oversample;
+      ranges[i].v_oversample = (unsigned char) spc->v_oversample;
+      for (j=0; j < ranges[i].num_chars; ++j) {
+         int x0,y0,x1,y1;
+         int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j];
+         int glyph = stbtt_FindGlyphIndex(info, codepoint);
+         if (glyph == 0 && spc->skip_missing) {
+            rects[k].w = rects[k].h = 0;
+         } else {
+            stbtt_GetGlyphBitmapBoxSubpixel(info,glyph,
+                                            scale * spc->h_oversample,
+                                            scale * spc->v_oversample,
+                                            0,0,
+                                            &x0,&y0,&x1,&y1);
+            rects[k].w = (stbrp_coord) (x1-x0 + spc->padding + spc->h_oversample-1);
+            rects[k].h = (stbrp_coord) (y1-y0 + spc->padding + spc->v_oversample-1);
+         }
+         ++k;
+      }
+   }
+
+   return k;
+}
+
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int prefilter_x, int prefilter_y, float *sub_x, float *sub_y, int glyph)
+{
+   stbtt_MakeGlyphBitmapSubpixel(info,
+                                 output,
+                                 out_w - (prefilter_x - 1),
+                                 out_h - (prefilter_y - 1),
+                                 out_stride,
+                                 scale_x,
+                                 scale_y,
+                                 shift_x,
+                                 shift_y,
+                                 glyph);
+
+   if (prefilter_x > 1)
+      stbtt__h_prefilter(output, out_w, out_h, out_stride, prefilter_x);
+
+   if (prefilter_y > 1)
+      stbtt__v_prefilter(output, out_w, out_h, out_stride, prefilter_y);
+
+   *sub_x = stbtt__oversample_shift(prefilter_x);
+   *sub_y = stbtt__oversample_shift(prefilter_y);
+}
+
+// rects array must be big enough to accommodate all characters in the given ranges
+STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects)
+{
+   int i,j,k, return_value = 1;
+
+   // save current values
+   int old_h_over = spc->h_oversample;
+   int old_v_over = spc->v_oversample;
+
+   k = 0;
+   for (i=0; i < num_ranges; ++i) {
+      float fh = ranges[i].font_size;
+      float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh);
+      float recip_h,recip_v,sub_x,sub_y;
+      spc->h_oversample = ranges[i].h_oversample;
+      spc->v_oversample = ranges[i].v_oversample;
+      recip_h = 1.0f / spc->h_oversample;
+      recip_v = 1.0f / spc->v_oversample;
+      sub_x = stbtt__oversample_shift(spc->h_oversample);
+      sub_y = stbtt__oversample_shift(spc->v_oversample);
+      for (j=0; j < ranges[i].num_chars; ++j) {
+         stbrp_rect *r = &rects[k];
+         if (r->was_packed && r->w != 0 && r->h != 0) {
+            stbtt_packedchar *bc = &ranges[i].chardata_for_range[j];
+            int advance, lsb, x0,y0,x1,y1;
+            int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j];
+            int glyph = stbtt_FindGlyphIndex(info, codepoint);
+            stbrp_coord pad = (stbrp_coord) spc->padding;
+
+            // pad on left and top
+            r->x += pad;
+            r->y += pad;
+            r->w -= pad;
+            r->h -= pad;
+            stbtt_GetGlyphHMetrics(info, glyph, &advance, &lsb);
+            stbtt_GetGlyphBitmapBox(info, glyph,
+                                    scale * spc->h_oversample,
+                                    scale * spc->v_oversample,
+                                    &x0,&y0,&x1,&y1);
+            stbtt_MakeGlyphBitmapSubpixel(info,
+                                          spc->pixels + r->x + r->y*spc->stride_in_bytes,
+                                          r->w - spc->h_oversample+1,
+                                          r->h - spc->v_oversample+1,
+                                          spc->stride_in_bytes,
+                                          scale * spc->h_oversample,
+                                          scale * spc->v_oversample,
+                                          0,0,
+                                          glyph);
+
+            if (spc->h_oversample > 1)
+               stbtt__h_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes,
+                                  r->w, r->h, spc->stride_in_bytes,
+                                  spc->h_oversample);
+
+            if (spc->v_oversample > 1)
+               stbtt__v_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes,
+                                  r->w, r->h, spc->stride_in_bytes,
+                                  spc->v_oversample);
+
+            bc->x0       = (stbtt_int16)  r->x;
+            bc->y0       = (stbtt_int16)  r->y;
+            bc->x1       = (stbtt_int16) (r->x + r->w);
+            bc->y1       = (stbtt_int16) (r->y + r->h);
+            bc->xadvance =                scale * advance;
+            bc->xoff     =       (float)  x0 * recip_h + sub_x;
+            bc->yoff     =       (float)  y0 * recip_v + sub_y;
+            bc->xoff2    =                (x0 + r->w) * recip_h + sub_x;
+            bc->yoff2    =                (y0 + r->h) * recip_v + sub_y;
+         } else {
+            return_value = 0; // if any fail, report failure
+         }
+
+         ++k;
+      }
+   }
+
+   // restore original values
+   spc->h_oversample = old_h_over;
+   spc->v_oversample = old_v_over;
+
+   return return_value;
+}
+
+STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects)
+{
+   stbrp_pack_rects((stbrp_context *) spc->pack_info, rects, num_rects);
+}
+
+STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges)
+{
+   stbtt_fontinfo info;
+   int i,j,n, return_value = 1;
+   //stbrp_context *context = (stbrp_context *) spc->pack_info;
+   stbrp_rect    *rects;
+
+   // flag all characters as NOT packed
+   for (i=0; i < num_ranges; ++i)
+      for (j=0; j < ranges[i].num_chars; ++j)
+         ranges[i].chardata_for_range[j].x0 =
+         ranges[i].chardata_for_range[j].y0 =
+         ranges[i].chardata_for_range[j].x1 =
+         ranges[i].chardata_for_range[j].y1 = 0;
+
+   n = 0;
+   for (i=0; i < num_ranges; ++i)
+      n += ranges[i].num_chars;
+         
+   rects = (stbrp_rect *) STBTT_malloc(sizeof(*rects) * n, spc->user_allocator_context);
+   if (rects == NULL)
+      return 0;
+
+   info.userdata = spc->user_allocator_context;
+   stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata,font_index));
+
+   n = stbtt_PackFontRangesGatherRects(spc, &info, ranges, num_ranges, rects);
+
+   stbtt_PackFontRangesPackRects(spc, rects, n);
+  
+   return_value = stbtt_PackFontRangesRenderIntoRects(spc, &info, ranges, num_ranges, rects);
+
+   STBTT_free(rects, spc->user_allocator_context);
+   return return_value;
+}
+
+STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size,
+            int first_unicode_codepoint_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range)
+{
+   stbtt_pack_range range;
+   range.first_unicode_codepoint_in_range = first_unicode_codepoint_in_range;
+   range.array_of_unicode_codepoints = NULL;
+   range.num_chars                   = num_chars_in_range;
+   range.chardata_for_range          = chardata_for_range;
+   range.font_size                   = font_size;
+   return stbtt_PackFontRanges(spc, fontdata, font_index, &range, 1);
+}
+
+STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap)
+{
+   int i_ascent, i_descent, i_lineGap;
+   float scale;
+   stbtt_fontinfo info;
+   stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata, index));
+   scale = size > 0 ? stbtt_ScaleForPixelHeight(&info, size) : stbtt_ScaleForMappingEmToPixels(&info, -size);
+   stbtt_GetFontVMetrics(&info, &i_ascent, &i_descent, &i_lineGap);
+   *ascent  = (float) i_ascent  * scale;
+   *descent = (float) i_descent * scale;
+   *lineGap = (float) i_lineGap * scale;
+}
+
+STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int align_to_integer)
+{
+   float ipw = 1.0f / pw, iph = 1.0f / ph;
+   const stbtt_packedchar *b = chardata + char_index;
+
+   if (align_to_integer) {
+      float x = (float) STBTT_ifloor((*xpos + b->xoff) + 0.5f);
+      float y = (float) STBTT_ifloor((*ypos + b->yoff) + 0.5f);
+      q->x0 = x;
+      q->y0 = y;
+      q->x1 = x + b->xoff2 - b->xoff;
+      q->y1 = y + b->yoff2 - b->yoff;
+   } else {
+      q->x0 = *xpos + b->xoff;
+      q->y0 = *ypos + b->yoff;
+      q->x1 = *xpos + b->xoff2;
+      q->y1 = *ypos + b->yoff2;
+   }
+
+   q->s0 = b->x0 * ipw;
+   q->t0 = b->y0 * iph;
+   q->s1 = b->x1 * ipw;
+   q->t1 = b->y1 * iph;
+
+   *xpos += b->xadvance;
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// sdf computation
+//
+
+#define STBTT_min(a,b)  ((a) < (b) ? (a) : (b))
+#define STBTT_max(a,b)  ((a) < (b) ? (b) : (a))
+
+static int stbtt__ray_intersect_bezier(float orig[2], float ray[2], float q0[2], float q1[2], float q2[2], float hits[2][2])
+{
+   float q0perp = q0[1]*ray[0] - q0[0]*ray[1];
+   float q1perp = q1[1]*ray[0] - q1[0]*ray[1];
+   float q2perp = q2[1]*ray[0] - q2[0]*ray[1];
+   float roperp = orig[1]*ray[0] - orig[0]*ray[1];
+
+   float a = q0perp - 2*q1perp + q2perp;
+   float b = q1perp - q0perp;
+   float c = q0perp - roperp;
+
+   float s0 = 0., s1 = 0.;
+   int num_s = 0;
+
+   if (a != 0.0) {
+      float discr = b*b - a*c;
+      if (discr > 0.0) {
+         float rcpna = -1 / a;
+         float d = (float) STBTT_sqrt(discr);
+         s0 = (b+d) * rcpna;
+         s1 = (b-d) * rcpna;
+         if (s0 >= 0.0 && s0 <= 1.0)
+            num_s = 1;
+         if (d > 0.0 && s1 >= 0.0 && s1 <= 1.0) {
+            if (num_s == 0) s0 = s1;
+            ++num_s;
+         }
+      }
+   } else {
+      // 2*b*s + c = 0
+      // s = -c / (2*b)
+      s0 = c / (-2 * b);
+      if (s0 >= 0.0 && s0 <= 1.0)
+         num_s = 1;
+   }
+
+   if (num_s == 0)
+      return 0;
+   else {
+      float rcp_len2 = 1 / (ray[0]*ray[0] + ray[1]*ray[1]);
+      float rayn_x = ray[0] * rcp_len2, rayn_y = ray[1] * rcp_len2;
+
+      float q0d =   q0[0]*rayn_x +   q0[1]*rayn_y;
+      float q1d =   q1[0]*rayn_x +   q1[1]*rayn_y;
+      float q2d =   q2[0]*rayn_x +   q2[1]*rayn_y;
+      float rod = orig[0]*rayn_x + orig[1]*rayn_y;
+
+      float q10d = q1d - q0d;
+      float q20d = q2d - q0d;
+      float q0rd = q0d - rod;
+
+      hits[0][0] = q0rd + s0*(2.0f - 2.0f*s0)*q10d + s0*s0*q20d;
+      hits[0][1] = a*s0+b;
+
+      if (num_s > 1) {
+         hits[1][0] = q0rd + s1*(2.0f - 2.0f*s1)*q10d + s1*s1*q20d;
+         hits[1][1] = a*s1+b;
+         return 2;
+      } else {
+         return 1;
+      }
+   }
+}
+
+static int equal(float *a, float *b)
+{
+   return (a[0] == b[0] && a[1] == b[1]);
+}
+
+static int stbtt__compute_crossings_x(float x, float y, int nverts, stbtt_vertex *verts)
+{
+   int i;
+   float orig[2], ray[2] = { 1, 0 };
+   float y_frac;
+   int winding = 0;
+
+   orig[0] = x;
+   orig[1] = y;
+
+   // make sure y never passes through a vertex of the shape
+   y_frac = (float) STBTT_fmod(y, 1.0f);
+   if (y_frac < 0.01f)
+      y += 0.01f;
+   else if (y_frac > 0.99f)
+      y -= 0.01f;
+   orig[1] = y;
+
+   // test a ray from (-infinity,y) to (x,y)
+   for (i=0; i < nverts; ++i) {
+      if (verts[i].type == STBTT_vline) {
+         int x0 = (int) verts[i-1].x, y0 = (int) verts[i-1].y;
+         int x1 = (int) verts[i  ].x, y1 = (int) verts[i  ].y;
+         if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) {
+            float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0;
+            if (x_inter < x)  
+               winding += (y0 < y1) ? 1 : -1;
+         }
+      }
+      if (verts[i].type == STBTT_vcurve) {
+         int x0 = (int) verts[i-1].x , y0 = (int) verts[i-1].y ;
+         int x1 = (int) verts[i  ].cx, y1 = (int) verts[i  ].cy;
+         int x2 = (int) verts[i  ].x , y2 = (int) verts[i  ].y ;
+         int ax = STBTT_min(x0,STBTT_min(x1,x2)), ay = STBTT_min(y0,STBTT_min(y1,y2));
+         int by = STBTT_max(y0,STBTT_max(y1,y2));
+         if (y > ay && y < by && x > ax) {
+            float q0[2],q1[2],q2[2];
+            float hits[2][2];
+            q0[0] = (float)x0;
+            q0[1] = (float)y0;
+            q1[0] = (float)x1;
+            q1[1] = (float)y1;
+            q2[0] = (float)x2;
+            q2[1] = (float)y2;
+            if (equal(q0,q1) || equal(q1,q2)) {
+               x0 = (int)verts[i-1].x;
+               y0 = (int)verts[i-1].y;
+               x1 = (int)verts[i  ].x;
+               y1 = (int)verts[i  ].y;
+               if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) {
+                  float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0;
+                  if (x_inter < x)  
+                     winding += (y0 < y1) ? 1 : -1;
+               }
+            } else {
+               int num_hits = stbtt__ray_intersect_bezier(orig, ray, q0, q1, q2, hits);
+               if (num_hits >= 1)
+                  if (hits[0][0] < 0)
+                     winding += (hits[0][1] < 0 ? -1 : 1);
+               if (num_hits >= 2)
+                  if (hits[1][0] < 0)
+                     winding += (hits[1][1] < 0 ? -1 : 1);
+            }
+         } 
+      }
+   }
+   return winding;
+}
+
+static float stbtt__cuberoot( float x )
+{
+   if (x<0)
+      return -(float) STBTT_pow(-x,1.0f/3.0f);
+   else
+      return  (float) STBTT_pow( x,1.0f/3.0f);
+}
+
+// x^3 + c*x^2 + b*x + a = 0
+static int stbtt__solve_cubic(float a, float b, float c, float* r)
+{
+	float s = -a / 3;
+	float p = b - a*a / 3;
+	float q = a * (2*a*a - 9*b) / 27 + c;
+   float p3 = p*p*p;
+	float d = q*q + 4*p3 / 27;
+	if (d >= 0) {
+		float z = (float) STBTT_sqrt(d);
+		float u = (-q + z) / 2;
+		float v = (-q - z) / 2;
+		u = stbtt__cuberoot(u);
+		v = stbtt__cuberoot(v);
+		r[0] = s + u + v;
+		return 1;
+	} else {
+	   float u = (float) STBTT_sqrt(-p/3);
+	   float v = (float) STBTT_acos(-STBTT_sqrt(-27/p3) * q / 2) / 3; // p3 must be negative, since d is negative
+	   float m = (float) STBTT_cos(v);
+      float n = (float) STBTT_cos(v-3.141592/2)*1.732050808f;
+	   r[0] = s + u * 2 * m;
+	   r[1] = s - u * (m + n);
+	   r[2] = s - u * (m - n);
+
+      //STBTT_assert( STBTT_fabs(((r[0]+a)*r[0]+b)*r[0]+c) < 0.05f);  // these asserts may not be safe at all scales, though they're in bezier t parameter units so maybe?
+      //STBTT_assert( STBTT_fabs(((r[1]+a)*r[1]+b)*r[1]+c) < 0.05f);
+      //STBTT_assert( STBTT_fabs(((r[2]+a)*r[2]+b)*r[2]+c) < 0.05f);
+   	return 3;
+   }
+}
+
+STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff)
+{
+   float scale_x = scale, scale_y = scale;
+   int ix0,iy0,ix1,iy1;
+   int w,h;
+   unsigned char *data;
+
+   // if one scale is 0, use same scale for both
+   if (scale_x == 0) scale_x = scale_y;
+   if (scale_y == 0) {
+      if (scale_x == 0) return NULL;  // if both scales are 0, return NULL
+      scale_y = scale_x;
+   }
+
+   stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale, scale, 0.0f,0.0f, &ix0,&iy0,&ix1,&iy1);
+
+   // if empty, return NULL
+   if (ix0 == ix1 || iy0 == iy1)
+      return NULL;
+
+   ix0 -= padding;
+   iy0 -= padding;
+   ix1 += padding;
+   iy1 += padding;
+
+   w = (ix1 - ix0);
+   h = (iy1 - iy0);
+
+   if (width ) *width  = w;
+   if (height) *height = h;
+   if (xoff  ) *xoff   = ix0;
+   if (yoff  ) *yoff   = iy0;
+
+   // invert for y-downwards bitmaps
+   scale_y = -scale_y;
+      
+   {
+      int x,y,i,j;
+      float *precompute;
+      stbtt_vertex *verts;
+      int num_verts = stbtt_GetGlyphShape(info, glyph, &verts);
+      data = (unsigned char *) STBTT_malloc(w * h, info->userdata);
+      precompute = (float *) STBTT_malloc(num_verts * sizeof(float), info->userdata);
+
+      for (i=0,j=num_verts-1; i < num_verts; j=i++) {
+         if (verts[i].type == STBTT_vline) {
+            float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y;
+            float x1 = verts[j].x*scale_x, y1 = verts[j].y*scale_y;
+            float dist = (float) STBTT_sqrt((x1-x0)*(x1-x0) + (y1-y0)*(y1-y0));
+            precompute[i] = (dist == 0) ? 0.0f : 1.0f / dist;
+         } else if (verts[i].type == STBTT_vcurve) {
+            float x2 = verts[j].x *scale_x, y2 = verts[j].y *scale_y;
+            float x1 = verts[i].cx*scale_x, y1 = verts[i].cy*scale_y;
+            float x0 = verts[i].x *scale_x, y0 = verts[i].y *scale_y;
+            float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2;
+            float len2 = bx*bx + by*by;
+            if (len2 != 0.0f)
+               precompute[i] = 1.0f / (bx*bx + by*by);
+            else
+               precompute[i] = 0.0f;
+         } else
+            precompute[i] = 0.0f;
+      }
+
+      for (y=iy0; y < iy1; ++y) {
+         for (x=ix0; x < ix1; ++x) {
+            float val;
+            float min_dist = 999999.0f;
+            float sx = (float) x + 0.5f;
+            float sy = (float) y + 0.5f;
+            float x_gspace = (sx / scale_x);
+            float y_gspace = (sy / scale_y);
+
+            int winding = stbtt__compute_crossings_x(x_gspace, y_gspace, num_verts, verts); // @OPTIMIZE: this could just be a rasterization, but needs to be line vs. non-tesselated curves so a new path
+
+            for (i=0; i < num_verts; ++i) {
+               float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y;
+
+               // check against every point here rather than inside line/curve primitives -- @TODO: wrong if multiple 'moves' in a row produce a garbage point, and given culling, probably more efficient to do within line/curve
+               float dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy);
+               if (dist2 < min_dist*min_dist)
+                  min_dist = (float) STBTT_sqrt(dist2);
+
+               if (verts[i].type == STBTT_vline) {
+                  float x1 = verts[i-1].x*scale_x, y1 = verts[i-1].y*scale_y;
+
+                  // coarse culling against bbox
+                  //if (sx > STBTT_min(x0,x1)-min_dist && sx < STBTT_max(x0,x1)+min_dist &&
+                  //    sy > STBTT_min(y0,y1)-min_dist && sy < STBTT_max(y0,y1)+min_dist)
+                  float dist = (float) STBTT_fabs((x1-x0)*(y0-sy) - (y1-y0)*(x0-sx)) * precompute[i];
+                  STBTT_assert(i != 0);
+                  if (dist < min_dist) {
+                     // check position along line
+                     // x' = x0 + t*(x1-x0), y' = y0 + t*(y1-y0)
+                     // minimize (x'-sx)*(x'-sx)+(y'-sy)*(y'-sy)
+                     float dx = x1-x0, dy = y1-y0;
+                     float px = x0-sx, py = y0-sy;
+                     // minimize (px+t*dx)^2 + (py+t*dy)^2 = px*px + 2*px*dx*t + t^2*dx*dx + py*py + 2*py*dy*t + t^2*dy*dy
+                     // derivative: 2*px*dx + 2*py*dy + (2*dx*dx+2*dy*dy)*t, set to 0 and solve
+                     float t = -(px*dx + py*dy) / (dx*dx + dy*dy);
+                     if (t >= 0.0f && t <= 1.0f)
+                        min_dist = dist;
+                  }
+               } else if (verts[i].type == STBTT_vcurve) {
+                  float x2 = verts[i-1].x *scale_x, y2 = verts[i-1].y *scale_y;
+                  float x1 = verts[i  ].cx*scale_x, y1 = verts[i  ].cy*scale_y;
+                  float box_x0 = STBTT_min(STBTT_min(x0,x1),x2);
+                  float box_y0 = STBTT_min(STBTT_min(y0,y1),y2);
+                  float box_x1 = STBTT_max(STBTT_max(x0,x1),x2);
+                  float box_y1 = STBTT_max(STBTT_max(y0,y1),y2);
+                  // coarse culling against bbox to avoid computing cubic unnecessarily
+                  if (sx > box_x0-min_dist && sx < box_x1+min_dist && sy > box_y0-min_dist && sy < box_y1+min_dist) {
+                     int num=0;
+                     float ax = x1-x0, ay = y1-y0;
+                     float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2;
+                     float mx = x0 - sx, my = y0 - sy;
+                     float res[3],px,py,t,it;
+                     float a_inv = precompute[i];
+                     if (a_inv == 0.0) { // if a_inv is 0, it's 2nd degree so use quadratic formula
+                        float a = 3*(ax*bx + ay*by);
+                        float b = 2*(ax*ax + ay*ay) + (mx*bx+my*by);
+                        float c = mx*ax+my*ay;
+                        if (a == 0.0) { // if a is 0, it's linear
+                           if (b != 0.0) {
+                              res[num++] = -c/b;
+                           }
+                        } else {
+                           float discriminant = b*b - 4*a*c;
+                           if (discriminant < 0)
+                              num = 0;
+                           else {
+                              float root = (float) STBTT_sqrt(discriminant);
+                              res[0] = (-b - root)/(2*a);
+                              res[1] = (-b + root)/(2*a);
+                              num = 2; // don't bother distinguishing 1-solution case, as code below will still work
+                           }
+                        }
+                     } else {
+                        float b = 3*(ax*bx + ay*by) * a_inv; // could precompute this as it doesn't depend on sample point
+                        float c = (2*(ax*ax + ay*ay) + (mx*bx+my*by)) * a_inv;
+                        float d = (mx*ax+my*ay) * a_inv;
+                        num = stbtt__solve_cubic(b, c, d, res);
+                     }
+                     if (num >= 1 && res[0] >= 0.0f && res[0] <= 1.0f) {
+                        t = res[0], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                     if (num >= 2 && res[1] >= 0.0f && res[1] <= 1.0f) {
+                        t = res[1], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                     if (num >= 3 && res[2] >= 0.0f && res[2] <= 1.0f) {
+                        t = res[2], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                  }
+               }
+            }
+            if (winding == 0)
+               min_dist = -min_dist;  // if outside the shape, value is negative
+            val = onedge_value + pixel_dist_scale * min_dist;
+            if (val < 0)
+               val = 0;
+            else if (val > 255)
+               val = 255;
+            data[(y-iy0)*w+(x-ix0)] = (unsigned char) val;
+         }
+      }
+      STBTT_free(precompute, info->userdata);
+      STBTT_free(verts, info->userdata);
+   }
+   return data;
+}   
+
+STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff)
+{
+   return stbtt_GetGlyphSDF(info, scale, stbtt_FindGlyphIndex(info, codepoint), padding, onedge_value, pixel_dist_scale, width, height, xoff, yoff);
+}
+
+STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata)
+{
+   STBTT_free(bitmap, userdata);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//
+// font name matching -- recommended not to use this
+//
+
+// check if a utf8 string contains a prefix which is the utf16 string; if so return length of matching utf8 string
+static stbtt_int32 stbtt__CompareUTF8toUTF16_bigendian_prefix(stbtt_uint8 *s1, stbtt_int32 len1, stbtt_uint8 *s2, stbtt_int32 len2) 
+{
+   stbtt_int32 i=0;
+
+   // convert utf16 to utf8 and compare the results while converting
+   while (len2) {
+      stbtt_uint16 ch = s2[0]*256 + s2[1];
+      if (ch < 0x80) {
+         if (i >= len1) return -1;
+         if (s1[i++] != ch) return -1;
+      } else if (ch < 0x800) {
+         if (i+1 >= len1) return -1;
+         if (s1[i++] != 0xc0 + (ch >> 6)) return -1;
+         if (s1[i++] != 0x80 + (ch & 0x3f)) return -1;
+      } else if (ch >= 0xd800 && ch < 0xdc00) {
+         stbtt_uint32 c;
+         stbtt_uint16 ch2 = s2[2]*256 + s2[3];
+         if (i+3 >= len1) return -1;
+         c = ((ch - 0xd800) << 10) + (ch2 - 0xdc00) + 0x10000;
+         if (s1[i++] != 0xf0 + (c >> 18)) return -1;
+         if (s1[i++] != 0x80 + ((c >> 12) & 0x3f)) return -1;
+         if (s1[i++] != 0x80 + ((c >>  6) & 0x3f)) return -1;
+         if (s1[i++] != 0x80 + ((c      ) & 0x3f)) return -1;
+         s2 += 2; // plus another 2 below
+         len2 -= 2;
+      } else if (ch >= 0xdc00 && ch < 0xe000) {
+         return -1;
+      } else {
+         if (i+2 >= len1) return -1;
+         if (s1[i++] != 0xe0 + (ch >> 12)) return -1;
+         if (s1[i++] != 0x80 + ((ch >> 6) & 0x3f)) return -1;
+         if (s1[i++] != 0x80 + ((ch     ) & 0x3f)) return -1;
+      }
+      s2 += 2;
+      len2 -= 2;
+   }
+   return i;
+}
+
+static int stbtt_CompareUTF8toUTF16_bigendian_internal(char *s1, int len1, char *s2, int len2) 
+{
+   return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((stbtt_uint8*) s1, len1, (stbtt_uint8*) s2, len2);
+}
+
+// returns results in whatever encoding you request... but note that 2-byte encodings
+// will be BIG-ENDIAN... use stbtt_CompareUTF8toUTF16_bigendian() to compare
+STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID)
+{
+   stbtt_int32 i,count,stringOffset;
+   stbtt_uint8 *fc = font->data;
+   stbtt_uint32 offset = font->fontstart;
+   stbtt_uint32 nm = stbtt__find_table(fc, offset, "name");
+   if (!nm) return NULL;
+
+   count = ttUSHORT(fc+nm+2);
+   stringOffset = nm + ttUSHORT(fc+nm+4);
+   for (i=0; i < count; ++i) {
+      stbtt_uint32 loc = nm + 6 + 12 * i;
+      if (platformID == ttUSHORT(fc+loc+0) && encodingID == ttUSHORT(fc+loc+2)
+          && languageID == ttUSHORT(fc+loc+4) && nameID == ttUSHORT(fc+loc+6)) {
+         *length = ttUSHORT(fc+loc+8);
+         return (const char *) (fc+stringOffset+ttUSHORT(fc+loc+10));
+      }
+   }
+   return NULL;
+}
+
+static int stbtt__matchpair(stbtt_uint8 *fc, stbtt_uint32 nm, stbtt_uint8 *name, stbtt_int32 nlen, stbtt_int32 target_id, stbtt_int32 next_id)
+{
+   stbtt_int32 i;
+   stbtt_int32 count = ttUSHORT(fc+nm+2);
+   stbtt_int32 stringOffset = nm + ttUSHORT(fc+nm+4);
+
+   for (i=0; i < count; ++i) {
+      stbtt_uint32 loc = nm + 6 + 12 * i;
+      stbtt_int32 id = ttUSHORT(fc+loc+6);
+      if (id == target_id) {
+         // find the encoding
+         stbtt_int32 platform = ttUSHORT(fc+loc+0), encoding = ttUSHORT(fc+loc+2), language = ttUSHORT(fc+loc+4);
+
+         // is this a Unicode encoding?
+         if (platform == 0 || (platform == 3 && encoding == 1) || (platform == 3 && encoding == 10)) {
+            stbtt_int32 slen = ttUSHORT(fc+loc+8);
+            stbtt_int32 off = ttUSHORT(fc+loc+10);
+
+            // check if there's a prefix match
+            stbtt_int32 matchlen = stbtt__CompareUTF8toUTF16_bigendian_prefix(name, nlen, fc+stringOffset+off,slen);
+            if (matchlen >= 0) {
+               // check for target_id+1 immediately following, with same encoding & language
+               if (i+1 < count && ttUSHORT(fc+loc+12+6) == next_id && ttUSHORT(fc+loc+12) == platform && ttUSHORT(fc+loc+12+2) == encoding && ttUSHORT(fc+loc+12+4) == language) {
+                  slen = ttUSHORT(fc+loc+12+8);
+                  off = ttUSHORT(fc+loc+12+10);
+                  if (slen == 0) {
+                     if (matchlen == nlen)
+                        return 1;
+                  } else if (matchlen < nlen && name[matchlen] == ' ') {
+                     ++matchlen;
+                     if (stbtt_CompareUTF8toUTF16_bigendian_internal((char*) (name+matchlen), nlen-matchlen, (char*)(fc+stringOffset+off),slen))
+                        return 1;
+                  }
+               } else {
+                  // if nothing immediately following
+                  if (matchlen == nlen)
+                     return 1;
+               }
+            }
+         }
+
+         // @TODO handle other encodings
+      }
+   }
+   return 0;
+}
+
+static int stbtt__matches(stbtt_uint8 *fc, stbtt_uint32 offset, stbtt_uint8 *name, stbtt_int32 flags)
+{
+   stbtt_int32 nlen = (stbtt_int32) STBTT_strlen((char *) name);
+   stbtt_uint32 nm,hd;
+   if (!stbtt__isfont(fc+offset)) return 0;
+
+   // check italics/bold/underline flags in macStyle...
+   if (flags) {
+      hd = stbtt__find_table(fc, offset, "head");
+      if ((ttUSHORT(fc+hd+44) & 7) != (flags & 7)) return 0;
+   }
+
+   nm = stbtt__find_table(fc, offset, "name");
+   if (!nm) return 0;
+
+   if (flags) {
+      // if we checked the macStyle flags, then just check the family and ignore the subfamily
+      if (stbtt__matchpair(fc, nm, name, nlen, 16, -1))  return 1;
+      if (stbtt__matchpair(fc, nm, name, nlen,  1, -1))  return 1;
+      if (stbtt__matchpair(fc, nm, name, nlen,  3, -1))  return 1;
+   } else {
+      if (stbtt__matchpair(fc, nm, name, nlen, 16, 17))  return 1;
+      if (stbtt__matchpair(fc, nm, name, nlen,  1,  2))  return 1;
+      if (stbtt__matchpair(fc, nm, name, nlen,  3, -1))  return 1;
+   }
+
+   return 0;
+}
+
+static int stbtt_FindMatchingFont_internal(unsigned char *font_collection, char *name_utf8, stbtt_int32 flags)
+{
+   stbtt_int32 i;
+   for (i=0;;++i) {
+      stbtt_int32 off = stbtt_GetFontOffsetForIndex(font_collection, i);
+      if (off < 0) return off;
+      if (stbtt__matches((stbtt_uint8 *) font_collection, off, (stbtt_uint8*) name_utf8, flags))
+         return off;
+   }
+}
+
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset,
+                                float pixel_height, unsigned char *pixels, int pw, int ph,
+                                int first_char, int num_chars, stbtt_bakedchar *chardata)
+{
+   return stbtt_BakeFontBitmap_internal((unsigned char *) data, offset, pixel_height, pixels, pw, ph, first_char, num_chars, chardata);
+}
+
+STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index)
+{
+   return stbtt_GetFontOffsetForIndex_internal((unsigned char *) data, index);   
+}
+
+STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data)
+{
+   return stbtt_GetNumberOfFonts_internal((unsigned char *) data);
+}
+
+STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset)
+{
+   return stbtt_InitFont_internal(info, (unsigned char *) data, offset);
+}
+
+STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags)
+{
+   return stbtt_FindMatchingFont_internal((unsigned char *) fontdata, (char *) name, flags);
+}
+
+STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2)
+{
+   return stbtt_CompareUTF8toUTF16_bigendian_internal((char *) s1, len1, (char *) s2, len2);
+}
+
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic pop
+#endif
+
+#endif // STB_TRUETYPE_IMPLEMENTATION
+
+
+// FULL VERSION HISTORY
+//
+//   1.19 (2018-02-11) OpenType GPOS kerning (horizontal only), STBTT_fmod
+//   1.18 (2018-01-29) add missing function
+//   1.17 (2017-07-23) make more arguments const; doc fix
+//   1.16 (2017-07-12) SDF support
+//   1.15 (2017-03-03) make more arguments const
+//   1.14 (2017-01-16) num-fonts-in-TTC function
+//   1.13 (2017-01-02) support OpenType fonts, certain Apple fonts
+//   1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual
+//   1.11 (2016-04-02) fix unused-variable warning
+//   1.10 (2016-04-02) allow user-defined fabs() replacement
+//                     fix memory leak if fontsize=0.0
+//                     fix warning from duplicate typedef
+//   1.09 (2016-01-16) warning fix; avoid crash on outofmem; use alloc userdata for PackFontRanges
+//   1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges
+//   1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints;
+//                     allow PackFontRanges to pack and render in separate phases;
+//                     fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?);
+//                     fixed an assert() bug in the new rasterizer
+//                     replace assert() with STBTT_assert() in new rasterizer
+//   1.06 (2015-07-14) performance improvements (~35% faster on x86 and x64 on test machine)
+//                     also more precise AA rasterizer, except if shapes overlap
+//                     remove need for STBTT_sort
+//   1.05 (2015-04-15) fix misplaced definitions for STBTT_STATIC
+//   1.04 (2015-04-15) typo in example
+//   1.03 (2015-04-12) STBTT_STATIC, fix memory leak in new packing, various fixes
+//   1.02 (2014-12-10) fix various warnings & compile issues w/ stb_rect_pack, C++
+//   1.01 (2014-12-08) fix subpixel position when oversampling to exactly match
+//                        non-oversampled; STBTT_POINT_SIZE for packed case only
+//   1.00 (2014-12-06) add new PackBegin etc. API, w/ support for oversampling
+//   0.99 (2014-09-18) fix multiple bugs with subpixel rendering (ryg)
+//   0.9  (2014-08-07) support certain mac/iOS fonts without an MS platformID
+//   0.8b (2014-07-07) fix a warning
+//   0.8  (2014-05-25) fix a few more warnings
+//   0.7  (2013-09-25) bugfix: subpixel glyph bug fixed in 0.5 had come back
+//   0.6c (2012-07-24) improve documentation
+//   0.6b (2012-07-20) fix a few more warnings
+//   0.6  (2012-07-17) fix warnings; added stbtt_ScaleForMappingEmToPixels,
+//                        stbtt_GetFontBoundingBox, stbtt_IsGlyphEmpty
+//   0.5  (2011-12-09) bugfixes:
+//                        subpixel glyph renderer computed wrong bounding box
+//                        first vertex of shape can be off-curve (FreeSans)
+//   0.4b (2011-12-03) fixed an error in the font baking example
+//   0.4  (2011-12-01) kerning, subpixel rendering (tor)
+//                    bugfixes for:
+//                        codepoint-to-glyph conversion using table fmt=12
+//                        codepoint-to-glyph conversion using table fmt=4
+//                        stbtt_GetBakedQuad with non-square texture (Zer)
+//                    updated Hello World! sample to use kerning and subpixel
+//                    fixed some warnings
+//   0.3  (2009-06-24) cmap fmt=12, compound shapes (MM)
+//                    userdata, malloc-from-userdata, non-zero fill (stb)
+//   0.2  (2009-03-11) Fix unsigned/signed char warnings
+//   0.1  (2009-03-09) First public release
+//
+
+/*
+------------------------------------------------------------------------------
+This software is available under 2 licenses -- choose whichever you prefer.
+------------------------------------------------------------------------------
+ALTERNATIVE A - MIT License
+Copyright (c) 2017 Sean Barrett
+Permission is hereby granted, free of charge, to any person obtaining a copy of 
+this software and associated documentation files (the "Software"), to deal in 
+the Software without restriction, including without limitation the rights to 
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies 
+of the Software, and to permit persons to whom the Software is furnished to do 
+so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all 
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 
+SOFTWARE.
+------------------------------------------------------------------------------
+ALTERNATIVE B - Public Domain (www.unlicense.org)
+This is free and unencumbered software released into the public domain.
+Anyone is free to copy, modify, publish, use, compile, sell, or distribute this 
+software, either in source code form or as a compiled binary, for any purpose, 
+commercial or non-commercial, and by any means.
+In jurisdictions that recognize copyright laws, the author or authors of this 
+software dedicate any and all copyright interest in the software to the public 
+domain. We make this dedication for the benefit of the public at large and to 
+the detriment of our heirs and successors. We intend this dedication to be an 
+overt act of relinquishment in perpetuity of all present and future rights to 
+this software under copyright law.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 
+AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 
+ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION 
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+------------------------------------------------------------------------------
+*/
diff --git a/src/external/utf8.h b/src/external/utf8.h
new file mode 100644
index 0000000..1554884
--- /dev/null
+++ b/src/external/utf8.h
@@ -0,0 +1,1258 @@
+// The latest version of this library is available on GitHub;
+// https://github.com/sheredom/utf8.h
+
+// This is free and unencumbered software released into the public domain.
+//
+// Anyone is free to copy, modify, publish, use, compile, sell, or
+// distribute this software, either in source code form or as a compiled
+// binary, for any purpose, commercial or non-commercial, and by any
+// means.
+//
+// In jurisdictions that recognize copyright laws, the author or authors
+// of this software dedicate any and all copyright interest in the
+// software to the public domain. We make this dedication for the benefit
+// of the public at large and to the detriment of our heirs and
+// successors. We intend this dedication to be an overt act of
+// relinquishment in perpetuity of all present and future rights to this
+// software under copyright law.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+// IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+// OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+// OTHER DEALINGS IN THE SOFTWARE.
+//
+// For more information, please refer to <http://unlicense.org/>
+
+#ifndef SHEREDOM_UTF8_H_INCLUDED
+#define SHEREDOM_UTF8_H_INCLUDED
+
+#if defined(_MSC_VER)
+#pragma warning(push)
+
+// disable 'bytes padding added after construct' warning
+#pragma warning(disable : 4820)
+#endif
+
+#include <stddef.h>
+#include <stdlib.h>
+
+#if defined(_MSC_VER)
+#pragma warning(pop)
+#endif
+
+#if defined(_MSC_VER)
+typedef __int32 utf8_int32_t;
+#else
+#include <stdint.h>
+typedef int32_t utf8_int32_t;
+#endif
+
+#if defined(__clang__)
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wold-style-cast"
+#pragma clang diagnostic ignored "-Wcast-qual"
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+	
+#if defined(__clang__) || defined(__GNUC__)
+#define utf8_nonnull __attribute__((nonnull))
+#define utf8_pure __attribute__((pure))
+#define utf8_restrict __restrict__
+#define utf8_weak __attribute__((weak))
+#elif defined(_MSC_VER)
+#define utf8_nonnull
+#define utf8_pure
+#define utf8_restrict __restrict
+#define utf8_weak __inline
+#else
+#error Non clang, non gcc, non MSVC compiler found!
+#endif
+	
+#ifdef __cplusplus
+#define utf8_null NULL
+#else
+#define utf8_null 0
+#endif
+	
+	// Return less than 0, 0, greater than 0 if src1 < src2, src1 == src2, src1 >
+	// src2 respectively, case insensitive.
+	utf8_nonnull utf8_pure utf8_weak int utf8casecmp(const void *src1,
+													 const void *src2);
+	
+	// Append the utf8 string src onto the utf8 string dst.
+	utf8_nonnull utf8_weak void *utf8cat(void *utf8_restrict dst,
+										 const void *utf8_restrict src);
+	
+	// Find the first match of the utf8 codepoint chr in the utf8 string src.
+	utf8_nonnull utf8_pure utf8_weak void *utf8chr(const void *src,
+												   utf8_int32_t chr);
+	
+	// Return less than 0, 0, greater than 0 if src1 < src2,
+	// src1 == src2, src1 > src2 respectively.
+	utf8_nonnull utf8_pure utf8_weak int utf8cmp(const void *src1,
+												 const void *src2);
+	
+	// Copy the utf8 string src onto the memory allocated in dst.
+	utf8_nonnull utf8_weak void *utf8cpy(void *utf8_restrict dst,
+										 const void *utf8_restrict src);
+	
+	// Number of utf8 codepoints in the utf8 string src that consists entirely
+	// of utf8 codepoints not from the utf8 string reject.
+	utf8_nonnull utf8_pure utf8_weak size_t utf8cspn(const void *src,
+													 const void *reject);
+	
+	// Duplicate the utf8 string src by getting its size, malloc'ing a new buffer
+	// copying over the data, and returning that. Or 0 if malloc failed.
+	utf8_nonnull utf8_weak void *utf8dup(const void *src);
+	
+	// Number of utf8 codepoints in the utf8 string str,
+	// excluding the null terminating byte.
+	utf8_nonnull utf8_pure utf8_weak size_t utf8len(const void *str);
+	
+	// Return less than 0, 0, greater than 0 if src1 < src2, src1 == src2, src1 >
+	// src2 respectively, case insensitive. Checking at most n bytes of each utf8
+	// string.
+	utf8_nonnull utf8_pure utf8_weak int utf8ncasecmp(const void *src1,
+													  const void *src2, size_t n);
+	
+	// Append the utf8 string src onto the utf8 string dst,
+	// writing at most n+1 bytes. Can produce an invalid utf8
+	// string if n falls partway through a utf8 codepoint.
+	utf8_nonnull utf8_weak void *utf8ncat(void *utf8_restrict dst,
+										  const void *utf8_restrict src, size_t n);
+	
+	// Return less than 0, 0, greater than 0 if src1 < src2,
+	// src1 == src2, src1 > src2 respectively. Checking at most n
+	// bytes of each utf8 string.
+	utf8_nonnull utf8_pure utf8_weak int utf8ncmp(const void *src1,
+												  const void *src2, size_t n);
+	
+	// Copy the utf8 string src onto the memory allocated in dst.
+	// Copies at most n bytes. If there is no terminating null byte in
+	// the first n bytes of src, the string placed into dst will not be
+	// null-terminated. If the size (in bytes) of src is less than n,
+	// extra null terminating bytes are appended to dst such that at
+	// total of n bytes are written. Can produce an invalid utf8
+	// string if n falls partway through a utf8 codepoint.
+	utf8_nonnull utf8_weak void *utf8ncpy(void *utf8_restrict dst,
+										  const void *utf8_restrict src, size_t n);
+	
+	// Similar to utf8dup, except that at most n bytes of src are copied. If src is
+	// longer than n, only n bytes are copied and a null byte is added.
+	//
+	// Returns a new string if successful, 0 otherwise
+	utf8_nonnull utf8_weak void *utf8ndup(const void *src, size_t n);
+	
+	// Locates the first occurence in the utf8 string str of any byte in the
+	// utf8 string accept, or 0 if no match was found.
+	utf8_nonnull utf8_pure utf8_weak void *utf8pbrk(const void *str,
+													const void *accept);
+	
+	// Find the last match of the utf8 codepoint chr in the utf8 string src.
+	utf8_nonnull utf8_pure utf8_weak void *utf8rchr(const void *src, int chr);
+	
+	// Number of bytes in the utf8 string str,
+	// including the null terminating byte.
+	utf8_nonnull utf8_pure utf8_weak size_t utf8size(const void *str);
+	
+	// Number of utf8 codepoints in the utf8 string src that consists entirely
+	// of utf8 codepoints from the utf8 string accept.
+	utf8_nonnull utf8_pure utf8_weak size_t utf8spn(const void *src,
+													const void *accept);
+	
+	// The position of the utf8 string needle in the utf8 string haystack.
+	utf8_nonnull utf8_pure utf8_weak void *utf8str(const void *haystack,
+												   const void *needle);
+	
+	// The position of the utf8 string needle in the utf8 string haystack, case
+	// insensitive.
+	utf8_nonnull utf8_pure utf8_weak void *utf8casestr(const void *haystack,
+													   const void *needle);
+	
+	// Return 0 on success, or the position of the invalid
+	// utf8 codepoint on failure.
+	utf8_nonnull utf8_pure utf8_weak void *utf8valid(const void *str);
+	
+	// Sets out_codepoint to the next utf8 codepoint in str, and returns the address
+	// of the utf8 codepoint after the current one in str.
+	utf8_nonnull utf8_weak void *
+		utf8codepoint(const void *utf8_restrict str,
+					  utf8_int32_t *utf8_restrict out_codepoint);
+	
+	// Returns the size of the given codepoint in bytes.
+	utf8_weak size_t utf8codepointsize(utf8_int32_t chr);
+	
+	// Write a codepoint to the given string, and return the address to the next
+	// place after the written codepoint. Pass how many bytes left in the buffer to
+	// n. If there is not enough space for the codepoint, this function returns
+	// null.
+	utf8_nonnull utf8_weak void *utf8catcodepoint(void *utf8_restrict str,
+												  utf8_int32_t chr, size_t n);
+	
+	// Returns 1 if the given character is lowercase, or 0 if it is not.
+	utf8_weak int utf8islower(utf8_int32_t chr);
+	
+	// Returns 1 if the given character is uppercase, or 0 if it is not.
+	utf8_weak int utf8isupper(utf8_int32_t chr);
+	
+	// Transform the given string into all lowercase codepoints.
+	utf8_nonnull utf8_weak void utf8lwr(void *utf8_restrict str);
+	
+	// Transform the given string into all uppercase codepoints.
+	utf8_nonnull utf8_weak void utf8upr(void *utf8_restrict str);
+	
+	// Make a codepoint lower case if possible.
+	utf8_weak utf8_int32_t utf8lwrcodepoint(utf8_int32_t cp);
+	
+	// Make a codepoint upper case if possible.
+	utf8_weak utf8_int32_t utf8uprcodepoint(utf8_int32_t cp);
+	
+#undef utf8_weak
+#undef utf8_pure
+#undef utf8_nonnull
+	
+	int utf8casecmp(const void *src1, const void *src2) {
+		utf8_int32_t src1_cp, src2_cp, src1_orig_cp, src2_orig_cp;
+		
+		for (;;) {
+			src1 = utf8codepoint(src1, &src1_cp);
+			src2 = utf8codepoint(src2, &src2_cp);
+			
+			// take a copy of src1 & src2
+			src1_orig_cp = src1_cp;
+			src2_orig_cp = src2_cp;
+			
+			// lower the srcs if required
+			src1_cp = utf8lwrcodepoint(src1_cp);
+			src2_cp = utf8lwrcodepoint(src2_cp);
+			
+			// check if the lowered codepoints match
+			if ((0 == src1_orig_cp) && (0 == src2_orig_cp)) {
+				return 0;
+			} else if (src1_cp == src2_cp) {
+				continue;
+			}
+			
+			// if they don't match, then we return the difference between the characters
+			return src1_cp - src2_cp;
+		}
+	}
+	
+	void *utf8cat(void *utf8_restrict dst, const void *utf8_restrict src) {
+		char *d = (char *)dst;
+		const char *s = (const char *)src;
+		
+		// find the null terminating byte in dst
+		while ('\0' != *d) {
+			d++;
+		}
+		
+		// overwriting the null terminating byte in dst, append src byte-by-byte
+		while ('\0' != *s) {
+			*d++ = *s++;
+		}
+		
+		// write out a new null terminating byte into dst
+		*d = '\0';
+		
+		return dst;
+	}
+	
+	void *utf8chr(const void *src, utf8_int32_t chr) {
+		char c[5] = {'\0', '\0', '\0', '\0', '\0'};
+		
+		if (0 == chr) {
+			// being asked to return position of null terminating byte, so
+			// just run s to the end, and return!
+			const char *s = (const char *)src;
+			while ('\0' != *s) {
+				s++;
+			}
+			return (void *)s;
+		} else if (0 == ((utf8_int32_t)0xffffff80 & chr)) {
+			// 1-byte/7-bit ascii
+			// (0b0xxxxxxx)
+			c[0] = (char)chr;
+		} else if (0 == ((utf8_int32_t)0xfffff800 & chr)) {
+			// 2-byte/11-bit utf8 code point
+			// (0b110xxxxx 0b10xxxxxx)
+			c[0] = 0xc0 | (char)(chr >> 6);
+			c[1] = 0x80 | (char)(chr & 0x3f);
+		} else if (0 == ((utf8_int32_t)0xffff0000 & chr)) {
+			// 3-byte/16-bit utf8 code point
+			// (0b1110xxxx 0b10xxxxxx 0b10xxxxxx)
+			c[0] = 0xe0 | (char)(chr >> 12);
+			c[1] = 0x80 | (char)((chr >> 6) & 0x3f);
+			c[2] = 0x80 | (char)(chr & 0x3f);
+		} else { // if (0 == ((int)0xffe00000 & chr)) {
+			// 4-byte/21-bit utf8 code point
+			// (0b11110xxx 0b10xxxxxx 0b10xxxxxx 0b10xxxxxx)
+			c[0] = 0xf0 | (char)(chr >> 18);
+			c[1] = 0x80 | (char)((chr >> 12) & 0x3f);
+			c[2] = 0x80 | (char)((chr >> 6) & 0x3f);
+			c[3] = 0x80 | (char)(chr & 0x3f);
+		}
+		
+		// we've made c into a 2 utf8 codepoint string, one for the chr we are
+		// seeking, another for the null terminating byte. Now use utf8str to
+		// search
+		return utf8str(src, c);
+	}
+	
+	int utf8cmp(const void *src1, const void *src2) {
+		const unsigned char *s1 = (const unsigned char *)src1;
+		const unsigned char *s2 = (const unsigned char *)src2;
+		
+		while (('\0' != *s1) || ('\0' != *s2)) {
+			if (*s1 < *s2) {
+				return -1;
+			} else if (*s1 > *s2) {
+				return 1;
+			}
+			
+			s1++;
+			s2++;
+		}
+		
+		// both utf8 strings matched
+		return 0;
+	}
+	
+	int utf8coll(const void *src1, const void *src2);
+	
+	void *utf8cpy(void *utf8_restrict dst, const void *utf8_restrict src) {
+		char *d = (char *)dst;
+		const char *s = (const char *)src;
+		
+		// overwriting anything previously in dst, write byte-by-byte
+		// from src
+		while ('\0' != *s) {
+			*d++ = *s++;
+		}
+		
+		// append null terminating byte
+		*d = '\0';
+		
+		return dst;
+	}
+	
+	size_t utf8cspn(const void *src, const void *reject) {
+		const char *s = (const char *)src;
+		size_t chars = 0;
+		
+		while ('\0' != *s) {
+			const char *r = (const char *)reject;
+			size_t offset = 0;
+			
+			while ('\0' != *r) {
+				// checking that if *r is the start of a utf8 codepoint
+				// (it is not 0b10xxxxxx) and we have successfully matched
+				// a previous character (0 < offset) - we found a match
+				if ((0x80 != (0xc0 & *r)) && (0 < offset)) {
+					return chars;
+				} else {
+					if (*r == s[offset]) {
+						// part of a utf8 codepoint matched, so move our checking
+						// onwards to the next byte
+						offset++;
+						r++;
+					} else {
+						// r could be in the middle of an unmatching utf8 code point,
+						// so we need to march it on to the next character beginning,
+						
+						do {
+							r++;
+						} while (0x80 == (0xc0 & *r));
+						
+						// reset offset too as we found a mismatch
+						offset = 0;
+					}
+				}
+			}
+			
+			// the current utf8 codepoint in src did not match reject, but src
+			// could have been partway through a utf8 codepoint, so we need to
+			// march it onto the next utf8 codepoint starting byte
+			do {
+				s++;
+			} while ((0x80 == (0xc0 & *s)));
+			chars++;
+		}
+		
+		return chars;
+	}
+	
+	size_t utf8size(const void *str);
+	
+	void *utf8dup(const void *src) {
+		const char *s = (const char *)src;
+		char *n = utf8_null;
+		
+		// figure out how many bytes (including the terminator) we need to copy first
+		size_t bytes = utf8size(src);
+		
+		n = (char *)malloc(bytes);
+		
+		if (utf8_null == n) {
+			// out of memory so we bail
+			return utf8_null;
+		} else {
+			bytes = 0;
+			
+			// copy src byte-by-byte into our new utf8 string
+			while ('\0' != s[bytes]) {
+				n[bytes] = s[bytes];
+				bytes++;
+			}
+			
+			// append null terminating byte
+			n[bytes] = '\0';
+			return n;
+		}
+	}
+	
+	void *utf8fry(const void *str);
+	
+	size_t utf8len(const void *str) {
+		const unsigned char *s = (const unsigned char *)str;
+		size_t length = 0;
+		
+		while ('\0' != *s) {
+			if (0xf0 == (0xf8 & *s)) {
+				// 4-byte utf8 code point (began with 0b11110xxx)
+				s += 4;
+			} else if (0xe0 == (0xf0 & *s)) {
+				// 3-byte utf8 code point (began with 0b1110xxxx)
+				s += 3;
+			} else if (0xc0 == (0xe0 & *s)) {
+				// 2-byte utf8 code point (began with 0b110xxxxx)
+				s += 2;
+			} else { // if (0x00 == (0x80 & *s)) {
+				// 1-byte ascii (began with 0b0xxxxxxx)
+				s += 1;
+			}
+			
+			// no matter the bytes we marched s forward by, it was
+			// only 1 utf8 codepoint
+			length++;
+		}
+		
+		return length;
+	}
+	
+	int utf8ncasecmp(const void *src1, const void *src2, size_t n) {
+		utf8_int32_t src1_cp, src2_cp, src1_orig_cp, src2_orig_cp;
+		
+		do {
+			const unsigned char *const s1 = (const unsigned char *)src1;
+			const unsigned char *const s2 = (const unsigned char *)src2;
+			
+			// first check that we have enough bytes left in n to contain an entire
+			// codepoint
+			if (0 == n) {
+				return 0;
+			}
+			
+			if ((1 == n) && ((0xc0 == (0xe0 & *s1)) || (0xc0 == (0xe0 & *s2)))) {
+				const utf8_int32_t c1 = (0xe0 & *s1);
+				const utf8_int32_t c2 = (0xe0 & *s2);
+				
+				if (c1 < c2) {
+					return -1;
+				} else if (c1 > c2) {
+					return 1;
+				} else {
+					return 0;
+				}
+			}
+			
+			if ((2 >= n) && ((0xe0 == (0xf0 & *s1)) || (0xe0 == (0xf0 & *s2)))) {
+				const utf8_int32_t c1 = (0xf0 & *s1);
+				const utf8_int32_t c2 = (0xf0 & *s2);
+				
+				if (c1 < c2) {
+					return -1;
+				} else if (c1 > c2) {
+					return 1;
+				} else {
+					return 0;
+				}
+			}
+			
+			if ((3 >= n) && ((0xf0 == (0xf8 & *s1)) || (0xf0 == (0xf8 & *s2)))) {
+				const utf8_int32_t c1 = (0xf8 & *s1);
+				const utf8_int32_t c2 = (0xf8 & *s2);
+				
+				if (c1 < c2) {
+					return -1;
+				} else if (c1 > c2) {
+					return 1;
+				} else {
+					return 0;
+				}
+			}
+			
+			src1 = utf8codepoint(src1, &src1_cp);
+			src2 = utf8codepoint(src2, &src2_cp);
+			n -= utf8codepointsize(src1_cp);
+			
+			// Take a copy of src1 & src2
+			src1_orig_cp = src1_cp;
+			src2_orig_cp = src2_cp;
+			
+			// Lower srcs if required
+			src1_cp = utf8lwrcodepoint(src1_cp);
+			src2_cp = utf8lwrcodepoint(src2_cp);
+			
+			// Check if the lowered codepoints match
+			if ((0 == src1_orig_cp) && (0 == src2_orig_cp)) {
+				return 0;
+			} else if (src1_cp == src2_cp) {
+				continue;
+			}
+			
+			// If they don't match, then we return which of the original's are less
+			if (src1_orig_cp < src2_orig_cp) {
+				return -1;
+			} else if (src1_orig_cp > src2_orig_cp) {
+				return 1;
+			}
+		} while (0 < n);
+		
+		// both utf8 strings matched
+		return 0;
+	}
+	
+	void *utf8ncat(void *utf8_restrict dst, const void *utf8_restrict src,
+				   size_t n) {
+		char *d = (char *)dst;
+		const char *s = (const char *)src;
+		
+		// find the null terminating byte in dst
+		while ('\0' != *d) {
+			d++;
+		}
+		
+		// overwriting the null terminating byte in dst, append src byte-by-byte
+		// stopping if we run out of space
+		do {
+			*d++ = *s++;
+		} while (('\0' != *s) && (0 != --n));
+		
+		// write out a new null terminating byte into dst
+		*d = '\0';
+		
+		return dst;
+	}
+	
+	int utf8ncmp(const void *src1, const void *src2, size_t n) {
+		const unsigned char *s1 = (const unsigned char *)src1;
+		const unsigned char *s2 = (const unsigned char *)src2;
+		
+		while ((0 != n--) && (('\0' != *s1) || ('\0' != *s2))) {
+			if (*s1 < *s2) {
+				return -1;
+			} else if (*s1 > *s2) {
+				return 1;
+			}
+			
+			s1++;
+			s2++;
+		}
+		
+		// both utf8 strings matched
+		return 0;
+	}
+	
+	void *utf8ncpy(void *utf8_restrict dst, const void *utf8_restrict src,
+				   size_t n) {
+		char *d = (char *)dst;
+		const char *s = (const char *)src;
+		size_t index;
+		
+		// overwriting anything previously in dst, write byte-by-byte
+		// from src
+		for (index = 0; index < n; index++) {
+			d[index] = s[index];
+			if ('\0' == s[index]) {
+				break;
+			}
+		}
+		
+		// append null terminating byte
+		for (; index < n; index++) {
+			d[index] = 0;
+		}
+		
+		return dst;
+	}
+	
+	void *utf8ndup(const void *src, size_t n) {
+		const char *s = (const char *)src;
+		char *c = utf8_null;
+		size_t bytes = 0;
+		
+		// Find the end of the string or stop when n is reached
+		while ('\0' != s[bytes] && bytes < n) {
+			bytes++;
+		}
+		
+		// In case bytes is actually less than n, we need to set it
+		// to be used later in the copy byte by byte.
+		n = bytes;
+		
+		c = (char *)malloc(bytes + 1);
+		if (utf8_null == c) {
+			// out of memory so we bail
+			return utf8_null;
+		}
+		
+		bytes = 0;
+		
+		// copy src byte-by-byte into our new utf8 string
+		while ('\0' != s[bytes] && bytes < n) {
+			c[bytes] = s[bytes];
+			bytes++;
+		}
+		
+		// append null terminating byte
+		c[bytes] = '\0';
+		return c;
+	}
+	
+	void *utf8rchr(const void *src, int chr) {
+		const char *s = (const char *)src;
+		const char *match = utf8_null;
+		char c[5] = {'\0', '\0', '\0', '\0', '\0'};
+		
+		if (0 == chr) {
+			// being asked to return position of null terminating byte, so
+			// just run s to the end, and return!
+			while ('\0' != *s) {
+				s++;
+			}
+			return (void *)s;
+		} else if (0 == ((int)0xffffff80 & chr)) {
+			// 1-byte/7-bit ascii
+			// (0b0xxxxxxx)
+			c[0] = (char)chr;
+		} else if (0 == ((int)0xfffff800 & chr)) {
+			// 2-byte/11-bit utf8 code point
+			// (0b110xxxxx 0b10xxxxxx)
+			c[0] = 0xc0 | (char)(chr >> 6);
+			c[1] = 0x80 | (char)(chr & 0x3f);
+		} else if (0 == ((int)0xffff0000 & chr)) {
+			// 3-byte/16-bit utf8 code point
+			// (0b1110xxxx 0b10xxxxxx 0b10xxxxxx)
+			c[0] = 0xe0 | (char)(chr >> 12);
+			c[1] = 0x80 | (char)((chr >> 6) & 0x3f);
+			c[2] = 0x80 | (char)(chr & 0x3f);
+		} else { // if (0 == ((int)0xffe00000 & chr)) {
+			// 4-byte/21-bit utf8 code point
+			// (0b11110xxx 0b10xxxxxx 0b10xxxxxx 0b10xxxxxx)
+			c[0] = 0xf0 | (char)(chr >> 18);
+			c[1] = 0x80 | (char)((chr >> 12) & 0x3f);
+			c[2] = 0x80 | (char)((chr >> 6) & 0x3f);
+			c[3] = 0x80 | (char)(chr & 0x3f);
+		}
+		
+		// we've created a 2 utf8 codepoint string in c that is
+		// the utf8 character asked for by chr, and a null
+		// terminating byte
+		
+		while ('\0' != *s) {
+			size_t offset = 0;
+			
+			while (s[offset] == c[offset]) {
+				offset++;
+			}
+			
+			if ('\0' == c[offset]) {
+				// we found a matching utf8 code point
+				match = s;
+				s += offset;
+			} else {
+				s += offset;
+				
+				// need to march s along to next utf8 codepoint start
+				// (the next byte that doesn't match 0b10xxxxxx)
+				if ('\0' != *s) {
+					do {
+						s++;
+					} while (0x80 == (0xc0 & *s));
+				}
+			}
+		}
+		
+		// return the last match we found (or 0 if no match was found)
+		return (void *)match;
+	}
+	
+	void *utf8pbrk(const void *str, const void *accept) {
+		const char *s = (const char *)str;
+		
+		while ('\0' != *s) {
+			const char *a = (const char *)accept;
+			size_t offset = 0;
+			
+			while ('\0' != *a) {
+				// checking that if *a is the start of a utf8 codepoint
+				// (it is not 0b10xxxxxx) and we have successfully matched
+				// a previous character (0 < offset) - we found a match
+				if ((0x80 != (0xc0 & *a)) && (0 < offset)) {
+					return (void *)s;
+				} else {
+					if (*a == s[offset]) {
+						// part of a utf8 codepoint matched, so move our checking
+						// onwards to the next byte
+						offset++;
+						a++;
+					} else {
+						// r could be in the middle of an unmatching utf8 code point,
+						// so we need to march it on to the next character beginning,
+						
+						do {
+							a++;
+						} while (0x80 == (0xc0 & *a));
+						
+						// reset offset too as we found a mismatch
+						offset = 0;
+					}
+				}
+			}
+			
+			// we found a match on the last utf8 codepoint
+			if (0 < offset) {
+				return (void *)s;
+			}
+			
+			// the current utf8 codepoint in src did not match accept, but src
+			// could have been partway through a utf8 codepoint, so we need to
+			// march it onto the next utf8 codepoint starting byte
+			do {
+				s++;
+			} while ((0x80 == (0xc0 & *s)));
+		}
+		
+		return utf8_null;
+	}
+	
+	size_t utf8size(const void *str) {
+		const char *s = (const char *)str;
+		size_t size = 0;
+		while ('\0' != s[size]) {
+			size++;
+		}
+		
+		// we are including the null terminating byte in the size calculation
+		size++;
+		return size;
+	}
+	
+	size_t utf8spn(const void *src, const void *accept) {
+		const char *s = (const char *)src;
+		size_t chars = 0;
+		
+		while ('\0' != *s) {
+			const char *a = (const char *)accept;
+			size_t offset = 0;
+			
+			while ('\0' != *a) {
+				// checking that if *r is the start of a utf8 codepoint
+				// (it is not 0b10xxxxxx) and we have successfully matched
+				// a previous character (0 < offset) - we found a match
+				if ((0x80 != (0xc0 & *a)) && (0 < offset)) {
+					// found a match, so increment the number of utf8 codepoints
+					// that have matched and stop checking whether any other utf8
+					// codepoints in a match
+					chars++;
+					s += offset;
+					break;
+				} else {
+					if (*a == s[offset]) {
+						offset++;
+						a++;
+					} else {
+						// a could be in the middle of an unmatching utf8 codepoint,
+						// so we need to march it on to the next character beginning,
+						do {
+							a++;
+						} while (0x80 == (0xc0 & *a));
+						
+						// reset offset too as we found a mismatch
+						offset = 0;
+					}
+				}
+			}
+			
+			// if a got to its terminating null byte, then we didn't find a match.
+			// Return the current number of matched utf8 codepoints
+			if ('\0' == *a) {
+				return chars;
+			}
+		}
+		
+		return chars;
+	}
+	
+	void *utf8str(const void *haystack, const void *needle) {
+		const char *h = (const char *)haystack;
+		utf8_int32_t throwaway_codepoint;
+		
+		// if needle has no utf8 codepoints before the null terminating
+		// byte then return haystack
+		if ('\0' == *((const char *)needle)) {
+			return (void *)haystack;
+		}
+		
+		while ('\0' != *h) {
+			const char *maybeMatch = h;
+			const char *n = (const char *)needle;
+			
+			while (*h == *n && (*h != '\0' && *n != '\0')) {
+				n++;
+				h++;
+			}
+			
+			if ('\0' == *n) {
+				// we found the whole utf8 string for needle in haystack at
+				// maybeMatch, so return it
+				return (void *)maybeMatch;
+			} else {
+				// h could be in the middle of an unmatching utf8 codepoint,
+				// so we need to march it on to the next character beginning
+				// starting from the current character
+				h = (const char*)utf8codepoint(maybeMatch, &throwaway_codepoint);
+			}
+		}
+		
+		// no match
+		return utf8_null;
+	}
+	
+	void *utf8casestr(const void *haystack, const void *needle) {
+		const void *h = haystack;
+		
+		// if needle has no utf8 codepoints before the null terminating
+		// byte then return haystack
+		if ('\0' == *((const char *)needle)) {
+			return (void *)haystack;
+		}
+		
+		for (;;) {
+			const void *maybeMatch = h;
+			const void *n = needle;
+			utf8_int32_t h_cp, n_cp;
+			
+			// Get the next code point and track it
+			const void *nextH = h = utf8codepoint(h, &h_cp);
+			n = utf8codepoint(n, &n_cp);
+			
+			while ((0 != h_cp) && (0 != n_cp)) {
+				h_cp = utf8lwrcodepoint(h_cp);
+				n_cp = utf8lwrcodepoint(n_cp);
+				
+				// if we find a mismatch, bail out!
+				if (h_cp != n_cp) {
+					break;
+				}
+				
+				h = utf8codepoint(h, &h_cp);
+				n = utf8codepoint(n, &n_cp);
+			}
+			
+			if (0 == n_cp) {
+				// we found the whole utf8 string for needle in haystack at
+				// maybeMatch, so return it
+				return (void *)maybeMatch;
+			}
+			
+			if (0 == h_cp) {
+				// no match
+				return utf8_null;
+			}
+			
+			// Roll back to the next code point in the haystack to test
+			h = nextH;
+		}
+	}
+	
+	void *utf8valid(const void *str) {
+		const char *s = (const char *)str;
+		
+		while ('\0' != *s) {
+			if (0xf0 == (0xf8 & *s)) {
+				// ensure each of the 3 following bytes in this 4-byte
+				// utf8 codepoint began with 0b10xxxxxx
+				if ((0x80 != (0xc0 & s[1])) || (0x80 != (0xc0 & s[2])) ||
+					(0x80 != (0xc0 & s[3]))) {
+					return (void *)s;
+				}
+				
+				// ensure that our utf8 codepoint ended after 4 bytes
+				if (0x80 == (0xc0 & s[4])) {
+					return (void *)s;
+				}
+				
+				// ensure that the top 5 bits of this 4-byte utf8
+				// codepoint were not 0, as then we could have used
+				// one of the smaller encodings
+				if ((0 == (0x07 & s[0])) && (0 == (0x30 & s[1]))) {
+					return (void *)s;
+				}
+				
+				// 4-byte utf8 code point (began with 0b11110xxx)
+				s += 4;
+			} else if (0xe0 == (0xf0 & *s)) {
+				// ensure each of the 2 following bytes in this 3-byte
+				// utf8 codepoint began with 0b10xxxxxx
+				if ((0x80 != (0xc0 & s[1])) || (0x80 != (0xc0 & s[2]))) {
+					return (void *)s;
+				}
+				
+				// ensure that our utf8 codepoint ended after 3 bytes
+				if (0x80 == (0xc0 & s[3])) {
+					return (void *)s;
+				}
+				
+				// ensure that the top 5 bits of this 3-byte utf8
+				// codepoint were not 0, as then we could have used
+				// one of the smaller encodings
+				if ((0 == (0x0f & s[0])) && (0 == (0x20 & s[1]))) {
+					return (void *)s;
+				}
+				
+				// 3-byte utf8 code point (began with 0b1110xxxx)
+				s += 3;
+			} else if (0xc0 == (0xe0 & *s)) {
+				// ensure the 1 following byte in this 2-byte
+				// utf8 codepoint began with 0b10xxxxxx
+				if (0x80 != (0xc0 & s[1])) {
+					return (void *)s;
+				}
+				
+				// ensure that our utf8 codepoint ended after 2 bytes
+				if (0x80 == (0xc0 & s[2])) {
+					return (void *)s;
+				}
+				
+				// ensure that the top 4 bits of this 2-byte utf8
+				// codepoint were not 0, as then we could have used
+				// one of the smaller encodings
+				if (0 == (0x1e & s[0])) {
+					return (void *)s;
+				}
+				
+				// 2-byte utf8 code point (began with 0b110xxxxx)
+				s += 2;
+			} else if (0x00 == (0x80 & *s)) {
+				// 1-byte ascii (began with 0b0xxxxxxx)
+				s += 1;
+			} else {
+				// we have an invalid 0b1xxxxxxx utf8 code point entry
+				return (void *)s;
+			}
+		}
+		
+		return utf8_null;
+	}
+	
+	void *utf8codepoint(const void *utf8_restrict str,
+						utf8_int32_t *utf8_restrict out_codepoint) {
+		const char *s = (const char *)str;
+		
+		if (0xf0 == (0xf8 & s[0])) {
+			// 4 byte utf8 codepoint
+			*out_codepoint = ((0x07 & s[0]) << 18) | ((0x3f & s[1]) << 12) |
+				((0x3f & s[2]) << 6) | (0x3f & s[3]);
+			s += 4;
+		} else if (0xe0 == (0xf0 & s[0])) {
+			// 3 byte utf8 codepoint
+			*out_codepoint =
+				((0x0f & s[0]) << 12) | ((0x3f & s[1]) << 6) | (0x3f & s[2]);
+			s += 3;
+		} else if (0xc0 == (0xe0 & s[0])) {
+			// 2 byte utf8 codepoint
+			*out_codepoint = ((0x1f & s[0]) << 6) | (0x3f & s[1]);
+			s += 2;
+		} else {
+			// 1 byte utf8 codepoint otherwise
+			*out_codepoint = s[0];
+			s += 1;
+		}
+		
+		return (void *)s;
+	}
+	
+	size_t utf8codepointsize(utf8_int32_t chr) {
+		if (0 == ((utf8_int32_t)0xffffff80 & chr)) {
+			return 1;
+		} else if (0 == ((utf8_int32_t)0xfffff800 & chr)) {
+			return 2;
+		} else if (0 == ((utf8_int32_t)0xffff0000 & chr)) {
+			return 3;
+		} else { // if (0 == ((int)0xffe00000 & chr)) {
+			return 4;
+		}
+	}
+	
+	void *utf8catcodepoint(void *utf8_restrict str, utf8_int32_t chr, size_t n) {
+		char *s = (char *)str;
+		
+		if (0 == ((utf8_int32_t)0xffffff80 & chr)) {
+			// 1-byte/7-bit ascii
+			// (0b0xxxxxxx)
+			if (n < 1) {
+				return utf8_null;
+			}
+			s[0] = (char)chr;
+			s += 1;
+		} else if (0 == ((utf8_int32_t)0xfffff800 & chr)) {
+			// 2-byte/11-bit utf8 code point
+			// (0b110xxxxx 0b10xxxxxx)
+			if (n < 2) {
+				return utf8_null;
+			}
+			s[0] = 0xc0 | (char)(chr >> 6);
+			s[1] = 0x80 | (char)(chr & 0x3f);
+			s += 2;
+		} else if (0 == ((utf8_int32_t)0xffff0000 & chr)) {
+			// 3-byte/16-bit utf8 code point
+			// (0b1110xxxx 0b10xxxxxx 0b10xxxxxx)
+			if (n < 3) {
+				return utf8_null;
+			}
+			s[0] = 0xe0 | (char)(chr >> 12);
+			s[1] = 0x80 | (char)((chr >> 6) & 0x3f);
+			s[2] = 0x80 | (char)(chr & 0x3f);
+			s += 3;
+		} else { // if (0 == ((int)0xffe00000 & chr)) {
+			// 4-byte/21-bit utf8 code point
+			// (0b11110xxx 0b10xxxxxx 0b10xxxxxx 0b10xxxxxx)
+			if (n < 4) {
+				return utf8_null;
+			}
+			s[0] = 0xf0 | (char)(chr >> 18);
+			s[1] = 0x80 | (char)((chr >> 12) & 0x3f);
+			s[2] = 0x80 | (char)((chr >> 6) & 0x3f);
+			s[3] = 0x80 | (char)(chr & 0x3f);
+			s += 4;
+		}
+		
+		return s;
+	}
+	
+	int utf8islower(utf8_int32_t chr) { return chr != utf8uprcodepoint(chr); }
+	
+	int utf8isupper(utf8_int32_t chr) { return chr != utf8lwrcodepoint(chr); }
+	
+	void utf8lwr(void *utf8_restrict str) {
+		void *p, *pn;
+		utf8_int32_t cp;
+		
+		p = (char *)str;
+		pn = utf8codepoint(p, &cp);
+		
+		while (cp != 0) {
+			const utf8_int32_t lwr_cp = utf8lwrcodepoint(cp);
+			const size_t size = utf8codepointsize(lwr_cp);
+			
+			if (lwr_cp != cp) {
+				utf8catcodepoint(p, lwr_cp, size);
+			}
+			
+			p = pn;
+			pn = utf8codepoint(p, &cp);
+		}
+	}
+	
+	void utf8upr(void *utf8_restrict str) {
+		void *p, *pn;
+		utf8_int32_t cp;
+		
+		p = (char *)str;
+		pn = utf8codepoint(p, &cp);
+		
+		while (cp != 0) {
+			const utf8_int32_t lwr_cp = utf8uprcodepoint(cp);
+			const size_t size = utf8codepointsize(lwr_cp);
+			
+			if (lwr_cp != cp) {
+				utf8catcodepoint(p, lwr_cp, size);
+			}
+			
+			p = pn;
+			pn = utf8codepoint(p, &cp);
+		}
+	}
+	
+	utf8_int32_t utf8lwrcodepoint(utf8_int32_t cp) {
+		if (((0x0041 <= cp) && (0x005a >= cp)) ||
+			((0x00c0 <= cp) && (0x00d6 >= cp)) ||
+			((0x00d8 <= cp) && (0x00de >= cp)) ||
+			((0x0391 <= cp) && (0x03a1 >= cp)) ||
+			((0x03a3 <= cp) && (0x03ab >= cp))) {
+			cp += 32;
+		} else if (((0x0100 <= cp) && (0x012f >= cp)) ||
+				   ((0x0132 <= cp) && (0x0137 >= cp)) ||
+				   ((0x014a <= cp) && (0x0177 >= cp)) ||
+				   ((0x0182 <= cp) && (0x0185 >= cp)) ||
+				   ((0x01a0 <= cp) && (0x01a5 >= cp)) ||
+				   ((0x01de <= cp) && (0x01ef >= cp)) ||
+				   ((0x01f8 <= cp) && (0x021f >= cp)) ||
+				   ((0x0222 <= cp) && (0x0233 >= cp)) ||
+				   ((0x0246 <= cp) && (0x024f >= cp)) ||
+				   ((0x03d8 <= cp) && (0x03ef >= cp))) {
+			cp |= 0x1;
+		} else if (((0x0139 <= cp) && (0x0148 >= cp)) ||
+				   ((0x0179 <= cp) && (0x017e >= cp)) ||
+				   ((0x01af <= cp) && (0x01b0 >= cp)) ||
+				   ((0x01b3 <= cp) && (0x01b6 >= cp)) ||
+				   ((0x01cd <= cp) && (0x01dc >= cp))) {
+			cp += 1;
+			cp &= ~0x1;
+		} else {
+			switch (cp) {
+				default: break;
+				case 0x0178: cp = 0x00ff; break;
+				case 0x0243: cp = 0x0180; break;
+				case 0x018e: cp = 0x01dd; break;
+				case 0x023d: cp = 0x019a; break;
+				case 0x0220: cp = 0x019e; break;
+				case 0x01b7: cp = 0x0292; break;
+				case 0x01c4: cp = 0x01c6; break;
+				case 0x01c7: cp = 0x01c9; break;
+				case 0x01ca: cp = 0x01cc; break;
+				case 0x01f1: cp = 0x01f3; break;
+				case 0x01f7: cp = 0x01bf; break;
+				case 0x0187: cp = 0x0188; break;
+				case 0x018b: cp = 0x018c; break;
+				case 0x0191: cp = 0x0192; break;
+				case 0x0198: cp = 0x0199; break;
+				case 0x01a7: cp = 0x01a8; break;
+				case 0x01ac: cp = 0x01ad; break;
+				case 0x01af: cp = 0x01b0; break;
+				case 0x01b8: cp = 0x01b9; break;
+				case 0x01bc: cp = 0x01bd; break;
+				case 0x01f4: cp = 0x01f5; break;
+				case 0x023b: cp = 0x023c; break;
+				case 0x0241: cp = 0x0242; break;
+				case 0x03fd: cp = 0x037b; break;
+				case 0x03fe: cp = 0x037c; break;
+				case 0x03ff: cp = 0x037d; break;
+				case 0x037f: cp = 0x03f3; break;
+				case 0x0386: cp = 0x03ac; break;
+				case 0x0388: cp = 0x03ad; break;
+				case 0x0389: cp = 0x03ae; break;
+				case 0x038a: cp = 0x03af; break;
+				case 0x038c: cp = 0x03cc; break;
+				case 0x038e: cp = 0x03cd; break;
+				case 0x038f: cp = 0x03ce; break;
+				case 0x0370: cp = 0x0371; break;
+				case 0x0372: cp = 0x0373; break;
+				case 0x0376: cp = 0x0377; break;
+				case 0x03f4: cp = 0x03d1; break;
+				case 0x03cf: cp = 0x03d7; break;
+				case 0x03f9: cp = 0x03f2; break;
+				case 0x03f7: cp = 0x03f8; break;
+				case 0x03fa: cp = 0x03fb; break;
+			};
+		}
+		
+		return cp;
+	}
+	
+	utf8_int32_t utf8uprcodepoint(utf8_int32_t cp) {
+		if (((0x0061 <= cp) && (0x007a >= cp)) ||
+			((0x00e0 <= cp) && (0x00f6 >= cp)) ||
+			((0x00f8 <= cp) && (0x00fe >= cp)) ||
+			((0x03b1 <= cp) && (0x03c1 >= cp)) ||
+			((0x03c3 <= cp) && (0x03cb >= cp))) {
+			cp -= 32;
+		} else if (((0x0100 <= cp) && (0x012f >= cp)) ||
+				   ((0x0132 <= cp) && (0x0137 >= cp)) ||
+				   ((0x014a <= cp) && (0x0177 >= cp)) ||
+				   ((0x0182 <= cp) && (0x0185 >= cp)) ||
+				   ((0x01a0 <= cp) && (0x01a5 >= cp)) ||
+				   ((0x01de <= cp) && (0x01ef >= cp)) ||
+				   ((0x01f8 <= cp) && (0x021f >= cp)) ||
+				   ((0x0222 <= cp) && (0x0233 >= cp)) ||
+				   ((0x0246 <= cp) && (0x024f >= cp)) ||
+				   ((0x03d8 <= cp) && (0x03ef >= cp))) {
+			cp &= ~0x1;
+		} else if (((0x0139 <= cp) && (0x0148 >= cp)) ||
+				   ((0x0179 <= cp) && (0x017e >= cp)) ||
+				   ((0x01af <= cp) && (0x01b0 >= cp)) ||
+				   ((0x01b3 <= cp) && (0x01b6 >= cp)) ||
+				   ((0x01cd <= cp) && (0x01dc >= cp))) {
+			cp -= 1;
+			cp |= 0x1;
+		} else {
+			switch (cp) {
+				default: break;
+				case 0x00ff: cp = 0x0178; break;
+				case 0x0180: cp = 0x0243; break;
+				case 0x01dd: cp = 0x018e; break;
+				case 0x019a: cp = 0x023d; break;
+				case 0x019e: cp = 0x0220; break;
+				case 0x0292: cp = 0x01b7; break;
+				case 0x01c6: cp = 0x01c4; break;
+				case 0x01c9: cp = 0x01c7; break;
+				case 0x01cc: cp = 0x01ca; break;
+				case 0x01f3: cp = 0x01f1; break;
+				case 0x01bf: cp = 0x01f7; break;
+				case 0x0188: cp = 0x0187; break;
+				case 0x018c: cp = 0x018b; break;
+				case 0x0192: cp = 0x0191; break;
+				case 0x0199: cp = 0x0198; break;
+				case 0x01a8: cp = 0x01a7; break;
+				case 0x01ad: cp = 0x01ac; break;
+				case 0x01b0: cp = 0x01af; break;
+				case 0x01b9: cp = 0x01b8; break;
+				case 0x01bd: cp = 0x01bc; break;
+				case 0x01f5: cp = 0x01f4; break;
+				case 0x023c: cp = 0x023b; break;
+				case 0x0242: cp = 0x0241; break;
+				case 0x037b: cp = 0x03fd; break;
+				case 0x037c: cp = 0x03fe; break;
+				case 0x037d: cp = 0x03ff; break;
+				case 0x03f3: cp = 0x037f; break;
+				case 0x03ac: cp = 0x0386; break;
+				case 0x03ad: cp = 0x0388; break;
+				case 0x03ae: cp = 0x0389; break;
+				case 0x03af: cp = 0x038a; break;
+				case 0x03cc: cp = 0x038c; break;
+				case 0x03cd: cp = 0x038e; break;
+				case 0x03ce: cp = 0x038f; break;
+				case 0x0371: cp = 0x0370; break;
+				case 0x0373: cp = 0x0372; break;
+				case 0x0377: cp = 0x0376; break;
+				case 0x03d1: cp = 0x03f4; break;
+				case 0x03d7: cp = 0x03cf; break;
+				case 0x03f2: cp = 0x03f9; break;
+				case 0x03f8: cp = 0x03f7; break;
+				case 0x03fb: cp = 0x03fa; break;
+			};
+		}
+		
+		return cp;
+	}
+	
+#undef utf8_restrict
+#undef utf8_null
+	
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#if defined(__clang__)
+#pragma clang diagnostic pop
+#endif
+
+#endif // SHEREDOM_UTF8_H_INCLUDED
diff --git a/src/input.c b/src/input.c
new file mode 100644
index 0000000..40f7b74
--- /dev/null
+++ b/src/input.c
@@ -0,0 +1,237 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+inline bool is_left_down(mouse_input *input)
+{
+	return input->left_state & MOUSE_DOWN;
+}
+
+inline bool is_left_released(mouse_input *input)
+{
+	return input->left_state & MOUSE_RELEASE;
+}
+
+inline bool is_left_clicked(mouse_input *input)
+{
+	return input->left_state & MOUSE_CLICK;
+}
+
+inline bool is_left_double_clicked(mouse_input *input)
+{
+	return input->left_state & MOUSE_DOUBLE_CLICK;
+}
+
+inline bool is_right_down(mouse_input *input)
+{
+	return input->right_state & MOUSE_DOWN;
+}
+
+inline bool is_right_released(mouse_input *input)
+{
+	return input->right_state & MOUSE_RELEASE;
+}
+
+inline bool is_right_clicked(mouse_input *input)
+{
+	return input->right_state & MOUSE_CLICK;
+}
+
+inline bool keyboard_is_key_down(keyboard_input *keyboard, s16 key)
+{
+	return keyboard->keys[key];
+}
+
+inline bool keyboard_is_key_pressed(keyboard_input *keyboard, s16 key)
+{
+	return keyboard->input_keys[key];
+}
+
+inline void keyboard_set_input_text(keyboard_input *keyboard, char *text)
+{
+	string_copyn(keyboard->input_text, text, MAX_INPUT_LENGTH);
+	u32 len = utf8len(keyboard->input_text);
+	keyboard->cursor = len;
+	keyboard->input_text_len = len;
+}
+
+mouse_input mouse_input_create()
+{
+	mouse_input mouse;
+	mouse.x = -1;
+	mouse.y = -1;
+	mouse.move_x = 0;
+	mouse.move_y = 0;
+	mouse.left_state = 0;
+	mouse.right_state = 0;
+	
+	return mouse;
+}
+
+keyboard_input keyboard_input_create()
+{
+	keyboard_input keyboard;
+	keyboard.modifier_state = 0;
+	keyboard.input_text = mem_alloc(MAX_INPUT_LENGTH);
+	keyboard.input_text[0] = '\0';
+	keyboard.take_input = false;
+	keyboard.cursor = 0;
+	keyboard.input_text_len = 0;
+	keyboard.input_mode = INPUT_FULL;
+	keyboard.has_selection = false;
+	keyboard.selection_begin_offset = 0;
+	keyboard.selection_length = 0;
+	memset(keyboard.keys, 0, MAX_KEYCODE);
+	
+	return keyboard;
+}
+
+void keyboard_set_input_mode(keyboard_input *keyboard, keyboard_input_mode mode)
+{
+	keyboard->input_mode = mode;
+}
+
+inline void keyboard_input_destroy(keyboard_input *keyboard)
+{
+	mem_free(keyboard->input_text);
+}
+
+static void keyboard_handle_input_copy_and_paste(platform_window *window, keyboard_input *keyboard)
+{
+	bool is_lctrl_down = keyboard->keys[KEY_LEFT_CONTROL];
+	
+	if (is_lctrl_down && keyboard_is_key_pressed(keyboard, KEY_V))
+	{
+		char buf[MAX_INPUT_LENGTH];
+		bool result = platform_get_clipboard(window, buf);
+		
+		if (keyboard->input_mode == INPUT_NUMERIC && !is_string_numeric(buf))
+		{
+			return;
+		}
+		
+		if (keyboard->has_selection)
+		{
+			keyboard->cursor = keyboard->selection_begin_offset;
+			utf8_str_remove_range(keyboard->input_text, keyboard->selection_begin_offset,
+								  keyboard->selection_begin_offset + keyboard->selection_length);
+			keyboard->has_selection = false;
+			keyboard->text_changed = true;
+		}
+		
+		if (result)
+		{
+			s32 len = utf8len(buf);
+			utf8_str_insert_utf8str(keyboard->input_text, keyboard->cursor, buf);
+			
+			keyboard->cursor += len;
+			keyboard->input_text_len += len;
+			keyboard->text_changed = true;
+		}
+	}
+	else if (is_lctrl_down && keyboard_is_key_pressed(keyboard, KEY_C))
+	{
+		char buffer[MAX_INPUT_LENGTH];
+		utf8_str_copy_range(keyboard->input_text, 
+							keyboard->selection_begin_offset,
+							keyboard->selection_begin_offset+keyboard->selection_length,
+							buffer);
+		
+		if (!string_equals(buffer, ""))
+			platform_set_clipboard(window, buffer);
+	}
+}
+
+void keyboard_handle_input_string(platform_window *window, keyboard_input *keyboard, char *ch)
+{
+	keyboard_handle_input_copy_and_paste(window, keyboard);
+	
+	bool is_lctrl_down = keyboard->keys[KEY_LEFT_CONTROL];
+	
+	if (ch)
+	{
+		if (keyboard->input_text_len < MAX_INPUT_LENGTH && !is_lctrl_down)
+		{
+			if (keyboard->has_selection)
+			{
+				keyboard->cursor = keyboard->selection_begin_offset;
+				utf8_str_remove_range(keyboard->input_text, keyboard->selection_begin_offset,
+									  keyboard->selection_begin_offset + keyboard->selection_length);
+				keyboard->has_selection = false;
+				keyboard->text_changed = true;
+			}
+			
+			if (keyboard->input_text_len)
+			{
+				utf8_str_insert_at(keyboard->input_text, keyboard->cursor, *ch);
+				keyboard->text_changed = true;
+			}
+			else
+			{
+				string_appendn(keyboard->input_text, ch, MAX_INPUT_LENGTH);
+				keyboard->text_changed = true;
+			}
+			
+			keyboard->cursor++;
+			keyboard->input_text_len = utf8len(keyboard->input_text);
+		}
+	}
+	else
+	{
+		bool is_lctrl_down = keyboard->keys[KEY_LEFT_CONTROL];
+		
+		// cursor movement
+		if (keyboard_is_key_down(keyboard, KEY_LEFT) && keyboard->cursor > 0)
+		{
+			if (is_lctrl_down)
+				keyboard->cursor = 0;
+			else
+				keyboard->cursor--;
+		}
+		if (keyboard_is_key_down(keyboard, KEY_RIGHT) && keyboard->cursor < keyboard->input_text_len)
+		{
+			if (is_lctrl_down)
+				keyboard->cursor = utf8len(keyboard->input_text);
+			else
+				keyboard->cursor++;
+		}
+	}
+	
+	if (keyboard_is_key_down(keyboard, KEY_BACKSPACE))
+	{
+		bool is_lctrl_down = keyboard->keys[KEY_LEFT_CONTROL];
+		
+		if (keyboard->has_selection)
+		{
+			utf8_str_remove_range(keyboard->input_text, keyboard->selection_begin_offset,
+								  keyboard->selection_begin_offset + keyboard->selection_length);
+			keyboard->has_selection = false;
+			keyboard->text_changed = true;
+			
+			if (keyboard->selection_begin_offset < keyboard->cursor)
+			{
+				keyboard->cursor -= keyboard->selection_length-1;
+			}
+		}
+		else if (is_lctrl_down)
+		{
+			for (s32 i = 0; i < keyboard->cursor; i++)
+			{
+				utf8_str_remove_at(keyboard->input_text, 0);
+			}
+			keyboard->cursor = 0;
+			keyboard->text_changed = true;
+		}
+		else if (keyboard->cursor > 0)
+		{
+			utf8_str_remove_at(keyboard->input_text, keyboard->cursor-1);
+			
+			keyboard->cursor--;
+			keyboard->text_changed = true;
+		}
+		
+		keyboard->input_text_len = utf8len(keyboard->input_text);
+	}
+}
\ No newline at end of file
diff --git a/src/input.h b/src/input.h
new file mode 100644
index 0000000..21f4df9
--- /dev/null
+++ b/src/input.h
@@ -0,0 +1,224 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_INPUT
+#define INCLUDE_INPUT
+
+/* The unknown key */
+#define KEY_UNKNOWN -1
+
+#define MOUSE_OFFSCREEN 32767
+
+/* Printable keys */
+#define KEY_SPACE              32
+#define KEY_APOSTROPHE         39  /* ' */
+#define KEY_COMMA              44  /* , */
+#define KEY_MINUS              45  /* - */
+#define KEY_PERIOD             46  /* . */
+#define KEY_SLASH              47  /* / */
+#define KEY_0                  48
+#define KEY_1                  49
+#define KEY_2                  50
+#define KEY_3                  51
+#define KEY_4                  52
+#define KEY_5                  53
+#define KEY_6                  54
+#define KEY_7                  55
+#define KEY_8                  56
+#define KEY_9                  57
+#define KEY_SEMICOLON          59  /* ; */
+#define KEY_EQUAL              61  /* = */
+#define KEY_A                  65
+#define KEY_B                  66
+#define KEY_C                  67
+#define KEY_D                  68
+#define KEY_E                  69
+#define KEY_F                  70
+#define KEY_G                  71
+#define KEY_H                  72
+#define KEY_I                  73
+#define KEY_J                  74
+#define KEY_K                  75
+#define KEY_L                  76
+#define KEY_M                  77
+#define KEY_N                  78
+#define KEY_O                  79
+#define KEY_P                  80
+#define KEY_Q                  81
+#define KEY_R                  82
+#define KEY_S                  83
+#define KEY_T                  84
+#define KEY_U                  85
+#define KEY_V                  86
+#define KEY_W                  87
+#define KEY_X                  88
+#define KEY_Y                  89
+#define KEY_Z                  90
+#define KEY_LEFT_BRACKET       91  /* [ */
+#define KEY_BACKSLASH          92  /* \ */
+#define KEY_RIGHT_BRACKET      93  /* ] */
+#define KEY_GRAVE_ACCENT       96  /* ` */
+#define KEY_WORLD_1            161 /* non-US #1 */
+#define KEY_WORLD_2            162 /* non-US #2 */
+
+/* Function keys */
+#define KEY_ESCAPE             256
+#define KEY_ENTER              257
+#define KEY_TAB                258
+#define KEY_BACKSPACE          259
+#define KEY_INSERT             260
+#define KEY_DELETE             261
+#define KEY_RIGHT              262
+#define KEY_LEFT               263
+#define KEY_DOWN               264
+#define KEY_UP                 265
+#define KEY_PAGE_UP            266
+#define KEY_PAGE_DOWN          267
+#define KEY_HOME               268
+#define KEY_END                269
+#define KEY_CAPS_LOCK          280
+#define KEY_SCROLL_LOCK        281
+#define KEY_NUM_LOCK           282
+#define KEY_PRINT_SCREEN       283
+#define KEY_PAUSE              284
+#define KEY_F1                 290
+#define KEY_F2                 291
+#define KEY_F3                 292
+#define KEY_F4                 293
+#define KEY_F5                 294
+#define KEY_F6                 295
+#define KEY_F7                 296
+#define KEY_F8                 297
+#define KEY_F9                 298
+#define KEY_F10                299
+#define KEY_F11                300
+#define KEY_F12                301
+#define KEY_F13                302
+#define KEY_F14                303
+#define KEY_F15                304
+#define KEY_F16                305
+#define KEY_F17                306
+#define KEY_F18                307
+#define KEY_F19                308
+#define KEY_F20                309
+#define KEY_F21                310
+#define KEY_F22                311
+#define KEY_F23                312
+#define KEY_F24                313
+#define KEY_F25                314
+#define KEY_KP_0               320
+#define KEY_KP_1               321
+#define KEY_KP_2               322
+#define KEY_KP_3               323
+#define KEY_KP_4               324
+#define KEY_KP_5               325
+#define KEY_KP_6               326
+#define KEY_KP_7               327
+#define KEY_KP_8               328
+#define KEY_KP_9               329
+#define KEY_KP_DECIMAL         330
+#define KEY_KP_DIVIDE          331
+#define KEY_KP_MULTIPLY        332
+#define KEY_KP_SUBTRACT        333
+#define KEY_KP_ADD             334
+#define KEY_KP_ENTER           335
+#define KEY_KP_EQUAL           336
+#define KEY_LEFT_SHIFT         340
+#define KEY_LEFT_CONTROL       341
+#define KEY_LEFT_ALT           342
+#define KEY_LEFT_SUPER         343
+#define KEY_RIGHT_SHIFT        344
+#define KEY_RIGHT_CONTROL      345
+#define KEY_RIGHT_ALT          346
+#define KEY_RIGHT_SUPER        347
+#define KEY_MENU               348
+
+#define KEY_LAST KEY_MENU
+
+#define MAX_KEYCODE 512
+
+#define MOUSE_DOWN (1 << 1)
+#define MOUSE_RELEASE (1 << 2)
+#define MOUSE_DOUBLE_CLICK (1 << 3)
+#define MOUSE_CLICK (1 << 4)
+
+#define SCROLL_UP 1
+#define SCROLL_DOWN -1
+
+
+// should be max path length
+#ifdef OS_LINUX
+#define MAX_INPUT_LENGTH 4096+1
+#define MAX_PATH_LENGTH 255+1
+#endif
+
+#ifdef OS_WIN
+#define MAX_INPUT_LENGTH 4096+1
+#define MAX_PATH_LENGTH 259+1
+#endif
+
+typedef struct t_mouse_input
+{
+	s16 x;
+	s16 y;
+	s16 move_x;
+	s16 move_y;
+	s16 total_move_x;
+	s16 total_move_y;
+	s8 left_state;
+	s8 right_state;
+	s8 scroll_state;
+} mouse_input;
+
+typedef enum t_keyboard_input_mode
+{
+	INPUT_NUMERIC,
+	INPUT_FULL,
+} keyboard_input_mode;
+
+typedef struct t_keyboard_input
+{
+	keyboard_input_mode input_mode;
+	int modifier_state;
+	bool take_input;
+	u32 cursor;
+	
+	// input
+	bool text_changed; // is set when text is pasted in, incase the new text is the same length as the old text so we know to store a new undo entry
+	bool has_selection;
+	s32 selection_begin_offset;
+	s32 selection_length;
+	char *input_text;
+	// input
+	
+	s32 input_text_len;
+	bool keys[MAX_KEYCODE];
+	bool input_keys[MAX_KEYCODE];
+} keyboard_input;
+
+int keycode_map[MAX_KEYCODE];
+
+bool is_left_down(mouse_input *input);
+bool is_left_released(mouse_input *input);
+bool is_left_clicked(mouse_input *input);
+bool is_left_double_clicked(mouse_input *input);
+bool is_right_down(mouse_input *input);
+bool is_right_released(mouse_input *input);
+bool is_right_clicked(mouse_input *input);
+
+bool keyboard_is_key_down(keyboard_input *keyboard, s16 key);
+bool keyboard_is_key_pressed(keyboard_input *keyboard, s16 key);
+void keyboard_set_input_text(keyboard_input *keyboard, char *text);
+void keyboard_set_input_mode(keyboard_input *keyboard, keyboard_input_mode mode);
+
+typedef struct t_platform_window platform_window;
+void keyboard_handle_input_string(platform_window *window, keyboard_input *keyboard, char *text);
+
+mouse_input mouse_input_create();
+keyboard_input keyboard_input_create();
+void keyboard_input_destroy(keyboard_input *keyboard);
+
+#endif
\ No newline at end of file
diff --git a/src/languages.h b/src/languages.h
new file mode 100644
index 0000000..a079141
--- /dev/null
+++ b/src/languages.h
@@ -0,0 +1,22 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#define LANGUAGE_CODE_SIZE 3
+
+typedef struct t_language
+{
+	char code[3];
+	char fullname[40];
+} language;
+
+
+language global_langues[] = {
+	{"AD","Andorra"},
+	{"AE","United Arab Emirates"},
+	{"AF","Afghanistan"},
+};
+
+#define COUNTRY_CODE_COUNT (sizeof(global_langues)/sizeof(language))
\ No newline at end of file
diff --git a/src/linux/platform.c b/src/linux/platform.c
new file mode 100644
index 0000000..e650097
--- /dev/null
+++ b/src/linux/platform.c
@@ -0,0 +1,1592 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#include <locale.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <X11/X.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/Xrandr.h>
+#include <X11/Xatom.h>
+#include <time.h>
+#include <X11/XKBlib.h>
+#include <unistd.h>
+#include <limits.h>
+#include <dirent.h> 
+#include <errno.h>
+#include <dlfcn.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <X11/cursorfont.h>
+
+#define GET_ATOM(X) window.X = XInternAtom(window.display, #X, False)
+
+struct t_platform_window
+{
+	Display *display;
+	Window parent;
+	XVisualInfo *visual_info;
+	Colormap cmap;
+	Window window;
+	GLXContext gl_context;
+	XWindowAttributes window_attributes;
+	XEvent event;
+	char *clipboard_str;
+	s32 clipboard_strlen;
+	
+	Atom xdnd_req;
+	Atom xdnd_source;
+	Atom XdndEnter;
+	Atom XdndPosition;
+	Atom XdndStatus;
+	Atom XdndTypeList;
+	Atom XdndActionCopy;
+	Atom XdndDrop;
+	Atom XdndFinished;
+	Atom XdndSelection;
+	Atom XdndLeave;
+	Atom quit;
+	Atom PRIMARY;
+	Atom CLIPBOARD;
+	Atom UTF8_STRING;
+	Atom COMPOUND_STRING;
+	Atom TARGETS;
+	Atom MULTIPLE;
+	Atom _NET_WM_STATE;
+	
+	// shared window properties
+	s32 width;
+	s32 height;
+	bool is_open;
+	bool has_focus;
+	cursor_type curr_cursor_type;
+	cursor_type next_cursor_type;
+};
+
+bool platform_get_clipboard(platform_window *window, char *buffer)
+{
+	char *result;
+	unsigned long ressize, restail;
+	int resbits;
+	Atom bufid = XInternAtom(window->display, "CLIPBOARD", False),
+	fmtid = XInternAtom(window->display, "UTF8_STRING", False),
+	propid = XInternAtom(window->display, "XSEL_DATA", False),
+	incrid = XInternAtom(window->display, "INCR", False);
+	XEvent event;
+	
+	if(window->CLIPBOARD != None) {
+		
+		if (settings_window && XGetSelectionOwner(window->display, window->CLIPBOARD) == settings_window->window)
+		{
+			snprintf(buffer, MAX_INPUT_LENGTH, "%s", settings_window->clipboard_str);
+			return true;
+		}
+		else if (XGetSelectionOwner(window->display, window->CLIPBOARD) == 
+				 main_window->window)
+		{
+			snprintf(buffer, MAX_INPUT_LENGTH, "%s", main_window->clipboard_str);
+			return true;
+		}
+	}
+	
+	XConvertSelection(window->display, bufid, fmtid, propid, window->window, CurrentTime);
+	do {
+		XNextEvent(window->display, &event);
+	} while (event.type != SelectionNotify || event.xselection.selection != bufid);
+	
+	if (event.xselection.property)
+	{
+		XGetWindowProperty(window->display, window->window, propid, 0, LONG_MAX/4, False, AnyPropertyType,
+						   &fmtid, &resbits, &ressize, &restail, (unsigned char**)&result);
+		
+		if (fmtid == incrid)
+			printf("Buffer is too large and INCR reading is not implemented yet.\n");
+		else
+			snprintf(buffer, MAX_INPUT_LENGTH, "%s", result);
+		
+		XFree(result);
+		return True;
+	}
+	else // request failed, e.g. owner can't convert to the target format
+		return False;
+}
+
+bool platform_set_clipboard(platform_window *window, char *buffer)
+{
+	if (buffer)
+	{
+		if(window->CLIPBOARD != None && XGetSelectionOwner(window->display, window->CLIPBOARD) != window->window) {
+			XSetSelectionOwner(window->display, window->CLIPBOARD, window->window, CurrentTime);
+		}
+		
+		window->clipboard_strlen = strlen(buffer);
+		if(!window->clipboard_str) {
+			window->clipboard_str = mem_alloc(window->clipboard_strlen+1);
+		} else {
+			window->clipboard_str = mem_realloc(window->clipboard_str, window->clipboard_strlen+1);
+		}
+		string_copyn(window->clipboard_str, buffer, window->clipboard_strlen+1);
+		
+		return true;
+	}
+	
+	return false;
+}
+
+void platform_create_config_directory()
+{
+	char *env = getenv("HOME");
+	char tmp[PATH_MAX];
+	snprintf(tmp, PATH_MAX, "%s%s", env, "/.config/text-search");
+	
+	if (!platform_directory_exists(tmp))
+	{
+		mkdir(tmp, S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
+	}
+}
+
+char* get_config_save_location(char *buffer)
+{
+	char *env = getenv("HOME");
+	snprintf(buffer, PATH_MAX, "%s%s", env, "/.config/text-search/config.txt");
+	return buffer;
+}
+
+inline void platform_set_cursor(platform_window *window, cursor_type type)
+{
+	if (window->next_cursor_type != type)
+	{
+		window->next_cursor_type = type;
+	}
+}
+
+bool is_platform_in_darkmode()
+{
+	return false;
+}
+
+bool get_active_directory(char *buffer)
+{
+	char cwd[PATH_MAX];
+	if (getcwd(cwd, sizeof(cwd)) != NULL) {
+		string_copyn(buffer, cwd, PATH_MAX);
+	} else {
+		return false;
+	}
+	return true;
+}
+
+bool set_active_directory(char *path)
+{
+	return !chdir(path);
+}
+
+bool platform_write_file_content(char *path, const char *mode, char *buffer, s32 len)
+{
+	bool result = false;
+	
+	FILE *file = fopen(path, mode);
+	
+	if (!file)
+	{
+		goto done_failure;
+	}
+	else
+	{
+		fprintf(file, "%s", buffer);
+	}
+	
+	//done:
+	fclose(file);
+	done_failure:
+	return result;
+}
+
+void platform_window_set_title(platform_window *window, char *name)
+{
+	Atom WM_NAME = XInternAtom(window->display, "WM_NAME", False);
+	Atom _NET_WM_NAME = XInternAtom(window->display, "_NET_WM_NAME", False);
+	Atom _NET_WM_ICON_NAME = XInternAtom(window->display, "_NET_WM_ICON_NAME", False);
+	
+	char *list[1] = { (char *) name };
+	XTextProperty property;
+	
+	XStoreName(window->display, window->window, name);
+	
+	Xutf8TextListToTextProperty(window->display, list, 1, XUTF8StringStyle,
+								&property);
+	XSetTextProperty(window->display, window->window, &property, WM_NAME);
+	XSetTextProperty(window->display, window->window, &property, _NET_WM_NAME);
+	XSetTextProperty(window->display, window->window, &property, XA_WM_NAME);
+	XSetTextProperty(window->display, window->window, &property, _NET_WM_ICON_NAME);
+	XFree(property.value);
+}
+
+bool platform_file_exists(char *path)
+{
+	if(access(path, F_OK) != -1) {
+		return 1;
+	}
+	
+	return 0;
+}
+
+bool platform_directory_exists(char *path)
+{
+	DIR* dir = opendir(path);
+	if (dir) {
+		/* Directory exists. */
+		closedir(dir);
+		return 1;
+	} else if (ENOENT == errno) {
+		return 0; // does not exist
+	} else {
+		return 0; // error opening dir
+	}
+}
+
+file_content platform_read_file_content(char *path, const char *mode)
+{
+	file_content result;
+	result.content = 0;
+	result.content_length = 0;
+	result.file_error = 0;
+	
+	FILE *file = fopen(path, mode);
+	
+	if (!file) 
+	{
+		if (errno == EMFILE)
+			result.file_error = FILE_ERROR_TOO_MANY_OPEN_FILES_PROCESS;
+		else if (errno == ENFILE)
+			result.file_error = FILE_ERROR_TOO_MANY_OPEN_FILES_SYSTEM;
+		else if (errno == EACCES)
+			result.file_error = FILE_ERROR_NO_ACCESS;
+		else if (errno == EPERM)
+			result.file_error = FILE_ERROR_NO_ACCESS;
+		else if (errno == ENOENT)
+			result.file_error = FILE_ERROR_NOT_FOUND;
+		else if (errno == ECONNABORTED)
+			result.file_error = FILE_ERROR_CONNECTION_ABORTED;
+		else if (errno == ECONNREFUSED)
+			result.file_error = FILE_ERROR_CONNECTION_REFUSED;
+		else if (errno == ENETDOWN)
+			result.file_error = FILE_ERROR_NETWORK_DOWN;
+		else if (errno == EREMOTEIO)
+			result.file_error = FILE_ERROR_REMOTE_IO_ERROR;
+		else if (errno == ESTALE)
+			result.file_error = FILE_ERROR_STALE;
+		else
+		{
+			result.file_error = FILE_ERROR_GENERIC;
+			printf("ERROR: %d\n", errno);
+		}
+		
+		goto done_failure;
+	}
+	
+	fseek(file, 0 , SEEK_END);
+	int length = ftell(file);
+	fseek(file, 0, SEEK_SET);
+	
+	s32 length_to_alloc = length+1;
+	
+	result.content = mem_alloc(length_to_alloc);
+	if (!result.content) goto done;
+	
+	memset(result.content, 0, length);
+	s32 read_result = fread(result.content, 1, length, file);
+	if (read_result == 0 && length != 0)
+	{
+		mem_free(result.content);
+		result.content = 0;
+		return result;
+	}
+	
+	result.content_length = read_result;
+	
+	((char*)result.content)[length] = 0;
+	
+	done:
+	fclose(file);
+	done_failure:
+	return result;
+}
+
+inline void platform_destroy_file_content(file_content *content)
+{
+	assert(content);
+	mem_free(content->content);
+}
+
+// Translate an X11 key code to a GLFW key code.
+//
+static s32 translate_keycode(platform_window *window, s32 scancode)
+{
+    s32 keySym;
+	
+    // Valid key code range is  [8,255], according to the Xlib manual
+	
+    if (1)
+    {
+        // Try secondary keysym, for numeric keypad keys
+        // Note: This way we always force "NumLock = ON", which is intentional
+        // since the returned key code should correspond to a physical
+        // location.
+        keySym = XkbKeycodeToKeysym(window->display, scancode, 0, 1);
+        switch (keySym)
+        {
+            case XK_KP_0:           return KEY_KP_0;
+            case XK_KP_1:           return KEY_KP_1;
+            case XK_KP_2:           return KEY_KP_2;
+            case XK_KP_3:           return KEY_KP_3;
+            case XK_KP_4:           return KEY_KP_4;
+            case XK_KP_5:           return KEY_KP_5;
+            case XK_KP_6:           return KEY_KP_6;
+            case XK_KP_7:           return KEY_KP_7;
+            case XK_KP_8:           return KEY_KP_8;
+            case XK_KP_9:           return KEY_KP_9;
+            case XK_KP_Separator:
+            case XK_KP_Decimal:     return KEY_KP_DECIMAL;
+            case XK_KP_Equal:       return KEY_KP_EQUAL;
+            case XK_KP_Enter:       return KEY_KP_ENTER;
+            default:                break;
+        }
+		
+        // Now try primary keysym for function keys (non-printable keys)
+        // These should not depend on the current keyboard layout
+        keySym = XkbKeycodeToKeysym(window->display, scancode, 0, 0);
+    }
+	
+    switch (keySym)
+    {
+        case XK_Escape:         return KEY_ESCAPE;
+        case XK_Tab:            return KEY_TAB;
+        case XK_Shift_L:        return KEY_LEFT_SHIFT;
+        case XK_Shift_R:        return KEY_RIGHT_SHIFT;
+        case XK_Control_L:      return KEY_LEFT_CONTROL;
+        case XK_Control_R:      return KEY_RIGHT_CONTROL;
+        case XK_Meta_L:
+        case XK_Alt_L:          return KEY_LEFT_ALT;
+        case XK_Mode_switch: // Mapped to Alt_R on many keyboards
+        case XK_ISO_Level3_Shift: // AltGr on at least some machines
+        case XK_Meta_R:
+        case XK_Alt_R:          return KEY_RIGHT_ALT;
+        case XK_Super_L:        return KEY_LEFT_SUPER;
+        case XK_Super_R:        return KEY_RIGHT_SUPER;
+        case XK_Menu:           return KEY_MENU;
+        case XK_Num_Lock:       return KEY_NUM_LOCK;
+        case XK_Caps_Lock:      return KEY_CAPS_LOCK;
+        case XK_Print:          return KEY_PRINT_SCREEN;
+        case XK_Scroll_Lock:    return KEY_SCROLL_LOCK;
+        case XK_Pause:          return KEY_PAUSE;
+        case XK_Delete:         return KEY_DELETE;
+        case XK_BackSpace:      return KEY_BACKSPACE;
+        case XK_Return:         return KEY_ENTER;
+        case XK_Home:           return KEY_HOME;
+        case XK_End:            return KEY_END;
+        case XK_Page_Up:        return KEY_PAGE_UP;
+        case XK_Page_Down:      return KEY_PAGE_DOWN;
+        case XK_Insert:         return KEY_INSERT;
+        case XK_Left:           return KEY_LEFT;
+        case XK_Right:          return KEY_RIGHT;
+        case XK_Down:           return KEY_DOWN;
+        case XK_Up:             return KEY_UP;
+        case XK_F1:             return KEY_F1;
+        case XK_F2:             return KEY_F2;
+        case XK_F3:             return KEY_F3;
+        case XK_F4:             return KEY_F4;
+        case XK_F5:             return KEY_F5;
+        case XK_F6:             return KEY_F6;
+        case XK_F7:             return KEY_F7;
+        case XK_F8:             return KEY_F8;
+        case XK_F9:             return KEY_F9;
+        case XK_F10:            return KEY_F10;
+        case XK_F11:            return KEY_F11;
+        case XK_F12:            return KEY_F12;
+        case XK_F13:            return KEY_F13;
+        case XK_F14:            return KEY_F14;
+        case XK_F15:            return KEY_F15;
+        case XK_F16:            return KEY_F16;
+        case XK_F17:            return KEY_F17;
+        case XK_F18:            return KEY_F18;
+        case XK_F19:            return KEY_F19;
+        case XK_F20:            return KEY_F20;
+        case XK_F21:            return KEY_F21;
+        case XK_F22:            return KEY_F22;
+        case XK_F23:            return KEY_F23;
+        case XK_F24:            return KEY_F24;
+        case XK_F25:            return KEY_F25;
+		
+        // Numeric keypad
+        case XK_KP_Divide:      return KEY_KP_DIVIDE;
+        case XK_KP_Multiply:    return KEY_KP_MULTIPLY;
+        case XK_KP_Subtract:    return KEY_KP_SUBTRACT;
+        case XK_KP_Add:         return KEY_KP_ADD;
+		
+        // These should have been detected in secondary keysym test above!
+        case XK_KP_Insert:      return KEY_KP_0;
+        case XK_KP_End:         return KEY_KP_1;
+        case XK_KP_Down:        return KEY_KP_2;
+        case XK_KP_Page_Down:   return KEY_KP_3;
+        case XK_KP_Left:        return KEY_KP_4;
+        case XK_KP_Right:       return KEY_KP_6;
+        case XK_KP_Home:        return KEY_KP_7;
+        case XK_KP_Up:          return KEY_KP_8;
+        case XK_KP_Page_Up:     return KEY_KP_9;
+        case XK_KP_Delete:      return KEY_KP_DECIMAL;
+        case XK_KP_Equal:       return KEY_KP_EQUAL;
+        case XK_KP_Enter:       return KEY_KP_ENTER;
+		
+        // Last resort: Check for printable keys (should not happen if the XKB
+        // extension is available). This will give a layout dependent mapping
+        // (which is wrong, and we may miss some keys, especially on non-US
+        // keyboards), but it's better than nothing...
+        case XK_a:              return KEY_A;
+        case XK_b:              return KEY_B;
+        case XK_c:              return KEY_C;
+        case XK_d:              return KEY_D;
+        case XK_e:              return KEY_E;
+        case XK_f:              return KEY_F;
+        case XK_g:              return KEY_G;
+        case XK_h:              return KEY_H;
+        case XK_i:              return KEY_I;
+        case XK_j:              return KEY_J;
+        case XK_k:              return KEY_K;
+        case XK_l:              return KEY_L;
+        case XK_m:              return KEY_M;
+        case XK_n:              return KEY_N;
+        case XK_o:              return KEY_O;
+        case XK_p:              return KEY_P;
+        case XK_q:              return KEY_Q;
+        case XK_r:              return KEY_R;
+        case XK_s:              return KEY_S;
+        case XK_t:              return KEY_T;
+        case XK_u:              return KEY_U;
+        case XK_v:              return KEY_V;
+        case XK_w:              return KEY_W;
+        case XK_x:              return KEY_X;
+        case XK_y:              return KEY_Y;
+        case XK_z:              return KEY_Z;
+        case XK_1:              return KEY_1;
+        case XK_2:              return KEY_2;
+        case XK_3:              return KEY_3;
+        case XK_4:              return KEY_4;
+        case XK_5:              return KEY_5;
+        case XK_6:              return KEY_6;
+        case XK_7:              return KEY_7;
+        case XK_8:              return KEY_8;
+        case XK_9:              return KEY_9;
+        case XK_0:              return KEY_0;
+        case XK_space:          return KEY_SPACE;
+        case XK_minus:          return KEY_MINUS;
+        case XK_equal:          return KEY_EQUAL;
+        case XK_bracketleft:    return KEY_LEFT_BRACKET;
+        case XK_bracketright:   return KEY_RIGHT_BRACKET;
+        case XK_backslash:      return KEY_BACKSLASH;
+        case XK_semicolon:      return KEY_SEMICOLON;
+        case XK_apostrophe:     return KEY_APOSTROPHE;
+        case XK_grave:          return KEY_GRAVE_ACCENT;
+        case XK_comma:          return KEY_COMMA;
+        case XK_period:         return KEY_PERIOD;
+        case XK_slash:          return KEY_SLASH;
+        case XK_less:           return KEY_WORLD_1; // At least in some layouts...
+        default:                break;
+    }
+	
+    // No matching translation was found
+    return KEY_UNKNOWN;
+}
+
+static void create_key_tables(platform_window window)
+{
+	s32 scancode, key;
+	char name[XkbKeyNameLength + 1];
+	XkbDescPtr desc = XkbGetMap(window.display, 0, XkbUseCoreKbd);
+	XkbGetNames(window.display, XkbKeyNamesMask, desc);
+	
+	// uncomment for layout independant input for games.
+#if 0
+	for (scancode = desc->min_key_code;  scancode <= desc->max_key_code;  scancode++)
+	{
+		memcpy(name, desc->names->keys[scancode].name, XkbKeyNameLength);
+		name[XkbKeyNameLength] = '\0';
+		
+		// Map the key name to a GLFW key code. Note: We only map printable
+		// keys here, and we use the US keyboard layout. The rest of the
+		// keys (function keys) are mapped using traditional KeySym
+		// translations.
+		if (strcmp(name, "TLDE") == 0) key = KEY_GRAVE_ACCENT;
+		else if (strcmp(name, "AE01") == 0) key = KEY_1;
+		else if (strcmp(name, "AE02") == 0) key = KEY_2;
+		else if (strcmp(name, "AE03") == 0) key = KEY_3;
+		else if (strcmp(name, "AE04") == 0) key = KEY_4;
+		else if (strcmp(name, "AE05") == 0) key = KEY_5;
+		else if (strcmp(name, "AE06") == 0) key = KEY_6;
+		else if (strcmp(name, "AE07") == 0) key = KEY_7;
+		else if (strcmp(name, "AE08") == 0) key = KEY_8;
+		else if (strcmp(name, "AE09") == 0) key = KEY_9;
+		else if (strcmp(name, "AE10") == 0) key = KEY_0;
+		else if (strcmp(name, "AE11") == 0) key = KEY_MINUS;
+		else if (strcmp(name, "AE12") == 0) key = KEY_EQUAL;
+		else if (strcmp(name, "AD01") == 0) key = KEY_Q;
+		else if (strcmp(name, "AD02") == 0) key = KEY_W;
+		else if (strcmp(name, "AD03") == 0) key = KEY_E;
+		else if (strcmp(name, "AD04") == 0) key = KEY_R;
+		else if (strcmp(name, "AD05") == 0) key = KEY_T;
+		else if (strcmp(name, "AD06") == 0) key = KEY_Y;
+		else if (strcmp(name, "AD07") == 0) key = KEY_U;
+		else if (strcmp(name, "AD08") == 0) key = KEY_I;
+		else if (strcmp(name, "AD09") == 0) key = KEY_O;
+		else if (strcmp(name, "AD10") == 0) key = KEY_P;
+		else if (strcmp(name, "AD11") == 0) key = KEY_LEFT_BRACKET;
+		else if (strcmp(name, "AD12") == 0) key = KEY_RIGHT_BRACKET;
+		else if (strcmp(name, "AC01") == 0) key = KEY_A;
+		else if (strcmp(name, "AC02") == 0) key = KEY_S;
+		else if (strcmp(name, "AC03") == 0) key = KEY_D;
+		else if (strcmp(name, "AC04") == 0) key = KEY_F;
+		else if (strcmp(name, "AC05") == 0) key = KEY_G;
+		else if (strcmp(name, "AC06") == 0) key = KEY_H;
+		else if (strcmp(name, "AC07") == 0) key = KEY_J;
+		else if (strcmp(name, "AC08") == 0) key = KEY_K;
+		else if (strcmp(name, "AC09") == 0) key = KEY_L;
+		else if (strcmp(name, "AC10") == 0) key = KEY_SEMICOLON;
+		else if (strcmp(name, "AC11") == 0) key = KEY_APOSTROPHE;
+		else if (strcmp(name, "AB01") == 0) key = KEY_Z;
+		else if (strcmp(name, "AB02") == 0) key = KEY_X;
+		else if (strcmp(name, "AB03") == 0) key = KEY_C;
+		else if (strcmp(name, "AB04") == 0) key = KEY_V;
+		else if (strcmp(name, "AB05") == 0) key = KEY_B;
+		else if (strcmp(name, "AB06") == 0) key = KEY_N;
+		else if (strcmp(name, "AB07") == 0) key = KEY_M;
+		else if (strcmp(name, "AB08") == 0) key = KEY_COMMA;
+		else if (strcmp(name, "AB09") == 0) key = KEY_PERIOD;
+		else if (strcmp(name, "AB10") == 0) key = KEY_SLASH;
+		else if (strcmp(name, "BKSL") == 0) key = KEY_BACKSLASH;
+		else if (strcmp(name, "LSGT") == 0) key = KEY_WORLD_1;
+		else key = KEY_UNKNOWN;
+		
+		if ((scancode >= 0) && (scancode < 256))
+			keycode_map[scancode] = key;
+	}
+#endif
+	
+	for (scancode = 0;  scancode < MAX_KEYCODE;  scancode++)
+	{
+		// Translate the un-translated key codes using traditional X11 KeySym
+		// lookups
+		
+		keycode_map[scancode] = translate_keycode(&window, scancode);
+	}
+	
+	XkbFreeNames(desc, XkbKeyNamesMask, True);
+	XkbFreeKeyboard(desc, 0, True);
+}
+
+inline void platform_init(int argc, char **argv)
+{
+#if 0
+	dlerror(); // clear error
+	void *x11 = dlopen("libX11.so.6", RTLD_NOW | RTLD_GLOBAL);
+	void *randr = dlopen("libXrandr.so", RTLD_NOW | RTLD_GLOBAL);
+#endif
+	
+	setlocale(LC_ALL, "en_US.UTF-8");
+	
+	XInitThreads();
+	
+	// get fullpath of the directory the binary is residing in
+	binary_path = platform_get_full_path(argv[0]);
+	
+	platform_create_config_directory();
+	
+	char buf[MAX_INPUT_LENGTH];
+	get_directory_from_path(buf, binary_path);
+	string_copyn(binary_path, buf, MAX_INPUT_LENGTH);
+	
+	assets_create();
+}
+
+inline void platform_destroy()
+{
+	assets_destroy();
+	
+#if defined(MODE_DEVELOPER)
+	memory_print_leaks();
+#endif
+}
+
+inline void platform_window_make_current(platform_window *window)
+{
+	glXMakeCurrent(window->display, window->window, window->gl_context);
+}
+
+void platform_window_set_size(platform_window *window, u16 width, u16 height)
+{
+	XResizeWindow(window->display, window->window, width, height);
+}
+
+void platform_window_set_position(platform_window *window, u16 x, u16 y)
+{
+	XMoveWindow(window->display, window->window, x, y);
+}
+
+
+vec2 platform_get_window_size(platform_window *window)
+{
+	vec2 res;
+	res.x = window->width;
+	res.y = window->height;
+	return res;
+}
+
+platform_window platform_open_window(char *name, u16 width, u16 height, u16 max_w, u16 max_h, u16 min_w, u16 min_h)
+{
+	bool has_max_size = max_w || max_h;
+	
+	platform_window window;
+	window.has_focus = true;
+	window.curr_cursor_type = CURSOR_DEFAULT;
+	window.next_cursor_type = CURSOR_DEFAULT;
+	window.clipboard_str = 0;
+	window.clipboard_strlen = 0;
+	
+	static int att[] =
+	{
+		GLX_X_RENDERABLE    , True,
+		GLX_DRAWABLE_TYPE   , GLX_WINDOW_BIT,
+		GLX_RENDER_TYPE     , GLX_RGBA_BIT,
+		GLX_X_VISUAL_TYPE   , GLX_TRUE_COLOR,
+		GLX_RED_SIZE        , 8,
+		GLX_GREEN_SIZE      , 8,
+		GLX_BLUE_SIZE       , 8,
+		GLX_ALPHA_SIZE      , 8,
+		GLX_DEPTH_SIZE      , 24,
+		GLX_STENCIL_SIZE    , 8,
+		GLX_DOUBLEBUFFER    , True,
+		//GLX_SAMPLE_BUFFERS  , 1,
+		//GLX_SAMPLES         , 4,
+		None
+	};
+	
+	window.display = XOpenDisplay(NULL);
+	
+	if(window.display == NULL) {
+		return window;
+	}
+	
+	window.parent = DefaultRootWindow(window.display);
+	
+	int fbcount;
+	GLXFBConfig* fbc = glXChooseFBConfig(window.display, DefaultScreen(window.display), att, &fbcount);
+	int best_fbc = -1, worst_fbc = -1, best_num_samp = -1, worst_num_samp = 999;
+	
+	int i;
+	for (i=0; i<fbcount; ++i)
+	{
+		XVisualInfo *vi = glXGetVisualFromFBConfig(window.display, fbc[i] );
+		if ( vi )
+		{
+			int samp_buf, samples;
+			glXGetFBConfigAttrib(window.display, fbc[i], GLX_SAMPLE_BUFFERS, &samp_buf );
+			glXGetFBConfigAttrib(window.display, fbc[i], GLX_SAMPLES       , &samples  );
+			
+			if ( best_fbc < 0 || (samp_buf && samples > best_num_samp))
+				best_fbc = i, best_num_samp = samples;
+			if ( worst_fbc < 0 || !samp_buf || samples < worst_num_samp )
+				worst_fbc = i, worst_num_samp = samples;
+		}
+		XFree(vi);
+	}
+	
+	GLXFBConfig bestFbc = fbc[best_fbc];
+	XFree(fbc);
+	
+	XVisualInfo *vi = glXGetVisualFromFBConfig(window.display, bestFbc );
+	window.visual_info = vi;
+	
+	if(window.visual_info == NULL) {
+		return window;
+	}
+	
+	window.cmap = XCreateColormap(window.display, window.parent, window.visual_info->visual, AllocNone);
+	
+	// calculate window center
+	XRRScreenResources *screens = XRRGetScreenResources(window.display, window.parent);
+	XRRCrtcInfo *info = XRRGetCrtcInfo(window.display, screens, screens->crtcs[0]);
+	
+	s32 center_x = (info->width / 2) - (width / 2);
+	s32 center_y = (info->height / 2) - (height / 2);
+	
+	XRRFreeCrtcInfo(info);
+	XRRFreeScreenResources(screens);
+	
+	XSetWindowAttributes window_attributes;
+	window_attributes.colormap = window.cmap;
+	window_attributes.border_pixel = 0;
+	window_attributes.event_mask = KeyPressMask | KeyReleaseMask | PointerMotionMask |
+		ButtonPressMask | ButtonReleaseMask | StructureNotifyMask | FocusChangeMask | LeaveWindowMask;
+	
+	window.window = XCreateWindow(window.display, window.parent, center_x, center_y, width, height, 0, window.visual_info->depth, InputOutput, window.visual_info->visual, CWColormap | CWEventMask | CWBorderPixel, &window_attributes);
+	
+	XMapWindow(window.display, window.window);
+	XFlush(window.display);
+	
+	XSync(window.display, False);
+	
+	XSizeHints hints;
+	
+	if (has_max_size)
+		hints.flags = PMaxSize | PMinSize | USPosition;
+	else
+		hints.flags = PMinSize | USPosition;
+	hints.x = center_x;
+	hints.y = center_y;
+	hints.max_width = width;
+	hints.max_height = height;
+	hints.min_width = min_w;
+	hints.min_height = min_h;
+	
+	XSetWMNormalHints(window.display, window.window, &hints);
+	
+	// window name
+	{
+		Atom WM_NAME = XInternAtom(window.display, "WM_NAME", False);
+		Atom _NET_WM_NAME = XInternAtom(window.display, "_NET_WM_NAME", False);
+		Atom _NET_WM_ICON_NAME = XInternAtom(window.display, "_NET_WM_ICON_NAME", False);
+		
+		char *list[1] = { (char *) name };
+		XTextProperty property;
+		
+		XStoreName(window.display, window.window, name);
+		
+		Xutf8TextListToTextProperty(window.display, list, 1, XUTF8StringStyle,
+									&property);
+		XSetTextProperty(window.display, window.window, &property, WM_NAME);
+		XSetTextProperty(window.display, window.window, &property, _NET_WM_NAME);
+		XSetTextProperty(window.display, window.window, &property, XA_WM_NAME);
+		XSetTextProperty(window.display, window.window, &property, _NET_WM_ICON_NAME);
+		XFree(property.value);
+		
+		XClassHint class_hint;
+		class_hint.res_name = name;
+		class_hint.res_class = name;
+		XSetClassHint(window.display, window.window, &class_hint);
+	}
+	
+	
+	// hide taskbar icon and stay on top
+#if 0
+	{
+		Atom wm_state = XInternAtom(window.display, 
+									"_NET_WM_STATE", False);
+		Atom taskbar_atom = XInternAtom(window.display, 
+										"_NET_WM_STATE_SKIP_TASKBAR", False);
+		Atom above_atom = XInternAtom(window.display, 
+									  "_NET_WM_STATE_ABOVE", False);
+		
+		Atom atom_list[2] = {taskbar_atom, above_atom};
+		XChangeProperty(window.display, window.window, wm_state, XA_ATOM, 32,
+						PropModeReplace, (const unsigned char*)&atom_list, 2);
+	}
+#endif
+	
+	{
+		XWMHints* win_hints = XAllocWMHints();
+		win_hints->flags = StateHint | IconPositionHint;
+		win_hints->initial_state = IconicState;
+		win_hints->icon_x = 0;
+		win_hints->icon_y = 0;
+		
+		/* pass the hints to the window manager. */
+		XSetWMHints(window.display, window.window, win_hints);
+		XFree(win_hints);
+	}
+	
+	static GLXContext share_list = 0;
+	
+	// get opengl context
+	window.gl_context = glXCreateContext(window.display, window.visual_info, 
+										 share_list, GL_TRUE);
+	
+	if (share_list == 0)
+		share_list = window.gl_context;
+	glXMakeCurrent(window.display, window.window, window.gl_context);
+	
+	// blending
+	glEnable(GL_DEPTH_TEST);
+	//glDepthMask(true);
+	//glClearDepth(50);
+	glDepthFunc(GL_LEQUAL);
+	
+	glEnable(GL_BLEND);
+	glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
+	
+	// setup multisampling
+#if 0
+	glEnable(GL_ALPHA_TEST);
+	glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE);
+	glEnable(GL_SAMPLE_ALPHA_TO_ONE);
+	glEnable(GL_MULTISAMPLE);
+	glHint(GL_MULTISAMPLE_FILTER_HINT_NV, GL_NICEST);
+#endif
+	
+	window.is_open = true;
+	window.width = width;
+	window.height = height;
+	
+	glMatrixMode(GL_PROJECTION);
+	glLoadIdentity();
+	glOrtho(0, width, height, 0, -1, 1);
+	
+	glMatrixMode(GL_MODELVIEW);
+	
+	create_key_tables(window);
+	
+	// recieve window close event
+	window.quit = XInternAtom(window.display, "WM_DELETE_WINDOW", False);
+	
+	GET_ATOM(XdndEnter);
+	GET_ATOM(XdndPosition);
+	GET_ATOM(XdndStatus);
+	GET_ATOM(XdndTypeList);
+	GET_ATOM(XdndActionCopy);
+	GET_ATOM(XdndDrop);
+	GET_ATOM(XdndFinished);
+	GET_ATOM(XdndSelection);
+	GET_ATOM(XdndLeave);
+	GET_ATOM(PRIMARY);
+	GET_ATOM(CLIPBOARD);
+	GET_ATOM(UTF8_STRING);
+	GET_ATOM(COMPOUND_STRING);
+	GET_ATOM(TARGETS);
+	GET_ATOM(MULTIPLE);
+	GET_ATOM(_NET_WM_STATE);
+	
+	array atoms = array_create(sizeof(Atom));
+	array_push(&atoms, &window.quit);
+	array_push(&atoms, &window.XdndEnter);
+	array_push(&atoms, &window.XdndPosition);
+	array_push(&atoms, &window.XdndStatus);
+	array_push(&atoms, &window.XdndTypeList);
+	array_push(&atoms, &window.XdndActionCopy);
+	array_push(&atoms, &window.XdndDrop);
+	array_push(&atoms, &window.XdndFinished);
+	array_push(&atoms, &window.XdndSelection);
+	array_push(&atoms, &window.XdndLeave);
+	array_push(&atoms, &window.PRIMARY);
+	array_push(&atoms, &window.CLIPBOARD);
+	array_push(&atoms, &window.UTF8_STRING);
+	array_push(&atoms, &window.COMPOUND_STRING);
+	array_push(&atoms, &window.TARGETS);
+	array_push(&atoms, &window.MULTIPLE);
+	array_push(&atoms, &window._NET_WM_STATE);
+	
+	XSetWMProtocols(window.display, window.window, atoms.data, atoms.length);
+	array_destroy(&atoms);
+	
+	Atom XdndAware = XInternAtom(window.display, "XdndAware", False);
+	Atom xdnd_version = 5;
+	XChangeProperty(window.display, window.window, XdndAware, XA_ATOM, 32,
+					PropModeReplace, (unsigned char*)&xdnd_version, 1);
+	
+	
+	XFlush(window.display);
+	XSync(window.display, True);
+	
+	return window;
+}
+
+inline bool platform_window_is_valid(platform_window *window)
+{
+	return window->window && window->display;
+}
+
+void platform_destroy_window(platform_window *window)
+{
+	glXMakeCurrent(window->display, None, NULL);
+	glXDestroyContext(window->display, window->gl_context);
+	XDestroyWindow(window->display, window->window);
+	XCloseDisplay(window->display);
+	XFree(window->visual_info);
+	mem_free(window->clipboard_str);
+	
+	window->window = 0;
+	window->display = 0;
+}
+
+void platform_hide_window_taskbar_icon(platform_window *window)
+{
+	XClientMessageEvent m;
+	memset(&m, 0, sizeof(XClientMessageEvent));
+	m.type = ClientMessage;
+	m.display = window->display;
+	m.window = window->window;
+	m.message_type = window->_NET_WM_STATE;
+	m.format=32;
+	m.data.l[0] = 1;
+	m.data.l[1] = XInternAtom(window->display, "_NET_WM_STATE_SKIP_TASKBAR", False);
+	m.data.l[2] = None;
+	m.data.l[3] = 1;
+	m.data.l[4] = 0;
+	XSendEvent(window->display, window->window, False, SubstructureRedirectMask | SubstructureNotifyMask, (XEvent*)&m);
+	
+	XFlush(window->display);
+}
+
+void platform_handle_events(platform_window *window, mouse_input *mouse, keyboard_input *keyboard)
+{
+	mouse->left_state &= ~MOUSE_CLICK;
+	mouse->right_state &= ~MOUSE_CLICK;
+	mouse->left_state &= ~MOUSE_DOUBLE_CLICK;
+	mouse->right_state &= ~MOUSE_DOUBLE_CLICK;
+	mouse->left_state &= ~MOUSE_RELEASE;
+	mouse->right_state &= ~MOUSE_RELEASE;
+	memset(keyboard->input_keys, 0, MAX_KEYCODE);
+	mouse->move_x = 0;
+	mouse->move_y = 0;
+	mouse->scroll_state = 0;
+	keyboard->text_changed = false;
+	
+	XClientMessageEvent m;
+	
+	s32 pending_events = XPending(window->display);
+	for (s32 i = 0; i < pending_events; i++)
+	{
+		XNextEvent(window->display, &window->event);
+		if (window->event.type == ClientMessage)
+		{
+			static int xdnd_version=0;
+			
+			if ((Atom)window->event.xclient.data.l[0] == window->quit) {
+				window->is_open = false;
+			}
+			
+			if (window->event.xclient.message_type == window->XdndDrop)
+			{
+				if (window->xdnd_req == None) {
+					/* say again - not interested! */
+					memset(&m, 0, sizeof(XClientMessageEvent));
+					m.type = ClientMessage;
+					m.display = window->event.xclient.display;
+					m.window = window->event.xclient.data.l[0];
+					m.message_type = window->XdndFinished;
+					m.format=32;
+					m.data.l[0] = window->window;
+					m.data.l[1] = 0;
+					m.data.l[2] = None; /* fail! */
+					XSendEvent(window->display, window->event.xclient.data.l[0], False, NoEventMask, (XEvent*)&m);
+				} else {
+					/* convert */
+					if(xdnd_version >= 1) {
+						XConvertSelection(window->display, window->XdndSelection, window->xdnd_req, window->PRIMARY, window->window, window->event.xclient.data.l[2]);
+					} else {
+						printf("time to find the time.\n");
+						//XConvertSelection(window->display, window->XdndSelection, window->xdnd_req, window->PRIMARY, window->xwindow, CurrentTime);
+					}
+				}
+			}
+		}
+		else if (window->event.type == LeaveNotify)
+		{
+			mouse->x = MOUSE_OFFSCREEN;
+			mouse->y = MOUSE_OFFSCREEN;
+		}
+		else if (window->event.type == ConfigureNotify)
+		{
+			XConfigureEvent xce = window->event.xconfigure;
+			window->width = xce.width;
+			window->height = xce.height;
+			glViewport(0, 0, window->width, window->height);
+		}
+		else if (window->event.type == FocusIn)
+		{
+			window->has_focus = true;
+		}
+		else if (window->event.type == FocusOut)
+		{
+			mouse->x = MOUSE_OFFSCREEN;
+			mouse->y = MOUSE_OFFSCREEN;
+			window->has_focus = false;
+			
+			// if windows loses focus, set all keys to not pressed
+			memset(keyboard->keys, 0, MAX_KEYCODE);
+		}
+		else if (window->event.type == MotionNotify)
+		{
+			s32 x = mouse->x;
+			s32 y = mouse->y;
+			
+			mouse->total_move_x += window->event.xmotion.x - mouse->x;
+			mouse->total_move_y += window->event.xmotion.y - mouse->y;
+			
+			mouse->x = window->event.xmotion.x;
+			mouse->y = window->event.xmotion.y;
+			
+			mouse->move_x = mouse->x - x;
+			mouse->move_y = mouse->y - y;
+		}
+		else if (window->event.type == ButtonPress)
+		{
+			Time ev_time = window->event.xbutton.time;
+			static Time last_ev_time = 0;
+			
+			bool is_left_down = window->event.xbutton.button == Button1;
+			bool is_right_down = window->event.xbutton.button == Button3;
+			bool is_middle_down = window->event.xbutton.button == Button2;
+			bool scroll_up = window->event.xbutton.button == Button4;
+			bool scroll_down = window->event.xbutton.button == Button5;
+			
+			if (scroll_up)
+				mouse->scroll_state = SCROLL_UP;
+			if (scroll_down)
+				mouse->scroll_state = SCROLL_DOWN;
+			
+			if (is_left_down)
+			{
+				if (ev_time - last_ev_time < 200)
+				{
+					mouse->left_state |= MOUSE_DOUBLE_CLICK;
+				}
+				
+				mouse->left_state |= MOUSE_DOWN;
+				mouse->left_state |= MOUSE_CLICK;
+				
+				mouse->total_move_x = 0;
+				mouse->total_move_y = 0;
+				last_ev_time = ev_time;
+			}
+			if (is_right_down)
+			{
+				mouse->right_state |= MOUSE_DOWN;
+				mouse->right_state |= MOUSE_CLICK;
+			}
+		}
+		else if (window->event.type == ButtonRelease)
+		{
+			bool is_left_up = window->event.xbutton.button == Button1;
+			bool is_right_up = window->event.xbutton.button == Button3;
+			bool is_middle_up = window->event.xbutton.button == Button2;
+			
+			if (is_left_up)
+			{
+				mouse->left_state = MOUSE_RELEASE;
+			}
+			if (is_right_up)
+			{
+				mouse->right_state = MOUSE_RELEASE;
+			}
+		}
+		else if(window->event.type == KeyPress) 
+		{
+			s32 key = window->event.xkey.keycode;
+			
+			keyboard->keys[keycode_map[key]] = true;
+			keyboard->input_keys[keycode_map[key]] = true;
+			
+			// https://gist.github.com/rickyzhang82/8581a762c9f9fc6ddb8390872552c250
+			//printf("state: %d\n", window->event.xkey.state);
+			
+			// remove key control key from mask so it doesnt block input
+			window->event.xkey.state &= ~ControlMask;
+			
+			// replace capslock with shiftkey else keylookup returns 0...
+			if (window->event.xkey.state == 2)
+				window->event.xkey.state = 1;
+			
+			KeySym ksym = XLookupKeysym(&window->event.xkey, window->event.xkey.state);
+			
+			if (keyboard->take_input)
+			{
+				char *ch = 0;
+				switch(ksym)
+				{
+					case XK_space: ch = " "; break;
+					case XK_exclam: ch = "!"; break;
+					case XK_quotedbl: ch = "\""; break;
+					case XK_numbersign: ch = "#"; break;
+					case XK_dollar: ch = "$"; break;
+					case XK_percent: ch = "%"; break;
+					case XK_ampersand: ch = "&"; break;
+					case XK_apostrophe: ch = "`"; break;
+					case XK_parenleft: ch = "("; break;
+					case XK_parenright: ch = ")"; break;
+					case XK_asterisk: ch = "*"; break;
+					case XK_plus: ch = "+"; break;
+					case XK_comma: ch = ","; break;
+					case XK_minus: ch = "-"; break;
+					case XK_period: ch = "."; break;
+					case XK_slash: ch = "/"; break;
+					case XK_0: ch = "0"; break;
+					case XK_1: ch = "1"; break;
+					case XK_2: ch = "2"; break;
+					case XK_3: ch = "3"; break;
+					case XK_4: ch = "4"; break;
+					case XK_5: ch = "5"; break;
+					case XK_6: ch = "6"; break;
+					case XK_7: ch = "7"; break;
+					case XK_8: ch = "8"; break;
+					case XK_9: ch = "9"; break;
+					
+					case XK_colon: ch = ":"; break;
+					case XK_semicolon: ch = ";"; break;
+					case XK_less: ch = "<"; break;
+					case XK_equal: ch = "="; break;
+					case XK_greater: ch = ">"; break;
+					case XK_question: ch = "?"; break;
+					case XK_at: ch = "@"; break;
+					case XK_bracketleft: ch = "["; break;
+					case XK_backslash: ch = "\\"; break;
+					case XK_bracketright: ch = "]"; break;
+					case XK_asciicircum: ch = "^"; break;
+					case XK_underscore: ch = "_"; break;
+					case XK_grave: ch = "`"; break;
+					case XK_braceleft: ch = "{"; break;
+					case XK_bar: ch = "|"; break;
+					case XK_braceright: ch = "}"; break;
+					case XK_asciitilde: ch = "~"; break;
+				}
+				
+				if ((ksym >= XK_A && ksym <= XK_Z) || (ksym >= XK_a && ksym <= XK_z))
+				{
+					ch = XKeysymToString(ksym);
+				}
+				
+				if (ch && keyboard->input_mode == INPUT_NUMERIC)
+				{
+					if (!(*ch >= 48 && *ch <= 57))
+					{
+						ch = 0;
+					}
+				}
+				
+				keyboard_handle_input_string(window, keyboard, ch);
+			}
+		}
+		else if (window->event.type == KeyRelease)
+		{
+			s32 key = window->event.xkey.keycode;
+			keyboard->keys[keycode_map[key]] = false;
+			
+			KeySym ksym = XLookupKeysym(&window->event.xkey, 0);
+		}
+		else if (window->event.type == SelectionClear)
+		{
+			window->clipboard_str = 0;
+			window->clipboard_strlen = 0;
+		}
+		else if (window->event.type == SelectionRequest)
+		{
+			Atom formats[] = {window->UTF8_STRING, window->COMPOUND_STRING, XA_STRING};
+			Atom targets[] = {window->TARGETS, window->MULTIPLE, window->UTF8_STRING, window->COMPOUND_STRING, XA_STRING};
+			int formatCount = sizeof(formats) / sizeof(formats[0]);
+			
+			XSelectionEvent event = {.type = SelectionNotify, .selection = window->event.xselectionrequest.selection, .target = window->event.xselectionrequest.target, .display = window->event.xselectionrequest.display, .requestor = window->event.xselectionrequest.requestor, .time = window->event.xselectionrequest.time};
+			
+			if(window->event.xselectionrequest.target == window->TARGETS) {
+				XChangeProperty(window->display, window->event.xselectionrequest.requestor, window->event.xselectionrequest.property, XA_ATOM, 32, PropModeReplace, (unsigned char*)targets, sizeof(targets) / sizeof(targets[0]));
+				
+				event.property = window->event.xselectionrequest.property;
+			} else {
+				event.property = None;
+				int i;
+				for(i = 0; i < formatCount; i++) {
+					if(window->event.xselectionrequest.target == formats[i]) {
+						XChangeProperty(window->display, window->event.xselectionrequest.requestor, window->event.xselectionrequest.property, window->event.xselectionrequest.target, 8, PropModeReplace, (unsigned char*)window->clipboard_str, window->clipboard_strlen);
+						
+						event.property = window->event.xselectionrequest.property;
+						break;
+					}
+				}
+			}
+			
+			XSendEvent(window->display, window->event.xselectionrequest.requestor, False, 0, (XEvent*)&event);
+			XFlush(window->display);
+		}
+	}
+}
+
+inline void platform_show_alert(char *title, char *message)
+{
+	char command[MAX_INPUT_LENGTH];
+	snprintf(command, MAX_INPUT_LENGTH, "notify-send \"%s\" \"%s\"", title, message);
+	platform_run_command(command);
+}
+
+inline void platform_window_swap_buffers(platform_window *window)
+{
+	// set cursor if changed
+	if (window->curr_cursor_type != window->next_cursor_type)
+	{
+		int cursor_shape = 0;
+		switch(window->next_cursor_type)
+		{
+			case CURSOR_DEFAULT: cursor_shape = XC_arrow; break;
+			case CURSOR_POINTER: cursor_shape = XC_hand1; break;
+		}
+		Cursor cursor = XCreateFontCursor(window->display, cursor_shape);
+		XDefineCursor(window->display, window->window, cursor);
+		window->curr_cursor_type = window->next_cursor_type;
+	}
+	
+	glXSwapBuffers(window->display, window->window);
+}
+
+u64 platform_get_time(time_type time_type, time_precision precision)
+{
+	s32 type = CLOCK_REALTIME;
+	switch(time_type)
+	{
+		case TIME_FULL: type = CLOCK_REALTIME; break;
+		case TIME_THREAD: type = CLOCK_THREAD_CPUTIME_ID; break;
+		case TIME_PROCESS: type = CLOCK_PROCESS_CPUTIME_ID; break;
+	}
+	
+	struct timespec tms;
+	if (clock_gettime(type,&tms)) {
+		return -1;
+	}
+	
+	long result = 0;
+	
+	if (precision == TIME_NS)
+	{
+		result = tms.tv_sec * 1000000000;
+		result += tms.tv_nsec;
+		if (tms.tv_nsec % 1000 >= 500) {
+			++result;
+		}
+	}
+	else if (precision == TIME_US)
+	{
+		result = tms.tv_sec * 1000000;
+		result += tms.tv_nsec/1000;
+		if (tms.tv_nsec % 1000 >= 500) {
+			++result;
+		}
+	}
+	else if (precision == TIME_MILI_S)
+	{
+		result = tms.tv_sec * 1000;
+		result += tms.tv_nsec/1000000;
+		if (tms.tv_nsec % 1000 >= 500) {
+			++result;
+		}
+	}
+	else if (precision == TIME_S)
+	{
+		result = tms.tv_sec;
+		result += tms.tv_nsec/1000000000;
+		if (tms.tv_nsec % 1000 >= 500) {
+			++result;
+		}
+	}
+	
+	return result;
+}
+
+inline s32 platform_get_cpu_count()
+{
+	return (int)sysconf(_SC_NPROCESSORS_ONLN);
+}
+
+inline s32 platform_get_memory_size()
+{
+	uint64_t aid = (uint64_t) sysconf(_SC_PHYS_PAGES);
+	aid         *= (uint64_t) sysconf(_SC_PAGESIZE);
+	aid         /= (uint64_t) (1024 * 1024);
+	return (int)(aid);
+}
+
+void platform_show_message(platform_window *window, char *message, char *title)
+{
+	char command[MAX_INPUT_LENGTH];
+	snprintf(command, MAX_INPUT_LENGTH, "zenity --info --text=\"%s\" --title=\"%s\" --width=240", message, title);
+	FILE *f = popen(command, "r");
+}
+
+static void* platform_open_file_dialog_thread(void *data)
+{
+	struct open_dialog_args *args = data;
+	
+	FILE *f;
+	
+	char current_val[MAX_INPUT_LENGTH];
+	string_copyn(current_val, args->buffer, MAX_INPUT_LENGTH);
+	
+	char file_filter[MAX_INPUT_LENGTH];
+	file_filter[0] = 0;
+	if (args->file_filter)
+		snprintf(file_filter, MAX_INPUT_LENGTH, "--file-filter=\"%s\"", args->file_filter);
+	
+	char start_path[MAX_INPUT_LENGTH];
+	start_path[0] = 0;
+	if (args->start_path)
+		snprintf(start_path, MAX_INPUT_LENGTH, "--filename=\"%s\"", args->start_path);
+	
+	
+	char command[MAX_INPUT_LENGTH];
+	
+	if (args->type == OPEN_FILE)
+	{
+		snprintf(command, MAX_INPUT_LENGTH, "zenity --file-selection %s %s", file_filter, start_path);
+	}
+	else if (args->type == OPEN_DIRECTORY)
+	{
+		snprintf(command, MAX_INPUT_LENGTH, "zenity --file-selection --directory %s %s", file_filter, start_path);
+	}
+	else if (args->type == SAVE_FILE)
+	{
+		snprintf(command, MAX_INPUT_LENGTH, "zenity --file-selection --save --confirm-overwrite %s %s", file_filter, start_path);
+	}
+	
+	f = popen(command, "r");
+	
+	char buffer[MAX_INPUT_LENGTH];
+	char *result = fgets(buffer, MAX_INPUT_LENGTH, f);
+	
+	if (!result)
+		return 0;
+	
+	// replace newlines with 0, we only want one file path
+	s32 len = strlen(buffer);
+	for (s32 x = 0; x < len; x++)
+	{
+		if (buffer[x] == '\n') buffer[x] = 0;
+	}
+	
+	if (strcmp(buffer, current_val) != 0 && strcmp(buffer, "") != 0)
+	{
+		string_copyn(args->buffer, buffer, MAX_INPUT_LENGTH);
+		s32 len = strlen(args->buffer);
+		args->buffer[len] = 0;
+	}
+	
+	return 0;
+}
+
+void *platform_open_file_dialog_block(void *arg)
+{
+	thread thr = thread_start(platform_open_file_dialog_thread, arg);
+	thread_join(&thr);
+	mem_free(arg);
+	return 0;
+}
+
+void platform_list_files_block(array *list, char *start_dir, array filters, bool recursive, memory_bucket *bucket, bool include_directories, bool *is_cancelled)
+{
+	assert(list);
+	
+	s32 len = 0;
+	char *matched_filter = 0;
+	
+	char *subdirname_buf;
+	if (bucket)
+		subdirname_buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+	else
+		subdirname_buf = mem_alloc(MAX_INPUT_LENGTH);
+	
+	DIR *d;
+	struct dirent *dir;
+	d = opendir(start_dir);
+	if (d) {
+		set_active_directory(start_dir);
+		while ((dir = readdir(d)) != NULL) {
+			if (*is_cancelled) break;
+			set_active_directory(start_dir);
+			
+			if (dir->d_type == DT_DIR)
+			{
+				if ((strcmp(dir->d_name, ".") == 0) || (strcmp(dir->d_name, "..") == 0))
+					continue;
+				
+				if (include_directories)
+				{
+					if ((len = filter_matches(&filters, dir->d_name, 
+											  &matched_filter)) && len != -1)
+					{
+						char *buf;
+						if (bucket)
+							buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+						else
+							buf = mem_alloc(MAX_INPUT_LENGTH);
+						
+						//realpath(dir->d_name, buf);
+						snprintf(buf, MAX_INPUT_LENGTH, "%s%s",start_dir, dir->d_name);
+						
+						found_file f;
+						f.path = buf;
+						
+						if (bucket)
+							f.matched_filter = memory_bucket_reserve(bucket, len+1);
+						else
+							f.matched_filter = mem_alloc(len+1);
+						
+						string_copyn(f.matched_filter, matched_filter, len+1);
+						
+						mutex_lock(&list->mutex);
+						array_push_size(list, &f, sizeof(found_file));
+						mutex_unlock(&list->mutex);
+					}
+				}
+				
+				if (recursive)
+				{
+					string_copyn(subdirname_buf, start_dir, MAX_INPUT_LENGTH);
+					string_appendn(subdirname_buf, dir->d_name, MAX_INPUT_LENGTH);
+					string_appendn(subdirname_buf, "/", MAX_INPUT_LENGTH);
+					
+					// do recursive search
+					platform_list_files_block(list, subdirname_buf, filters, recursive, bucket, include_directories, is_cancelled);
+				}
+			}
+			// we handle DT_UNKNOWN for file systems that do not support type lookup.
+			else if (dir->d_type == DT_REG || dir->d_type == DT_UNKNOWN)
+			{
+				// check if name matches pattern
+				if ((len = filter_matches(&filters, dir->d_name, 
+										  &matched_filter)) && len != -1)
+				{
+					char *buf;
+					if (bucket)
+						buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+					else
+						buf = mem_alloc(MAX_INPUT_LENGTH);
+					
+					//realpath(dir->d_name, buf);
+					snprintf(buf, MAX_INPUT_LENGTH, "%s%s",start_dir, dir->d_name);
+					
+					found_file f;
+					f.path = buf;
+					
+					if (bucket)
+						f.matched_filter = memory_bucket_reserve(bucket, len+1);
+					else
+						f.matched_filter = mem_alloc(len+1);
+					
+					string_copyn(f.matched_filter, matched_filter, len+1);
+					
+					mutex_lock(&list->mutex);
+					array_push_size(list, &f, sizeof(found_file));
+					mutex_unlock(&list->mutex);
+					
+				}
+			}
+		}
+		closedir(d);
+	}
+	
+	if (!bucket)
+		mem_free(subdirname_buf);
+}
+
+char *platform_get_full_path(char *file)
+{
+	char *buf = mem_alloc(PATH_MAX);
+	buf[0] = 0;
+	
+	char *result = realpath(file, buf);
+	
+	if (!result)
+	{
+		buf[0] = 0;
+		return buf;
+	}
+	
+	return buf;
+}
+
+inline u64 string_to_u64(char *str)
+{
+	return (u64)strtoull(str, 0, 10);
+}
+
+inline u32 string_to_u32(char *str)
+{
+	return (u32)strtoul(str, 0, 10);
+}
+
+inline u16 string_to_u16(char *str)
+{
+	return (u16)strtoul(str, 0, 10);
+}
+
+inline u8 string_to_u8(char *str)
+{
+	return (u8)strtoul(str, 0, 10);
+}
+
+inline s64 string_to_s64(char *str)
+{
+	return (s64)strtoll(str, 0, 10);
+}
+
+inline s32 string_to_s32(char *str)
+{
+	return (u32)strtol(str, 0, 10);
+}
+
+inline s16 string_to_s16(char *str)
+{
+	return (s16)strtol(str, 0, 10);
+}
+
+inline s8 string_to_s8(char *str)
+{
+	return (s8)strtol(str, 0, 10);
+}
+
+inline void platform_open_url(char *url)
+{
+	char buffer[MAX_INPUT_LENGTH];
+	snprintf(buffer, MAX_INPUT_LENGTH, "xdg-open %s", url);
+	platform_run_command(buffer);
+}
+
+inline void platform_run_command(char *command)
+{
+	s32 result = system(command);
+}
+
+void platform_set_icon(platform_window *window, image *img)
+{
+	s32 w = img->width;
+	s32 h = img->height;
+	
+	s32 nelements = (w * h) + 2;
+	
+	unsigned long data[nelements];
+	int i = 0;
+	(data)[i++] = w;
+	(data)[i++] = h;
+	
+	for (s32 y = 0; y < h; y++)
+	{
+		for (s32 x = 0; x < w; x++)
+		{
+			s32 *pixel = (s32*)(&((data)[i++]));
+			
+			s32 img_pixel = *(((s32*)img->data+(x+(y*w))));
+			
+			// 0xAABBGGRR
+			s32 a = (img_pixel>>24) & 0x000000FF;
+			s32 b = (img_pixel>>16) & 0x000000FF;
+			s32 g = (img_pixel>> 8) & 0x000000FF;
+			s32 r = (img_pixel>> 0) & 0x000000FF;
+			
+			//s32 c = (r << 24) | (g << 16) | (b << 8) | (a << 0);
+			s32 c = (a << 24) | (r << 16) | (g << 8) | (b << 0);
+			*pixel = c;
+		}
+	}
+	
+	Atom property = XInternAtom(window->display, "_NET_WM_ICON", 0);
+	Atom cardinal = XInternAtom(window->display, "CARDINAL", False);
+	
+	int result = XChangeProperty(window->display, window->window, 
+								 property, cardinal, 32, PropModeReplace, 
+								 (unsigned char *)data, nelements);
+}
diff --git a/src/linux/thread.c b/src/linux/thread.c
new file mode 100644
index 0000000..b6295ef
--- /dev/null
+++ b/src/linux/thread.c
@@ -0,0 +1,129 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+// stop gcc from reporting implicit declaration warning..
+extern long int syscall (long int __sysno, ...);
+extern int pthread_tryjoin_np(pthread_t thread, void **retval);
+
+thread thread_start(void *(*start_routine) (void *), void *arg)
+{
+	thread result;
+	result.valid = false;
+	
+	pthread_attr_t attr;
+	int attr_init_result = pthread_attr_init(&attr);
+	if (attr_init_result)
+		return result;
+	
+	int start_thread_result = pthread_create(&result.thread, &attr, start_routine, arg);
+	if (start_thread_result)
+	{
+		pthread_attr_destroy(&attr);
+		return result;
+	}
+	
+	result.valid = true;
+	pthread_attr_destroy(&attr);
+	
+	return result;
+}
+
+inline void thread_detach(thread *thread)
+{
+	if (thread->valid)
+	{
+		pthread_detach(thread->thread);
+	}
+}
+
+inline void thread_join(thread *thread)
+{
+	if (thread->valid)
+	{
+		void *retval;
+		pthread_join(thread->thread, &retval);
+	}
+}
+
+bool thread_tryjoin(thread *thread)
+{
+	if (thread->valid)
+	{
+		void *retval;
+		bool thread_joined = !pthread_tryjoin_np(thread->thread, &retval);
+		return thread_joined;
+	}
+	return false;
+}
+
+inline void thread_stop(thread *thread)
+{
+	if (thread->valid)
+	{
+		pthread_cancel(thread->thread);
+	}
+}
+
+mutex mutex_create()
+{
+	mutex result;
+	
+	pthread_mutexattr_t attr;
+	pthread_mutexattr_init(&attr);
+	pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_DEFAULT);
+	
+	pthread_mutex_init(&result.mutex, &attr);
+	
+	pthread_mutexattr_destroy(&attr);
+	
+	return result;
+}
+
+mutex mutex_create_recursive()
+{
+	mutex result;
+	
+	pthread_mutexattr_t attr;
+	pthread_mutexattr_init(&attr);
+	pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
+	
+	pthread_mutex_init(&result.mutex, &attr);
+	
+	pthread_mutexattr_destroy(&attr);
+	
+	return result;
+}
+
+inline void mutex_lock(mutex *mutex)
+{
+	pthread_mutex_lock(&mutex->mutex);
+}
+
+inline bool mutex_trylock(mutex *mutex)
+{
+	return !pthread_mutex_trylock(&mutex->mutex);
+}
+
+inline void mutex_unlock(mutex *mutex)
+{
+	pthread_mutex_unlock(&mutex->mutex);
+}
+
+inline void mutex_destroy(mutex *mutex)
+{
+	mutex_unlock(mutex);
+	pthread_mutex_destroy(&mutex->mutex);
+}
+
+inline u32 thread_get_id()
+{
+	return (u32)syscall(__NR_gettid);
+}
+
+inline void thread_sleep(u64 microseconds)
+{
+	usleep(microseconds);
+}
diff --git a/src/localization.c b/src/localization.c
new file mode 100644
index 0000000..41b1fcf
--- /dev/null
+++ b/src/localization.c
@@ -0,0 +1,149 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+mo_file load_localization_file(u8 *start_addr, u8 *end_addr, u8 *img_start, u8 *img_end,
+							   char *locale_id, char *locale_name)
+{
+	mo_file mo;
+	mo.translations = array_create(sizeof(mo_translation));
+	
+	{
+		mo.header = *(mo_header*)start_addr;
+		mo.locale_id = mem_alloc(strlen(locale_id)+1);
+		string_copyn(mo.locale_id, locale_id, strlen(locale_id)+1);
+		
+		mo.locale_full = mem_alloc(strlen(locale_name)+1);
+		string_copyn(mo.locale_full, locale_name, strlen(locale_name)+1);
+		
+		mo.icon = assets_load_image(img_start, img_end, false);
+		
+		char *buffer = (char*)start_addr;
+		mo_entry *identifiers = (mo_entry*)(buffer + mo.header.identifier_table_offset);
+		mo_entry *translations = (mo_entry*)(buffer + mo.header.translation_table_offset);
+		
+		// skip first one because this is file information
+		for (s32 i = 1; i < mo.header.number_of_strings; i++)
+		{
+			mo_entry *entry = &identifiers[i];
+			mo_entry *trans = &translations[i];
+			
+			mo_translation translation;
+			translation.identifier_len = entry->length;
+			translation.identifier = buffer+entry->offset;
+			translation.translation = buffer+trans->offset;
+			
+			array_push(&mo.translations, &translation);
+			//printf("%s=%s\n", translation.identifier, translation.translation);
+		}
+	}
+	
+	return mo;
+}
+
+char* locale_get_name()
+{
+	if (!global_localization.active_localization)
+	{
+		return "[NO LOCALE]";
+	}
+	
+	return global_localization.active_localization->locale_full;
+}
+
+char* locale_get_id()
+{
+	if (!global_localization.active_localization)
+	{
+		return "[NO LOCALE]";
+	}
+	
+	return global_localization.active_localization->locale_id;
+}
+
+bool set_locale(char *country_id)
+{
+	if (country_id == 0 && global_localization.mo_files.length)
+	{
+		global_localization.active_localization = array_at(&global_localization.mo_files, 0);
+		return true;
+	}
+	
+	for (s32 i = 0; i < global_localization.mo_files.length; i++)
+	{
+		mo_file *file = array_at(&global_localization.mo_files, i);
+		if (strcmp(file->locale_id, country_id) == 0)
+		{
+			global_localization.active_localization = file;
+			return true;
+		}
+	}
+	
+	// if localization is not found, default to first in list (english), return false to report error
+	if (global_localization.mo_files.length)
+		global_localization.active_localization = array_at(&global_localization.mo_files, 0);
+	else
+		global_localization.active_localization = 0;
+	
+	return false;
+}
+
+char* localize(const char *identifier)
+{
+	if (!global_localization.active_localization)
+	{
+		//printf("NO LOCALE SELECTED.");
+		return (char*)identifier;
+	}
+	
+	s32 len = strlen(identifier);
+	for (s32 i = 0; i < global_localization.active_localization->translations.length; i++)
+	{
+		mo_translation *trans = array_at(&global_localization.active_localization->translations, i);
+		
+		if (trans->identifier_len == len && strcmp(identifier, trans->identifier) == 0)
+		{
+			return trans->translation;
+		}
+	}
+	printf("MISSING TRANSLATION: [%s][%s]\n", identifier, global_localization.active_localization->locale_id);
+	return "MISSING";
+}
+
+void load_available_localizations()
+{
+	global_localization.mo_files = array_create(sizeof(mo_file));
+	array_reserve(&global_localization.mo_files, 10);
+	
+	mo_file en = load_localization_file(_binary____data_translations_en_English_mo_start,
+										_binary____data_translations_en_English_mo_end,
+										_binary____data_imgs_en_png_start,
+										_binary____data_imgs_en_png_end,
+										"en", "English");
+	
+	mo_file nl = load_localization_file(_binary____data_translations_nl_Dutch_mo_start,
+										_binary____data_translations_nl_Dutch_mo_end,
+										_binary____data_imgs_nl_png_start,
+										_binary____data_imgs_nl_png_end,
+										"nl", "Dutch");
+	
+	array_push(&global_localization.mo_files, &en);
+	array_push(&global_localization.mo_files, &nl);
+}
+
+void destroy_available_localizations()
+{
+	for (s32 i = 0; i < global_localization.mo_files.length; i++)
+	{
+		mo_file *file = array_at(&global_localization.mo_files, i);
+		array_destroy(&file->translations);
+		mem_free(file->locale_id);
+		mem_free(file->locale_full);
+		
+		if (file->icon)
+			assets_destroy_image(file->icon);
+	}
+	array_destroy(&global_localization.mo_files);
+}
diff --git a/src/localization.h b/src/localization.h
new file mode 100644
index 0000000..4a35bf1
--- /dev/null
+++ b/src/localization.h
@@ -0,0 +1,59 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_LOCALIZATION
+#define INCLUDE_LOCALIZATION
+
+// https://www.science.co.il/language/Locale-codes.php
+
+typedef struct t_mo_entry
+{
+	s32 length;
+	s32 offset;
+} mo_entry;
+
+typedef struct t_mo_translation
+{
+	s32 identifier_len;
+	char *identifier;
+	char *translation;
+} mo_translation;
+
+typedef struct t_mo_header
+{
+	s32 magic_number;
+	s32 file_format_revision;
+	s32 number_of_strings;
+	s32 identifier_table_offset;
+	s32 translation_table_offset;
+	s32 hashtable_size;
+	s32 hashtable_offset;
+} mo_header;
+
+typedef struct t_mo_file
+{
+	mo_header header;
+	array translations;
+	char *locale_id;
+	char *locale_full;
+	image *icon;
+} mo_file;
+
+typedef struct t_localization
+{
+	array mo_files;
+	mo_file *active_localization;
+} localization;
+
+localization global_localization;
+
+char* locale_get_id();
+char* locale_get_name();
+char* localize(const char *identifier);
+bool set_locale(char *country_id);
+void load_available_localizations();
+
+#endif
\ No newline at end of file
diff --git a/src/memory.h b/src/memory.h
new file mode 100644
index 0000000..f63edef
--- /dev/null
+++ b/src/memory.h
@@ -0,0 +1,19 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_MEMORY
+#define INCLUDE_MEMORY
+
+#define mem_alloc(size) malloc(size)
+#define mem_free(p) free(p)
+#define mem_realloc(p, size) realloc(p, size)
+#define memory_print_leaks() {}
+
+#define STBI_MALLOC(sz) mem_alloc(sz)
+#define STBI_REALLOC(p, newsz) mem_realloc(p, newsz)
+#define STBI_FREE(p) mem_free(p)
+
+#endif
\ No newline at end of file
diff --git a/src/memory_bucket.c b/src/memory_bucket.c
new file mode 100644
index 0000000..bc9f7e9
--- /dev/null
+++ b/src/memory_bucket.c
@@ -0,0 +1,74 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+inline memory_bucket memory_bucket_init(s32 bucket_size)
+{
+	assert(bucket_size >= MAX_INPUT_LENGTH);
+	
+	memory_bucket collection;
+	collection.bucket_mutex = mutex_create();
+	collection.buckets = array_create(sizeof(memory_bucket_entry));
+	
+	memory_bucket_entry bucket;
+	bucket.data = mem_alloc(bucket_size);
+	bucket.length = bucket_size;
+	bucket.cursor = 0;
+	array_push(&collection.buckets, &bucket);
+	return collection;
+}
+
+void* memory_bucket_reserve(memory_bucket *bucket, s32 reserve_length)
+{
+	mutex_lock(&bucket->bucket_mutex);
+	memory_bucket_entry *bucket_entry = 0;
+	for (s32 i = 0; i < bucket->buckets.length; i++)
+	{
+		bucket_entry = array_at(&bucket->buckets, i);
+		
+		if (bucket_entry->length - bucket_entry->cursor < reserve_length) continue;
+		
+		void *space = bucket_entry->data+bucket_entry->cursor;
+		bucket_entry->cursor += reserve_length;
+		mutex_unlock(&bucket->bucket_mutex);
+		
+		return space;
+	}
+	
+	// failed to find suitable space, allocate new bucket
+	memory_bucket_entry new_bucket;
+	new_bucket.data = mem_alloc(bucket_entry->length);
+	new_bucket.length = bucket_entry->length;
+	new_bucket.cursor = 0;
+	array_push(&bucket->buckets, &new_bucket);
+	mutex_unlock(&bucket->bucket_mutex);
+	
+	return new_bucket.data;
+}
+
+inline void memory_bucket_reset(memory_bucket *bucket)
+{
+	mutex_lock(&bucket->bucket_mutex);
+	for (s32 i = 0; i < bucket->buckets.length; i++)
+	{
+		memory_bucket_entry *bucket_entry = array_at(&bucket->buckets, i);
+		bucket_entry->cursor = 0;
+	}
+	mutex_unlock(&bucket->bucket_mutex);
+}
+
+inline void memory_bucket_destroy(memory_bucket *bucket)
+{
+	mutex_lock(&bucket->bucket_mutex);
+	for (s32 i = 0; i < bucket->buckets.length; i++)
+	{
+		memory_bucket_entry *bucket_entry = array_at(&bucket->buckets, i);
+		mem_free(bucket_entry->data);
+	}
+	array_destroy(&bucket->buckets);
+	mutex_unlock(&bucket->bucket_mutex);
+	
+	mutex_destroy(&bucket->bucket_mutex);
+}
\ No newline at end of file
diff --git a/src/memory_bucket.h b/src/memory_bucket.h
new file mode 100644
index 0000000..180af80
--- /dev/null
+++ b/src/memory_bucket.h
@@ -0,0 +1,26 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#define kilobytes(num) num*1000
+#define megabytes(num) kilobytes(num*1000)
+
+typedef struct t_memory_bucket_entry
+{
+	char *data;
+	s32 length;
+	s32 cursor;
+} memory_bucket_entry;
+
+typedef struct t_memory_bucket
+{
+	mutex bucket_mutex;
+	array buckets;
+} memory_bucket;
+
+memory_bucket memory_bucket_init(s32 bucket_size);
+void* memory_bucket_reserve(memory_bucket *bucket, s32 reserve_length);
+void memory_bucket_reset(memory_bucket *bucket);
+void memory_bucket_destroy(memory_bucket *bucket);
\ No newline at end of file
diff --git a/src/mo_edit.c b/src/mo_edit.c
new file mode 100644
index 0000000..019dbbe
--- /dev/null
+++ b/src/mo_edit.c
@@ -0,0 +1,531 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#include "config.h"
+#include "project_base.h"
+
+#include "languages.h"
+
+// TODO(Aldrik): option to disable menu item
+
+typedef struct t_translation
+{
+	s32 country_index;
+	char *value;
+} translation;
+
+typedef struct t_term
+{
+	char *name;
+	translation translations[COUNTRY_CODE_COUNT];
+} term;
+
+typedef struct t_translation_project
+{
+	array languages;
+	array terms;
+	int selected_term_index;
+} translation_project;
+
+translation_project *current_project = 0;
+
+scroll_state term_scroll;
+scroll_state lang_scroll;
+button_state btn_new_project;
+button_state btn_new_language;
+button_state btn_summary;
+dropdown_state dd_available_countries;
+textbox_state tb_filter;
+
+image *list_img;
+image *exclaim_img;
+image *delete_img;
+image *logo_small_img;
+
+font *font_medium;
+font *font_small;
+font *font_mini;
+s32 scroll_y = 0;
+
+#include "settings.h"
+#include "settings.c"
+
+static void load_assets()
+{
+	list_img = assets_load_image(_binary____data_imgs_list_png_start, 
+								 _binary____data_imgs_list_png_end, false);
+	exclaim_img = assets_load_image(_binary____data_imgs_exclaim_png_start, 
+									_binary____data_imgs_exclaim_png_end, false);
+	logo_small_img = assets_load_image(_binary____data_imgs_logo_64_png_start,
+									   _binary____data_imgs_logo_64_png_end, true);
+	delete_img = assets_load_image(_binary____data_imgs_delete_png_start,
+								   _binary____data_imgs_delete_png_end, false);
+	
+	font_medium = assets_load_font(_binary____data_fonts_mono_ttf_start,
+								   _binary____data_fonts_mono_ttf_end, 18);
+	font_small = assets_load_font(_binary____data_fonts_mono_ttf_start,
+								  _binary____data_fonts_mono_ttf_end, 15);
+	font_mini = assets_load_font(_binary____data_fonts_mono_ttf_start,
+								 _binary____data_fonts_mono_ttf_end, 12);
+}
+
+s32 get_available_country_index()
+{
+	s32 found_index = -1;
+	for (s32 x = 0; x < COUNTRY_CODE_COUNT; x++)
+	{
+		bool found = false;
+		for (s32 i = 0; i < current_project->languages.length; i++)
+		{
+			s32 ind = *(s32*)array_at(&current_project->languages, i);
+			if (ind == x) found = true;
+		}
+		
+		if (!found) found_index = x;
+	}
+	
+	return found_index;
+}
+
+bool country_has_been_added_to_project(s32 index)
+{
+	for (s32 i = 0; i < current_project->languages.length; i++)
+	{
+		s32 ind = *(s32*)array_at(&current_project->languages, i);
+		
+		if (index == ind) return true;
+	}
+	
+	return false;
+}
+
+s32 get_translated_count_for_language(s32 index)
+{
+	s32 count = 0;
+	for (s32 i = 0; i < current_project->terms.length; i++)
+	{
+		term *t = array_at(&current_project->terms, i);
+		
+		for (s32 x = 0; x < COUNTRY_CODE_COUNT; x++)
+		{
+			translation *tr = &t->translations[x];
+			if (tr->country_index == i)
+			{
+				count++;
+			}
+		}
+	}
+	
+	return count;
+}
+
+void start_new_project()
+{
+	current_project = mem_alloc(sizeof(translation_project));
+	
+	current_project->terms = array_create(sizeof(term));
+	array_reserve(&current_project->terms, 100);
+	current_project->terms.reserve_jump = 100;
+	
+	current_project->languages = array_create(sizeof(s32));
+	array_reserve(&current_project->languages, 100);
+	current_project->languages.reserve_jump = 100;
+	
+	for (s32 i = 0; i < 3; i++)
+	{
+		term t;
+		t.name = mem_alloc(10);
+		string_copyn(t.name, "meme_1", 10);
+		
+		for (s32 x = 0; x < COUNTRY_CODE_COUNT; x++)
+		{
+			translation tr;
+			tr.value = 0;
+			tr.country_index = -1;
+			t.translations[x] = tr;
+		}
+		
+		array_push(&current_project->terms, &t);
+	}
+	
+	current_project->selected_term_index = -1;
+}
+
+void load_config(settings_config *config)
+{
+	char *path = settings_config_get_string(config, "ACTIVE_PROJECT");
+	char *locale_id = settings_config_get_string(config, "LOCALE");
+	
+	if (locale_id)
+		set_locale(locale_id);
+	else
+		set_locale("en");
+}
+
+#if defined(OS_LINUX) || defined(OS_WIN)
+int main(int argc, char **argv)
+{
+	platform_init(argc, argv);
+	
+	bool is_command_line_run = (argc > 1);
+	if (is_command_line_run)
+	{
+		handle_command_line_arguments(argc, argv);
+		return 0;
+	}
+	
+	char config_path_buffer[PATH_MAX];
+	get_config_save_location(config_path_buffer);
+	
+	// load config
+	settings_config config = settings_config_load_from_file(config_path_buffer);
+	
+	s32 window_w = settings_config_get_number(&config, "WINDOW_WIDTH");
+	s32 window_h = settings_config_get_number(&config, "WINDOW_HEIGHT");
+	if (window_w <= 800 || window_h <= 600)
+	{
+		window_w = 800;
+		window_h = 600;
+	}
+	
+	platform_window window = platform_open_window("mo-edit", window_w, window_h, 0, 0, 800, 600);
+	main_window = &window;
+	
+	settings_page_create();
+	
+	load_available_localizations();
+	set_locale("en");
+	
+	load_assets();
+	
+	keyboard_input keyboard = keyboard_input_create();
+	mouse_input mouse = mouse_input_create();
+	
+	camera camera;
+	camera.x = 0;
+	camera.y = 0;
+	camera.rotation = 0;
+	
+	ui_create(&window, &keyboard, &mouse, &camera, font_small);
+	dd_available_countries = ui_create_dropdown();
+	term_scroll = ui_create_scroll(1);
+	lang_scroll = ui_create_scroll(1);
+	btn_summary = ui_create_button();
+	btn_new_project = ui_create_button();
+	btn_new_language = ui_create_button();
+	tb_filter = ui_create_textbox(MAX_INPUT_LENGTH);
+	
+	// asset worker
+	thread asset_queue_worker1 = thread_start(assets_queue_worker, NULL);
+	thread asset_queue_worker2 = thread_start(assets_queue_worker, NULL);
+	thread_detach(&asset_queue_worker1);
+	thread_detach(&asset_queue_worker2);
+	
+	load_config(&config);
+	
+	while(window.is_open) {
+        u64 last_stamp = platform_get_time(TIME_FULL, TIME_US);
+		platform_handle_events(&window, &mouse, &keyboard);
+		platform_set_cursor(&window, CURSOR_DEFAULT);
+		
+		settings_page_update_render();
+		
+		platform_window_make_current(&window);
+		
+		static bool icon_loaded = false;
+		if (!icon_loaded && logo_small_img->loaded)
+		{
+			icon_loaded = true;
+			platform_set_icon(&window, logo_small_img);
+		}
+		
+		if (global_asset_collection.queue.queue.length == 0 && !global_asset_collection.done_loading_assets)
+		{
+			global_asset_collection.done_loading_assets = true;
+		}
+		
+		global_ui_context.layout.active_window = &window;
+		global_ui_context.keyboard = &keyboard;
+		global_ui_context.mouse = &mouse;
+		
+		render_clear();
+		camera_apply_transformations(&window, &camera);
+		
+		global_ui_context.layout.width = global_ui_context.layout.active_window->width;
+		// begin ui
+		
+		ui_begin(1);
+		{
+			render_rectangle(0, 0, main_window->width, main_window->height, global_ui_context.style.background);
+			
+			ui_begin_menu_bar();
+			{
+				if (ui_push_menu(localize("file")))
+				{
+					// TODO(Aldrik): translate
+					if (ui_push_menu_item("Load Project", "Ctrl + O")) 
+					{ 
+					}
+					// TODO(Aldrik): translate
+					if (ui_push_menu_item("Save Project", "Ctrl + S")) 
+					{ 
+					}
+					// TODO(Aldrik): translate
+					if (ui_push_menu_item("Save Project As", "Ctrl + E"))  
+					{ 
+					}
+					// TODO(Aldrik): translate
+					if (ui_push_menu_item("Export MO files", "Ctrl + X"))  
+					{ 
+					}
+					ui_push_menu_item_separator();
+					if (ui_push_menu_item(localize("quit"), "Ctrl + Q")) 
+					{ 
+						window.is_open = false; 
+					}
+				}
+			}
+			ui_end_menu_bar();
+			
+			
+			// TODO(Aldrik): make this a setting, resizable panel
+			global_ui_context.layout.width = 300;
+			ui_push_vertical_dragbar();
+			
+			if (current_project)
+			{
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_button_image(&btn_summary, "", list_img);
+					// TODO(Aldrik): translate
+					ui_push_textf_width(font_medium, "Terms", global_ui_context.layout.width-150);
+					
+					ui_push_button_image(&btn_summary, "", delete_img);
+					ui_push_button_image(&btn_summary, "", delete_img);
+					
+				}
+				ui_block_end();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					// TODO(Aldrik): translate
+					ui_push_textbox(&tb_filter, "Filter terms..");
+				}
+				ui_block_end();
+				
+				ui_push_separator();
+				
+				term_scroll.height = main_window->height-global_ui_context.layout.offset_y;
+				ui_scroll_begin(&term_scroll);
+				{
+					for (s32 i = 0; i < current_project->terms.length; i++)
+					{
+						term *t = array_at(&current_project->terms, i);
+						
+						ui_push_button_image(&btn_summary, "", delete_img);
+						//ui_push_image(exclaim_img, 14, 14, 1, rgb(255,255,255));
+						ui_push_text_width(t->name, global_ui_context.layout.width-100);
+						
+						ui_block_end();
+					}
+				}
+				ui_scroll_end();
+			}
+			else
+			{
+				// TODO(Aldrik): translate
+				if (ui_push_button(&btn_new_project, "Create new project"))
+				{
+					start_new_project();
+				}
+			}
+			
+			global_ui_context.layout.width = main_window->width - 310;
+			
+			global_ui_context.layout.offset_x = 310;
+			global_ui_context.layout.offset_y = MENU_BAR_HEIGHT + WIDGET_PADDING;
+			
+			if (current_project && current_project->selected_term_index >= 0)
+			{
+				
+			}
+			else if (current_project)
+			{
+				// overview
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					// TODO(Aldrik): translate
+					ui_push_textf_width(font_medium, "Overview", 200);
+					
+					char info_text[60];
+					sprintf(info_text, "%d terms, %d languages", current_project->terms.length, current_project->languages.length);
+					
+					color c = global_ui_context.style.foreground;
+					global_ui_context.style.foreground = rgb(110,110,110);
+					ui_push_textf(font_small, info_text);
+					global_ui_context.style.foreground = c;
+				}
+				ui_block_end();
+				
+				ui_push_separator();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					s32 av_index = get_available_country_index();
+					
+					if (dd_available_countries.selected_index == -1 && av_index >= 0)
+						dd_available_countries.selected_index = av_index;
+					
+					if (dd_available_countries.selected_index >= 0)
+					{
+						if (ui_push_dropdown(&dd_available_countries, 
+											 global_langues[dd_available_countries.selected_index].fullname))
+						{
+							for (s32 i = 0; i < COUNTRY_CODE_COUNT; i++)
+							{
+								if (!country_has_been_added_to_project(i))
+								{
+									if (ui_push_dropdown_item(0, global_langues[i].fullname, i))
+									{
+										
+									}
+								}
+							}
+						}
+						
+						// TODO(Aldrik): translate
+						if (ui_push_button(&btn_new_language, "Add"))
+						{
+							array_push(&current_project->languages,&dd_available_countries.selected_index);
+							
+							dd_available_countries.selected_index = -1;
+						}
+					}
+				}
+				ui_block_end();
+				
+				if (dd_available_countries.selected_index >= 0)
+					ui_push_separator();
+				
+				// languages
+				lang_scroll.height = main_window->height-global_ui_context.layout.offset_y;
+				ui_scroll_begin(&lang_scroll);
+				{
+					for (s32 i = 0; i < current_project->languages.length; i++)
+					{
+						button_state btn_remove = ui_create_button();
+						
+						bool pressed = false;
+						if (ui_push_button_image(&btn_remove, "", delete_img))
+						{
+							pressed = true;
+						}
+						
+						s32 index = *(s32*)array_at(&current_project->languages, i);
+						ui_push_text_width(global_langues[index].fullname, global_ui_context.layout.width-200);
+						
+						color c = global_ui_context.style.foreground;
+						global_ui_context.style.foreground = rgb(110,110,110);
+						
+						char stats[50];
+						sprintf(stats, "%d/%d translated", get_translated_count_for_language(index), current_project->terms.length);
+						ui_push_text(stats);
+						
+						global_ui_context.style.foreground = c;
+						
+						if (pressed)
+						{
+							array_remove_at(&current_project->languages, i);
+							i--;
+						}
+						
+						ui_block_end();
+					}
+				}
+				ui_scroll_end();
+			}
+			else
+			{
+				// show no project loaded message/image
+			}
+		}
+		ui_end();
+		// end ui
+		
+		assets_do_post_process();
+		platform_window_swap_buffers(&window);
+		
+		u64 current_stamp = platform_get_time(TIME_FULL, TIME_US);
+		u64 diff = current_stamp - last_stamp;
+		float diff_ms = diff / 1000.0f;
+		last_stamp = current_stamp;
+		
+		if (diff_ms < TARGET_FRAMERATE)
+		{
+			double time_to_wait = (TARGET_FRAMERATE) - diff_ms;
+			thread_sleep(time_to_wait*1000);
+		}
+    }
+	
+	settings_page_hide_without_save();
+	
+	// write config file
+#if 0
+	settings_config_set_string(&config, "SEARCH_DIRECTORY", textbox_path.buffer);
+	settings_config_set_number(&config, "SEARCH_DIRECTORIES", checkbox_recursive.state);
+	settings_config_set_string(&config, "SEARCH_TEXT", textbox_search_text.buffer);
+	settings_config_set_string(&config, "FILE_FILTER", textbox_file_filter.buffer);
+	settings_config_set_number(&config, "MAX_THEAD_COUNT", global_settings_page.max_thread_count);
+	settings_config_set_number(&config, "MAX_FILE_SIZE", global_settings_page.max_file_size);
+	
+	vec2 win_size = platform_get_window_size(&window);
+	settings_config_set_number(&config, "WINDOW_WIDTH", win_size.x);
+	settings_config_set_number(&config, "WINDOW_HEIGHT", win_size.y);
+	
+	settings_config_set_number(&config, "STYLE", global_ui_context.style.id);
+	settings_config_set_number(&config, "DOUBLE_CLICK_ACTION", global_settings_page.selected_double_click_selection_option);
+	
+	if (global_localization.active_localization != 0)
+	{
+		char *current_locale_id = locale_get_id();
+		if (current_locale_id)
+		{
+			settings_config_set_string(&config, "LOCALE", current_locale_id);
+		}
+	}
+	
+	settings_config_write_to_file(&config, config_path_buffer);
+	settings_config_destroy(&config);
+#endif
+	
+	settings_page_destroy();
+	
+	destroy_available_localizations();
+	
+#if 0
+	// cleanup ui
+	ui_destroy_textbox(&textbox_path);
+	ui_destroy_textbox(&textbox_search_text);
+	ui_destroy_textbox(&textbox_file_filter);
+#endif
+	ui_destroy();
+	
+	// delete assets
+	assets_destroy_image(list_img);
+	assets_destroy_image(logo_small_img);
+	assets_destroy_image(delete_img);
+	
+	assets_destroy_font(font_small);
+	assets_destroy_font(font_mini);
+	
+	keyboard_input_destroy(&keyboard);
+	platform_destroy_window(&window);
+	
+	platform_destroy();
+	
+	return 0;
+}
+#endif
\ No newline at end of file
diff --git a/src/platform.h b/src/platform.h
new file mode 100644
index 0000000..d256c5f
--- /dev/null
+++ b/src/platform.h
@@ -0,0 +1,221 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_PLATFORM
+#define INCLUDE_PLATFORM
+
+typedef struct t_platform_window platform_window;
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+typedef struct t_found_file
+{
+	char *matched_filter;
+	char *path;
+} found_file;
+
+typedef struct t_file_match
+{
+	found_file file;
+	s16 file_error;
+	s32 file_size;
+	
+	u32 line_nr;
+	s32 word_match_offset;
+	s32 word_match_length;
+	s32 word_match_offset_x; // highlight render offset
+	s32 word_match_width; // highlight render width
+	char *line_info; // will be null when no match is found
+} file_match;
+
+typedef struct t_search_result
+{
+	array work_queue;
+	array files;
+	array matches;
+	u64 find_duration_us;
+	array errors;
+	bool show_error_message; // error occured
+	bool found_file_matches; // found/finding file matches
+	s32 files_searched;
+	s32 files_matched;
+	s32 search_result_source_dir_len;
+	bool match_found; // found text match
+	mutex mutex;
+	bool walking_file_system;
+	bool cancel_search;
+	bool done_finding_matches;
+	s32 search_id;
+	u64 start_time;
+	bool done_finding_files;
+	memory_bucket mem_bucket;
+	bool is_command_line_search;
+	bool threads_closed;
+	
+	char *export_path;
+	char *file_filter;
+	char *directory_to_search;
+	char *text_to_find;
+	s32 max_thread_count;
+	s32 max_file_size;
+	bool is_recursive;
+} search_result;
+
+typedef struct t_find_text_args
+{
+	file_match file;
+	search_result *search_result_buffer;
+} find_text_args;
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+typedef struct t_file_content
+{
+	s64 content_length;
+	void *content;
+	s16 file_error;
+} file_content;
+
+typedef enum t_time_type
+{
+	TIME_FULL,     // realtime
+	TIME_THREAD,   // run time for calling thread
+	TIME_PROCESS,  // run time for calling process
+} time_type;
+
+typedef enum t_time_precision
+{
+	TIME_NS, // nanoseconds
+	TIME_US, // microseconds
+	TIME_MILI_S, // miliseconds
+	TIME_S,  // seconds
+} time_precision;
+
+typedef struct t_cpu_info
+{
+	s32 model;
+	char model_name[255];
+	float32 frequency;
+	u32 cache_size;
+	u32 cache_alignment;
+} cpu_info;
+
+typedef enum t_file_dialog_type
+{
+	OPEN_FILE,
+	OPEN_DIRECTORY,
+	SAVE_FILE,
+} file_dialog_type;
+
+typedef enum t_file_open_error
+{
+	FILE_ERROR_TOO_MANY_OPEN_FILES_PROCESS = 1,
+	FILE_ERROR_TOO_MANY_OPEN_FILES_SYSTEM = 2,
+	FILE_ERROR_NO_ACCESS = 3,
+	FILE_ERROR_NOT_FOUND = 4,
+	FILE_ERROR_CONNECTION_ABORTED = 5,
+	FILE_ERROR_CONNECTION_REFUSED = 6,
+	FILE_ERROR_NETWORK_DOWN = 7,
+	FILE_ERROR_REMOTE_IO_ERROR = 8,
+	FILE_ERROR_STALE = 9, // NFS server file is removed/renamed
+	FILE_ERROR_GENERIC = 10,
+} file_open_error;
+
+struct open_dialog_args
+{
+	char *buffer;
+	char *file_filter;
+	char *start_path;
+	char *default_save_file_extension;
+	file_dialog_type type;
+};
+
+typedef struct t_list_file_args
+{
+	array *list;
+	char *start_dir;
+	char *pattern;
+	bool recursive;
+	bool include_directories;
+	bool *state;
+	bool *is_cancelled;
+	memory_bucket *bucket;
+} list_file_args;
+
+typedef enum t_cursor_type
+{
+	CURSOR_DEFAULT,
+	CURSOR_POINTER,
+} cursor_type;
+
+typedef struct t_vec2
+{
+	s32 x;
+	s32 y;
+} vec2;
+
+platform_window *main_window = 0;
+platform_window *settings_window = 0;
+
+bool platform_window_is_valid(platform_window *window);
+platform_window platform_open_window(char *name, u16 width, u16 height, u16 max_w, u16 max_h, u16 min_w, u16 min_h);
+void platform_get_focus(platform_window *window);
+bool platform_set_clipboard(platform_window *window, char *buffer);
+bool platform_get_clipboard(platform_window *window, char *buffer);
+void platform_window_set_size(platform_window *window, u16 width, u16 height);
+void platform_window_set_position(platform_window *window, u16 x, u16 y);
+void platform_destroy_window(platform_window *window);
+void platform_handle_events(platform_window *window, mouse_input *mouse, keyboard_input *keyboard);
+void platform_window_swap_buffers(platform_window *window);
+void platform_set_cursor(platform_window *window, cursor_type type);
+void platform_window_set_title(platform_window *window, char *name);
+file_content platform_read_file_content(char *path, const char *mode);
+bool platform_write_file_content(char *path, const char *mode, char *buffer, s32 len);
+void platform_destroy_file_content(file_content *content);
+bool get_active_directory(char *buffer);
+bool set_active_directory(char *path);
+void platform_show_message(platform_window *window, char *message, char *title);
+array get_filters(char *filter);
+void platform_list_files_block(array *list, char *start_dir, array filters, bool recursive, memory_bucket *bucket, bool include_directories, bool *is_cancelled);
+void platform_list_files(array *list, char *start_dir, char *filter, bool recursive, memory_bucket *bucket, bool *is_cancelled, bool *state);
+void platform_open_file_dialog(file_dialog_type type, char *buffer, char *file_filter, char *start_path);
+bool is_platform_in_darkmode();
+void *platform_open_file_dialog_block(void *arg);
+char *platform_get_full_path(char *file);
+void platform_open_url(char *command);
+void platform_run_command(char *command);
+void platform_window_make_current(platform_window *window);
+void platform_init(int argc, char **argv);
+void platform_destroy();
+void platform_set_icon(platform_window *window, image *img);
+void platform_autocomplete_path(char *buffer, bool want_dir);
+bool platform_directory_exists(char *path);
+bool platform_file_exists(char *path);
+void platform_show_alert(char *title, char *message);
+char *get_config_save_location(char *buffer);
+char *get_file_extension(char *path);
+void get_name_from_path(char *buffer, char *path);
+void get_directory_from_path(char *buffer, char *path);
+vec2 platform_get_window_size(platform_window *window);
+s32 filter_matches(array *filters, char *string, char **matched_filter);
+
+u64 platform_get_time(time_type time_type, time_precision precision);
+s32 platform_get_memory_size();
+s32 platform_get_cpu_count();
+
+u64 string_to_u64(char *str);
+u32 string_to_u32(char *str);
+u16 string_to_u16(char *str);
+u8 string_to_u8(char *str);
+
+s64 string_to_s64(char *str);
+s32 string_to_s32(char *str);
+s16 string_to_s16(char *str);
+s8 string_to_s8(char *str);
+
+#endif
\ No newline at end of file
diff --git a/src/platform_shared.c b/src/platform_shared.c
new file mode 100644
index 0000000..4107ca3
--- /dev/null
+++ b/src/platform_shared.c
@@ -0,0 +1,244 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+void get_name_from_path(char *buffer, char *path)
+{
+	buffer[0] = 0;
+	
+	s32 len = strlen(path);
+	if (len == 1)
+	{
+		return;
+	}
+	
+	char *path_end = path + len;
+#ifdef OS_LINUX
+	while (*path_end != '/' && path_end >= path)
+	{
+		--path_end;
+	}
+#endif
+#ifdef OS_WIN
+	while (*path_end != '\\' && path_end >= path)
+	{
+		--path_end;
+	}
+#endif
+	
+	string_copyn(buffer, path_end+1, MAX_INPUT_LENGTH);
+}
+
+void get_directory_from_path(char *buffer, char *path)
+{
+	buffer[0] = 0;
+	
+	s32 len = strlen(path);
+	if (len == 1)
+	{
+		return;
+	}
+	
+	char *path_end = path + len;
+#ifdef OS_LINUX
+	while (*path_end != '/' && path_end >= path)
+	{
+		--path_end;
+	}
+#endif
+#ifdef OS_WIN
+	while (*path_end != '\\' && path_end >= path)
+	{
+		--path_end;
+	}
+#endif
+	
+	s32 offset = path_end - path;
+	char ch = path[offset+1];
+	path[offset+1] = 0;
+	string_copyn(buffer, path, MAX_INPUT_LENGTH);
+	path[offset+1] = ch;
+}
+
+void platform_autocomplete_path(char *buffer, bool want_dir)
+{
+	char dir[MAX_INPUT_LENGTH]; 
+	char name[MAX_INPUT_LENGTH]; 
+	get_directory_from_path(dir, buffer);
+	get_name_from_path(name, buffer);
+	
+	// nothing to autocomplete
+	if (name[0] == 0)
+	{
+		return;
+	}
+	
+	// create filter
+	string_appendn(name, "*", MAX_INPUT_LENGTH);
+	
+	array files = array_create(sizeof(found_file));
+	array filters = get_filters(name);
+	bool is_cancelled = false;
+	platform_list_files_block(&files, dir, filters, false, 0, want_dir, &is_cancelled);
+	
+	s32 index_to_take = -1;
+	if (want_dir)
+	{
+		for (s32 i = 0; i < files.length; i++)
+		{
+			found_file *file = array_at(&files, i);
+			
+			if (platform_directory_exists(file->path))
+			{
+				index_to_take = i;
+				break;
+			}
+		}
+	}
+	else
+	{
+		index_to_take = 0;
+	}
+	
+	array_destroy(&filters);
+	
+	if (files.length > 0 && index_to_take != -1)
+	{
+		found_file *file = array_at(&files, index_to_take);
+		string_copyn(buffer, file->path, MAX_INPUT_LENGTH);
+	}
+	
+	for (s32 i = 0; i < files.length; i++)
+	{
+		found_file *match = array_at(&files, i);
+		mem_free(match->matched_filter);
+		mem_free(match->path);
+	}
+	array_destroy(&files);
+}
+
+array get_filters(char *pattern)
+{
+	array result = array_create(MAX_INPUT_LENGTH);
+	
+	char current_filter[MAX_INPUT_LENGTH];
+	s32 filter_len = 0;
+	while(*pattern)
+	{
+		char ch = *pattern;
+		
+		if (ch == ',')
+		{
+			current_filter[filter_len] = 0;
+			array_push(&result, current_filter);
+			filter_len = 0;
+		}
+		else
+		{
+			if(filter_len < MAX_INPUT_LENGTH-1)
+			{
+				current_filter[filter_len++] = ch;
+			}
+			else
+			{
+				current_filter[filter_len] = ch;
+			}
+		}
+		
+		pattern++;
+	}
+	current_filter[filter_len] = 0;
+	array_push(&result, current_filter);
+	
+	return result;
+}
+
+void *platform_list_files_thread(void *args)
+{
+	list_file_args *info = args;
+	
+	array filters = get_filters(info->pattern);
+	
+	array *list = info->list;
+	char *start_dir = info->start_dir;
+	bool recursive = info->recursive;
+	
+	platform_list_files_block(info->list, info->start_dir, filters, info->recursive, info->bucket, info->include_directories, info->is_cancelled);
+	
+	mutex_lock(&info->list->mutex);
+	//if (!(*info->is_cancelled))
+	*(info->state) = true;
+	mutex_unlock(&info->list->mutex);
+	
+	array_destroy(&filters);
+	
+	return 0;
+}
+
+void platform_list_files(array *list, char *start_dir, char *filter, bool recursive, memory_bucket *bucket, bool *is_cancelled, bool *state)
+{
+	list_file_args *args = memory_bucket_reserve(bucket, sizeof(list_file_args));
+	args->list = list;
+	args->start_dir = start_dir;
+	args->pattern = filter;
+	args->recursive = recursive;
+	args->state = state;
+	args->include_directories = 0;
+	args->bucket = bucket;
+	args->is_cancelled = is_cancelled;
+	
+	thread thr = thread_start(platform_list_files_thread, args);
+	thread_detach(&thr);
+}
+
+void platform_open_file_dialog(file_dialog_type type, char *buffer, char *file_filter, char *start_path)
+{
+	struct open_dialog_args *args = mem_alloc(sizeof(struct open_dialog_args));
+	args->buffer = buffer;
+	args->type = type;
+	args->file_filter = file_filter;
+	args->start_path = start_path;
+	
+	thread thr;
+	thr.valid = false;
+	
+	while (!thr.valid)
+		thr = thread_start(platform_open_file_dialog_block, args);
+	thread_detach(&thr);
+}
+
+void destroy_found_file_array(array *found_files)
+{
+	for (s32 i = 0; i < found_files->length; i++)
+	{
+		found_file *f = array_at(found_files, i);
+		mem_free(f->matched_filter);
+		mem_free(f->path);
+	}
+	array_destroy(found_files);
+}
+
+char *get_file_extension(char *path)
+{
+	while(*path != '.' && *path)
+	{
+		path++;
+	}
+	return path;
+}
+
+s32 filter_matches(array *filters, char *string, char **matched_filter)
+{
+	for (s32 i = 0; i < filters->length; i++)
+	{
+		char *filter = array_at(filters, i);
+		if (string_match(filter, string))
+		{
+			*matched_filter = filter;
+			return strlen(filter);
+		}
+	}
+	return -1;
+}
diff --git a/src/project_base.h b/src/project_base.h
new file mode 100644
index 0000000..0563934
--- /dev/null
+++ b/src/project_base.h
@@ -0,0 +1,111 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_PROJECT_BASE
+#define INCLUDE_PROJECT_BASE
+
+#ifdef _WIN32
+#define OS_WIN
+#include <windows.h>
+#include <time.h>
+#endif
+#ifdef __linux__
+#define OS_LINUX
+#include <sys/times.h>
+#include <sys/vtimes.h>
+#endif
+#ifdef __APPLE__
+#define OS_OSX
+#error platform not supported
+#endif
+
+#include "stdint.h"
+#include "string.h"
+#include "assert.h"
+
+#include <GL/gl.h>
+#ifdef OS_LINUX
+#include <GL/glx.h>
+#endif
+#include <GL/glu.h>
+#include <GL/glext.h>
+
+#define s8 int8_t
+#define s16 int16_t
+#define s32 int32_t
+#define s64 int64_t
+
+#define u8 uint8_t
+#define u16 uint16_t
+#define u32 uint32_t
+#define u64 uint64_t
+
+#define float32 float
+#define float64 double
+
+#ifdef OS_LINUX
+#define bool uint8_t
+#endif
+#ifdef OS_WIN
+#define bool _Bool
+#endif
+
+#define true 1
+#define false 0
+
+#include "thread.h"
+#include "array.h"
+#include "memory.h"
+
+#define STB_IMAGE_IMPLEMENTATION
+#include "external/stb_image.h"
+
+#define STB_TRUETYPE_IMPLEMENTATION
+#include "external/stb_truetype.h"
+
+#include "external/utf8.h"
+#include "input.h"
+#include "assets.h"
+#include "memory_bucket.h"
+#include "platform.h"
+#include "render.h"
+#include "camera.h"
+#include "ui.h"
+#include "string_utils.h"
+#include "settings_config.h"
+#include "localization.h"
+#include "command_line.h"
+
+#include "platform_shared.c"
+
+#ifdef OS_LINUX
+#define DEFAULT_DIRECTORY "/home/"
+#include "linux/thread.c"
+#include "linux/platform.c"
+#endif
+
+#ifdef OS_WIN
+#define DEFAULT_DIRECTORY "C:/"
+#include "windows/thread.c"
+#include "windows/platform.c"
+#endif
+
+#include "input.c"
+#include "array.c"
+#include "assets.c"
+#include "render.c"
+#include "camera.c"
+#include "ui.c"
+#include "string_utils.c"
+#include "settings_config.c"
+#include "localization.c"
+#include "memory_bucket.c"
+#include "command_line.c"
+
+#include "external/cJSON.h"
+#include "external/cJSON.c"
+
+#endif
\ No newline at end of file
diff --git a/src/render.c b/src/render.c
new file mode 100644
index 0000000..e126133
--- /dev/null
+++ b/src/render.c
@@ -0,0 +1,446 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+inline void render_clear()
+{
+	glClearColor(255/255.0, 255/255.0, 255/255.0, 1.0);
+	glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+}
+
+inline void render_set_rotation(float32 rotation, float32 x, float32 y, s32 depth)
+{
+	glRotatef(rotation, x, y, depth);
+}
+
+inline void set_render_depth(s32 depth)
+{
+	render_depth = depth;
+}
+
+void render_image(image *image, s32 x, s32 y, s32 width, s32 height)
+{
+	assert(image);
+	if (image->loaded)
+	{
+		glBindTexture(GL_TEXTURE_2D, image->textureID);
+		glEnable(GL_TEXTURE_2D);
+		glBegin(GL_QUADS);
+		glColor4f(1., 1., 1., 1.);
+		glTexCoord2i(0, 0); glVertex3i(x, y, render_depth);
+		glTexCoord2i(0, 1); glVertex3i(x, y+height, render_depth);
+		glTexCoord2i(1, 1); glVertex3i(x+width, y+height, render_depth);
+		glTexCoord2i(1, 0); glVertex3i(x+width, y, render_depth);
+		glEnd();
+		
+		glDisable(GL_TEXTURE_2D);
+	}
+}
+
+void render_image_tint(image *image, s32 x, s32 y, s32 width, s32 height, color tint)
+{
+	assert(image);
+	if (image->loaded)
+	{
+		glBindTexture(GL_TEXTURE_2D, image->textureID);
+		glEnable(GL_TEXTURE_2D);
+		glBegin(GL_QUADS);
+		glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+		glTexCoord2i(0, 0); glVertex3i(x, y, render_depth);
+		glTexCoord2i(0, 1); glVertex3i(x, y+height, render_depth);
+		glTexCoord2i(1, 1); glVertex3i(x+width, y+height, render_depth);
+		glTexCoord2i(1, 0); glVertex3i(x+width, y, render_depth);
+		glEnd();
+		
+		glDisable(GL_TEXTURE_2D);
+	}
+}
+
+s32 render_text_ellipsed(font *font, s32 x, s32 y, s32 maxw, char *text, color tint)
+{
+	if (!font->loaded)
+		return 0;
+	
+	glEnable(GL_TEXTURE_2D);
+	glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+	
+	char *ellipse = "...";
+	bool in_ellipse = false;
+	
+	s32 x_ = x;
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		glyph g = font->glyphs[ch];
+		
+		glBindTexture(GL_TEXTURE_2D, g.textureID);
+		glBegin(GL_QUADS);
+		
+		s32 width = g.width;
+		
+		s32 y_ = y + font->px_h + g.yoff;
+		
+		glTexCoord2i(0, 0); glVertex3i(x_,y_, render_depth);
+		glTexCoord2i(0, 1); glVertex3i(x_,y_+g.height, render_depth);
+		glTexCoord2i(1, 1); glVertex3i(x_+g.width,y_+g.height, render_depth);
+		glTexCoord2i(1, 0); glVertex3i(x_+g.width,y_, render_depth);
+		
+		glEnd();
+		
+		/* add kerning */
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x_ += kern * font->scale;
+		x_ += g.width+g.xoff;
+		
+		if (!in_ellipse && (x_-x) > maxw-(font->glyphs['.'].width*3))
+		{
+			in_ellipse = true;
+			text = ellipse;
+		}
+	}
+	
+	glDisable(GL_TEXTURE_2D);
+	
+	return maxw;
+}
+
+s32 render_text(font *font, s32 x, s32 y, char *text, color tint)
+{
+	if (!font->loaded)
+		return 0;
+	
+	glEnable(GL_TEXTURE_2D);
+	glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+	
+	s32 x_ = x;
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		glyph g = font->glyphs[ch];
+		
+		glBindTexture(GL_TEXTURE_2D, g.textureID);
+		glBegin(GL_QUADS);
+		
+		s32 width = g.width;
+		
+		s32 y_ = y + font->px_h + g.yoff;
+		
+		glTexCoord2i(0, 0); glVertex3i(x_,y_, render_depth);
+		glTexCoord2i(0, 1); glVertex3i(x_,y_+g.height, render_depth);
+		glTexCoord2i(1, 1); glVertex3i(x_+g.width,y_+g.height, render_depth);
+		glTexCoord2i(1, 0); glVertex3i(x_+g.width,y_, render_depth);
+		
+		glEnd();
+		
+		/* add kerning */
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x_ += kern * font->scale;
+		x_ += g.width+g.xoff;
+	}
+	
+	glDisable(GL_TEXTURE_2D);
+	
+	return x_ - x;
+}
+
+s32 render_text_cutoff(font *font, s32 x, s32 y, char *text, color tint, u16 cutoff_width)
+{
+	if (!font->loaded)
+		return 0;
+	
+	glEnable(GL_TEXTURE_2D);
+	glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+	
+	s32 x_ = x;
+	s32 y_ = y;
+	bool is_new_line = false;
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		if (ch == '\n')
+		{
+			x_ = x;
+			y_ += font->size;
+			is_new_line = true;
+			continue;
+		}
+		
+		if (is_new_line && ch == ' ')
+		{
+			is_new_line = false;
+			continue;
+		}
+		else if (is_new_line && ch != ' ')
+		{
+			is_new_line = false;
+		}
+		
+		
+		glyph g = font->glyphs[ch];
+		
+		glBindTexture(GL_TEXTURE_2D, g.textureID);
+		glBegin(GL_QUADS);
+		
+		s32 width = g.width;
+		
+		s32 y__ = y_ + font->px_h + g.yoff;
+		
+		glTexCoord2i(0, 0); glVertex3i(x_,y__, render_depth);
+		glTexCoord2i(0, 1); glVertex3i(x_,y__+g.height, render_depth);
+		glTexCoord2i(1, 1); glVertex3i(x_+g.width,y__+g.height, render_depth);
+		glTexCoord2i(1, 0); glVertex3i(x_+g.width,y__, render_depth);
+		
+		glEnd();
+		
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x_ += kern * font->scale;
+		
+		x_ += g.width+g.xoff;
+		if (x_ > x+cutoff_width)
+		{
+			x_ = x;
+			y_ += font->size;
+			is_new_line = true;
+		}
+	}
+	
+	glDisable(GL_TEXTURE_2D);
+	
+	return (y_ - y) + font->size;
+	
+}
+
+s32 calculate_cursor_position(font *font, char *text, s32 click_x)
+{
+	if (!font->loaded)
+		return 0;
+	
+	s32 x = 0;
+	s32 index = 0;
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		
+		glyph g = font->glyphs[ch];
+		
+		s32 width = g.width;
+		s32 width_next = font->glyphs[ch_next].width;
+		
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x += kern * font->scale;
+		
+		x += g.width+g.xoff;
+		
+		if (x - (width_next/5) > click_x)
+		{
+			return index;
+		}
+		
+		++index;
+	}
+	
+	return index;
+}
+
+s32 calculate_text_width_from_upto(font *font, char *text, s32 from, s32 index)
+{
+	if (!font->loaded)
+		return 0;
+	
+	s32 x = 0;
+	utf8_int32_t ch;
+	s32 i = 0;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (index == i) return x;
+		
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		glyph g = font->glyphs[ch];
+		s32 width = g.width;
+		
+		if (i >= from)
+		{
+			int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+			if (kern != 0) x += kern * font->scale;
+			
+			x += g.width+g.xoff;
+		}
+		
+		i++;
+	}
+	
+	return x;
+}
+
+s32 calculate_text_width_upto(font *font, char *text, s32 index)
+{
+	if (!font->loaded)
+		return 0;
+	
+	s32 x = 0;
+	utf8_int32_t ch;
+	s32 i = 0;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (index == i) return x;
+		
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		glyph g = font->glyphs[ch];
+		s32 width = g.width;
+		
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x += kern * font->scale;
+		
+		x += g.width+g.xoff;
+		
+		i++;
+	}
+	
+	return x;
+}
+
+s32 calculate_text_width(font *font, char *text)
+{
+	if (!font->loaded)
+		return 0;
+	
+	s32 x = 0;
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch == 9) ch = 32;
+		utf8_int32_t ch_next;
+		utf8codepoint(text, &ch_next);
+		if (ch < TEXT_CHARSET_START || ch > TEXT_CHARSET_END) 
+		{
+			ch = 0x3f;
+		}
+		
+		glyph g = font->glyphs[ch];
+		s32 width = g.width;
+		
+		int kern = stbtt_GetCodepointKernAdvance(&font->info, ch, ch_next);
+		if (kern != 0) x += kern * font->scale;
+		
+		x += g.width+g.xoff;
+	}
+	
+	return x;
+}
+
+void render_triangle(s32 x, s32 y, s32 w, s32 h, color tint)
+{
+	glBegin(GL_TRIANGLES);
+	glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+	glVertex3i(x+(w/2), y+h, render_depth);
+	glVertex3i(x, y, render_depth);
+	glVertex3i(x+w, y, render_depth);
+	glEnd();
+}
+
+void render_rectangle(s32 x, s32 y, s32 width, s32 height, color tint)
+{
+	glBegin(GL_QUADS);
+	glColor4f(tint.r/255.0f, tint.g/255.0f, tint.b/255.0f, tint.a/255.0f); 
+	glVertex3i(x, y, render_depth);
+	glVertex3i(x, y+height, render_depth);
+	glVertex3i(x+width, y+height, render_depth);
+	glVertex3i(x+width, y, render_depth);
+	glEnd();
+}
+
+void render_rectangle_tint(s32 x, s32 y, s32 width, s32 height, color tint[4])
+{
+	glBegin(GL_QUADS);
+	glColor4f(tint[0].r/255.0f, tint[0].g/255.0f, tint[0].b/255.0f, tint[0].a/255.0f);
+	glVertex3i(x, y, render_depth);
+	glColor4f(tint[1].r/255.0f, tint[1].g/255.0f, tint[1].b/255.0f, tint[1].a/255.0f);
+	glVertex3i(x, y+height, render_depth);
+	glColor4f(tint[2].r/255.0f, tint[2].g/255.0f, tint[2].b/255.0f, tint[2].a/255.0f); 
+	glVertex3i(x+width, y+height, render_depth);
+	glColor4f(tint[3].r/255.0f, tint[3].g/255.0f, tint[3].b/255.0f, tint[3].a/255.0f);
+	glVertex3i(x+width, y, render_depth);
+	glEnd();
+}
+
+void render_rectangle_outline(s32 x, s32 y, s32 width, s32 height, u16 outline_w, color tint)
+{
+	// left
+	render_rectangle(x, y, outline_w, height, tint);
+	// right
+	render_rectangle(x+width-outline_w, y, outline_w, height, tint);
+	// top
+	render_rectangle(x+outline_w, y, width-(outline_w*2), outline_w, tint);
+	// bottom
+	render_rectangle(x+outline_w, y+height-outline_w, width-(outline_w*2), outline_w, tint);
+}
+
+void render_set_scissor(platform_window *window, s32 x, s32 y, s32 w, s32 h)
+{
+	glEnable(GL_SCISSOR_TEST);
+	glScissor(x-1, window->height-h-y-1, w+1, h+1);
+}
+
+vec4 render_get_scissor()
+{
+	vec4 vec;
+	glGetIntegerv(GL_SCISSOR_BOX, (GLint*)(&vec));
+	vec.x += 1;
+	vec.y += 1;
+	vec.w -= 1;
+	vec.h -= 1;
+	return vec;
+}
+
+void render_reset_scissor()
+{
+	glDisable(GL_SCISSOR_TEST);
+}
\ No newline at end of file
diff --git a/src/render.h b/src/render.h
new file mode 100644
index 0000000..4d29eea
--- /dev/null
+++ b/src/render.h
@@ -0,0 +1,61 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_RENDER
+#define INCLUDE_RENDER
+
+typedef struct t_color {
+	u8 r;
+	u8 g;
+	u8 b;
+	u8 a;
+} color;
+
+typedef struct t_vec4
+{
+	s32 x;
+	s32 y;
+	s32 w;
+	s32 h;
+} vec4;
+
+s32 render_depth = 1;
+void set_render_depth(s32 depth);
+
+#define rgb(r_,g_,b_) (color){ r_, g_, b_, 255 }
+#define rgba(r_,g_,b_,a_) (color){r_,g_,b_,a_}
+
+void render_clear();
+
+// images
+void render_image(image *image, s32 x, s32 y, s32 width, s32 height);
+void render_image_tint(image *image, s32 x, s32 y, s32 width, s32 height, color tint);
+
+// text
+s32 render_text(font *font, s32 x, s32 y, char *text, color tint);
+s32 render_text_ellipsed(font *font, s32 x, s32 y, s32 maxw, char *text, color tint);
+s32 render_text_cutoff(font *font, s32 x, s32 y, char *text, color tint, u16 cutoff_width);
+s32 render_text_vertical(font *font, s32 x, s32 y, char *text, color tint);
+
+s32 calculate_cursor_position(font *font, char *text, s32 click_x);
+s32 calculate_text_width(font *font, char *text);
+s32 calculate_text_width_upto(font *font, char *text, s32 index);
+s32 calculate_text_width_from_upto(font *font, char *text, s32 from, s32 index);
+
+// primitives
+void render_rectangle(s32 x, s32 y, s32 width, s32 height, color tint);
+void render_rectangle_tint(s32 x, s32 y, s32 width, s32 height, color tint[4]);
+void render_rectangle_outline(s32 x, s32 y, s32 width, s32 height, u16 outline_w, color tint);
+void render_triangle(s32 x, s32 y, s32 w, s32 h, color tint);
+
+// utils
+void render_set_scissor(platform_window *window, s32 x, s32 y, s32 w, s32 h);
+vec4 render_get_scissor();
+void render_reset_scissor();
+
+void render_set_rotation(float32 rotation, float32 x, float32 y, s32 depth);
+
+#endif
\ No newline at end of file
diff --git a/src/save.c b/src/save.c
new file mode 100644
index 0000000..82ca623
--- /dev/null
+++ b/src/save.c
@@ -0,0 +1,380 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+search_result *create_empty_search_result();
+void* destroy_search_result_thread(void *arg);
+
+static void write_json_file(char *buffer, s32 length, search_result *search_result)
+{
+	array matches = search_result->matches;
+	
+	cJSON *result = cJSON_CreateObject();
+	if (cJSON_AddStringToObject(result, "search_directory", 
+								search_result->directory_to_search) == NULL)
+		return;
+	
+	if (cJSON_AddStringToObject(result, "filter", 
+								search_result->file_filter) == NULL)
+		return;
+	
+	if (cJSON_AddStringToObject(result, "search_query", 
+								search_result->text_to_find) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "duration_us", 
+								search_result->find_duration_us) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "show_error", 
+								search_result->show_error_message) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "file_match_found", 
+								search_result->found_file_matches) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "files_searched", 
+								search_result->files_searched) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "files_matched", 
+								search_result->files_matched) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "query_match_found", 
+								search_result->match_found) == NULL)
+		return;
+	
+	if (cJSON_AddNumberToObject(result, "recursive_search", 
+								search_result->is_recursive) == NULL)
+		return;
+	
+	cJSON *match_list = cJSON_AddArrayToObject(result, "match_list");
+	
+	if (!match_list) return;
+	
+	for (s32 i = 0; i < matches.length; i++)
+	{
+		file_match* m = array_at(&matches, i);
+		
+		cJSON *item = cJSON_CreateObject();
+		
+		if (cJSON_AddStringToObject(item, "path", 
+									m->file.path) == NULL)
+			return;
+		
+		if (cJSON_AddStringToObject(item, "matched_filter", 
+									m->file.matched_filter) == NULL)
+			return;
+		
+		if (cJSON_AddNumberToObject(item, "word_offset", 
+									m->word_match_offset) == NULL)
+			return;
+		
+		if (cJSON_AddNumberToObject(item, "word_length", 
+									m->word_match_length) == NULL)
+			return;
+		
+		if (cJSON_AddNumberToObject(item, "file_error", 
+									m->file_error) == NULL)
+			return;
+		
+		if (cJSON_AddNumberToObject(item, "line_nr", 
+									m->line_nr) == NULL)
+			return;
+		
+		if (cJSON_AddNumberToObject(item, "file_size", 
+									m->file_size) == NULL)
+			return;
+		
+		if (m->line_info)
+		{
+			if (cJSON_AddStringToObject(item, "line_info", 
+										m->line_info) == NULL)
+				return;
+		}
+		else
+		{
+			if (cJSON_AddNumberToObject(item, "line_info", 0) == NULL)
+				return;
+		}
+		
+		cJSON_AddItemToArray(match_list, item);
+	}
+	
+	cJSON_PrintPreallocated(result, buffer, length, true);
+	cJSON_Delete(result);
+}
+
+static void *export_result_d(void *arg)
+{
+	search_result *search_result = arg;
+	
+	array matches = search_result->files;
+	
+	char path_buf[MAX_INPUT_LENGTH];
+	path_buf[0] = 0;
+	
+	char start_path[MAX_INPUT_LENGTH];
+	snprintf(start_path, MAX_INPUT_LENGTH, "%s%s", binary_path, "");
+	
+	char default_save_file_extension[50];
+	string_copyn(default_save_file_extension, "json", 50);
+	
+	if (!search_result->is_command_line_search)
+	{
+		struct open_dialog_args *args = mem_alloc(sizeof(struct open_dialog_args));
+		args->buffer = path_buf;
+		args->type = SAVE_FILE;
+		args->file_filter = SEARCH_RESULT_FILE_EXTENSION;
+		args->start_path = start_path;
+		args->default_save_file_extension = default_save_file_extension;
+		
+		platform_open_file_dialog_block(args);
+	}
+	else
+	{
+		string_copyn(path_buf, search_result->export_path, MAX_INPUT_LENGTH);
+	}
+	
+	char tmp_dir_buffer[MAX_INPUT_LENGTH];
+	get_directory_from_path(tmp_dir_buffer, path_buf);
+	
+	char tmp_name_buffer[MAX_INPUT_LENGTH];
+	get_name_from_path(tmp_name_buffer, path_buf);
+	
+	if (string_equals(path_buf, "")) return 0;
+	if (string_equals(tmp_name_buffer, "")) return 0;
+	if (!platform_directory_exists(tmp_dir_buffer)) return 0;
+	
+	s32 size = matches.length * (MAX_INPUT_LENGTH*10);
+	char *buffer = mem_alloc(size);
+	memset(buffer, 0, size);
+	
+	char *file_extension = get_file_extension(path_buf);
+	if (string_equals(file_extension, ".json") || string_equals(file_extension, ""))
+	{
+		write_json_file(buffer, size, search_result);
+	}
+	
+	if (string_equals(file_extension, ""))
+	{
+		string_appendn(path_buf, ".json", MAX_INPUT_LENGTH);
+	}
+	
+	platform_write_file_content(path_buf, "w", buffer, size);
+	
+	return 0;
+}
+
+bool export_results(search_result *search_result)
+{
+	thread thr;
+	thr.valid = false;
+	
+	while (!thr.valid)
+		thr = thread_start(export_result_d, search_result);
+	
+	if (!search_result->is_command_line_search)
+		thread_detach(&thr);
+	else
+		thread_join(&thr);
+	
+	return true;
+}
+
+static bool read_json_file(char *buffer, s32 size, search_result *search_result)
+{
+	cJSON *result = cJSON_Parse(buffer);
+	if (!result) return false;
+	
+	cJSON *search_directory = cJSON_GetObjectItemCaseSensitive(result, "search_directory");
+	string_copyn(textbox_path.buffer, search_directory->valuestring, MAX_INPUT_LENGTH);
+	string_copyn(search_result->directory_to_search, search_directory->valuestring, MAX_INPUT_LENGTH);
+	
+	cJSON *filter = cJSON_GetObjectItemCaseSensitive(result, "filter");
+	string_copyn(textbox_file_filter.buffer, filter->valuestring, MAX_INPUT_LENGTH);
+	string_copyn(search_result->file_filter, filter->valuestring, MAX_INPUT_LENGTH);
+	
+	cJSON *search_query = cJSON_GetObjectItemCaseSensitive(result, "search_query");
+	string_copyn(textbox_search_text.buffer, search_query->valuestring, MAX_INPUT_LENGTH);
+	string_copyn(search_result->text_to_find, search_query->valuestring, MAX_INPUT_LENGTH);
+	
+	cJSON *duration_us = cJSON_GetObjectItemCaseSensitive(result, "duration_us");
+	search_result->find_duration_us = duration_us->valueint;
+	
+	cJSON *show_error = cJSON_GetObjectItemCaseSensitive(result, "show_error");
+	search_result->show_error_message = show_error->valueint;
+	
+	cJSON *file_match_found = cJSON_GetObjectItemCaseSensitive(result, "file_match_found");
+	search_result->found_file_matches = file_match_found->valueint;
+	
+	cJSON *files_searched = cJSON_GetObjectItemCaseSensitive(result, "files_searched");
+	search_result->files_searched = files_searched->valueint;
+	
+	cJSON *files_matched = cJSON_GetObjectItemCaseSensitive(result, "files_matched");
+	search_result->files_matched = files_matched->valueint;
+	
+	cJSON *query_match_found = cJSON_GetObjectItemCaseSensitive(result, "query_match_found");
+	search_result->match_found = query_match_found->valueint;
+	
+	cJSON *recursive = cJSON_GetObjectItemCaseSensitive(result, "recursive_search");
+	search_result->is_recursive = recursive->valueint;
+	checkbox_recursive.state = search_result->is_recursive;
+	
+	search_result->search_result_source_dir_len = strlen(search_result->directory_to_search);
+	
+	cJSON *file_list = cJSON_GetObjectItem(result, "match_list");
+	cJSON *file;
+	cJSON_ArrayForEach(file, file_list)
+	{
+		file_match new_match;
+		
+		////
+		cJSON *path = cJSON_GetObjectItem(file, "path");
+		new_match.file.path = memory_bucket_reserve(&search_result->mem_bucket, strlen(path->valuestring)+1);
+		string_copyn(new_match.file.path, path->valuestring, strlen(path->valuestring)+1);
+		
+		////
+        cJSON *matched_filter = cJSON_GetObjectItem(file, "matched_filter");
+		new_match.file.matched_filter = memory_bucket_reserve(&search_result->mem_bucket, strlen(matched_filter->valuestring)+1);
+		string_copyn(new_match.file.matched_filter, matched_filter->valuestring, strlen(matched_filter->valuestring)+1);
+		
+		////
+		cJSON *word_offset = cJSON_GetObjectItem(file, "word_offset");
+		new_match.word_match_offset = word_offset->valueint;
+		
+		////
+		cJSON *word_length = cJSON_GetObjectItem(file, "word_length");
+		new_match.word_match_length = word_length->valueint;
+		
+		////
+		cJSON *file_error = cJSON_GetObjectItem(file, "file_error");
+		new_match.file_error = file_error->valueint;
+		
+		////
+		cJSON *line_nr = cJSON_GetObjectItem(file, "line_nr");
+		new_match.line_nr = line_nr->valueint;
+		
+		////
+		cJSON *file_size = cJSON_GetObjectItem(file, "file_size");
+		new_match.file_size = file_size->valueint;
+		
+		////
+		cJSON *line_info = cJSON_GetObjectItem(file, "line_info");
+		if (cJSON_IsString(line_info))
+		{
+			new_match.line_info = memory_bucket_reserve(&search_result->mem_bucket, strlen(line_info->valuestring)+1);
+			string_copyn(new_match.line_info, line_info->valuestring, strlen(line_info->valuestring)+1);
+			search_result->match_found = true;
+		}
+		else
+		{
+			new_match.line_info = 0;
+		}
+		
+		// calculate highlight offsets
+		if (new_match.line_info)
+		{
+			new_match.word_match_offset_x = 
+				calculate_text_width_upto(font_mini, new_match.line_info, new_match.word_match_offset);
+			
+			new_match.word_match_width =
+				calculate_text_width_from_upto(font_mini, new_match.line_info, new_match.word_match_offset, new_match.word_match_offset + new_match.word_match_length);
+		}
+		
+		array_push(&search_result->matches, &new_match);
+	}
+	
+	return true;
+}
+
+void import_results_from_file(char *path_buf)
+{
+	char *file_extension = get_file_extension(path_buf);
+	if (!string_equals(file_extension, ".json") && !string_equals(file_extension, ".xml") && !string_equals(file_extension, ".yaml"))
+	{
+		platform_show_message(main_window, localize("invalid_search_result_file"), localize("error_importing_results"));
+		return;
+	}
+	
+	scroll_y = 0;
+	file_content content = platform_read_file_content(path_buf, "r");
+	
+	if (!content.content || content.file_error)
+	{
+		platform_destroy_file_content(&content);
+		return;
+	}
+	
+	search_result *new_result = create_empty_search_result();
+	search_result *old_result = current_search_result;
+	current_search_result = new_result;
+	
+	thread cleanup_thread = thread_start(destroy_search_result_thread, old_result);
+	thread_detach(&cleanup_thread);
+	
+	if (string_equals(file_extension, ".json"))
+	{
+		bool result = read_json_file(content.content, content.content_length, new_result);
+		if (!result) goto failed_to_load_file;
+	}
+	else
+	{
+		goto failed_to_load_file;
+	}
+	
+	new_result->walking_file_system = false;
+	new_result->done_finding_matches = true;
+	new_result->done_finding_files = true;
+	
+	snprintf(global_status_bar.result_status_text, MAX_INPUT_LENGTH, localize("files_matches_comparison"), current_search_result->matches.length, current_search_result->files_searched, current_search_result->find_duration_us/1000.0);
+	
+	array_destroy(&new_result->files);
+	platform_destroy_file_content(&content);
+	return;
+	
+	failed_to_load_file:
+	platform_show_message(main_window, localize("invalid_search_result_file"), localize("error_importing_results"));
+	platform_destroy_file_content(&content);
+}
+
+static void* import_results_d(void *arg)
+{
+	char path_buf[MAX_INPUT_LENGTH];
+	path_buf[0] = 0;
+	
+	char start_path[MAX_INPUT_LENGTH];
+	snprintf(start_path, MAX_INPUT_LENGTH, "%s%s", binary_path, "");
+	
+	char default_save_file_extension[50];
+	string_copyn(default_save_file_extension, "json", 50);
+	
+	struct open_dialog_args *args = mem_alloc(sizeof(struct open_dialog_args));
+	args->buffer = path_buf;
+	args->type = OPEN_FILE;
+	args->file_filter = SEARCH_RESULT_FILE_EXTENSION;
+	args->start_path = start_path;
+	args->default_save_file_extension = default_save_file_extension;
+	
+	platform_open_file_dialog_block(args);
+	
+	if (string_equals(path_buf, "")) return 0;
+	if (!platform_file_exists(path_buf)) return 0;
+	
+	import_results_from_file(path_buf);
+	return 0;
+}
+
+void import_results()
+{
+	thread thr;
+	thr.valid = false;
+	
+	while (!thr.valid)
+		thr = thread_start(import_results_d, 0);
+	thread_detach(&thr);
+}
\ No newline at end of file
diff --git a/src/save.h b/src/save.h
new file mode 100644
index 0000000..0772daa
--- /dev/null
+++ b/src/save.h
@@ -0,0 +1,16 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_SAVE
+#define INCLUDE_SAVE
+
+#define SEARCH_RESULT_FILE_EXTENSION "*.json"
+
+bool export_results(search_result *result);
+void import_results_from_file(char *path_buf);
+void import_results();
+
+#endif
\ No newline at end of file
diff --git a/src/settings.c b/src/settings.c
new file mode 100644
index 0000000..b379470
--- /dev/null
+++ b/src/settings.c
@@ -0,0 +1,313 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+void reset_status_text();
+void set_status_text_to_finished_search();
+
+void settings_page_create()
+{
+	global_settings_page.active = false;
+	global_settings_page.font_small = assets_load_font(_binary____data_fonts_mono_ttf_start,
+													   _binary____data_fonts_mono_ttf_end, 16);
+	global_settings_page.logo_img = assets_load_image(_binary____data_imgs_logo_64_png_start,
+													  _binary____data_imgs_logo_64_png_end, true);
+	global_settings_page.keyboard = keyboard_input_create();
+	global_settings_page.mouse = mouse_input_create();
+    
+	camera cam;
+	cam.x = 0;
+	cam.y = 0;
+	cam.rotation = 0;
+	
+	global_settings_page.camera = cam;
+    
+	global_settings_page.btn_close = ui_create_button();
+	global_settings_page.btn_save = ui_create_button();
+	global_settings_page.dropdown_language = ui_create_dropdown();
+	global_settings_page.dropdown_doubleclick = ui_create_dropdown();
+	global_settings_page.textbox_max_file_size = ui_create_textbox(9);
+	global_settings_page.textbox_max_thread_count = ui_create_textbox(5);
+	global_settings_page.checkbox_parallelize_search = ui_create_checkbox(false);
+}
+
+static void load_current_settings_into_ui()
+{
+	if (global_settings_page.max_thread_count != 0)
+		snprintf(global_settings_page.textbox_max_thread_count.buffer, global_settings_page.textbox_max_thread_count.max_len, "%d",global_settings_page.max_thread_count);
+	
+	if (global_settings_page.max_file_size != 0)
+		snprintf(global_settings_page.textbox_max_file_size.buffer, global_settings_page.textbox_max_file_size.max_len, "%d", global_settings_page.max_file_size);
+}
+
+void settings_page_update_render()
+{
+	if (global_settings_page.active)
+	{
+		platform_window_make_current(&global_settings_page.window);
+		platform_handle_events(&global_settings_page.window, &global_settings_page.mouse, &global_settings_page.keyboard);
+		platform_set_cursor(&global_settings_page.window, CURSOR_DEFAULT);
+		
+		render_clear();
+		
+        camera_apply_transformations(&global_settings_page.window, &global_settings_page.camera);
+        
+		global_ui_context.layout.active_window = &global_settings_page.window;
+		global_ui_context.keyboard = &global_settings_page.keyboard;
+		global_ui_context.mouse = &global_settings_page.mouse;
+		
+		ui_begin(3);
+		{
+			ui_begin_menu_bar();
+			{
+				if (ui_push_menu(localize("general")))
+				{
+					global_settings_page.selected_tab_index = 0;
+				}
+				if (ui_push_menu(localize("interface")))
+				{
+					global_settings_page.selected_tab_index = 1;
+				}
+			}
+			ui_end_menu_bar();
+			
+			ui_push_separator();
+			
+			if (global_settings_page.selected_tab_index == 0)
+			{
+				/////////////////////////////////////
+				// max file size
+				/////////////////////////////////////
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_text(localize("max_file_size"));
+					ui_push_text(localize("zero_for_no_limit"));
+				}
+				ui_block_end();
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					if (ui_push_textbox(&global_settings_page.textbox_max_file_size, "0"))
+					{
+						keyboard_set_input_mode(&global_settings_page.keyboard, INPUT_NUMERIC);
+					}
+					ui_push_text("KB");
+				}
+				ui_block_end();
+				
+				/////////////////////////////////////
+				// max threads
+				/////////////////////////////////////
+				global_ui_context.layout.offset_y += 10;
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_text(localize("max_threads"));
+					ui_push_text(localize("minimum_of_1"));
+				}
+				ui_block_end();
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					if (ui_push_textbox(&global_settings_page.textbox_max_thread_count, "0"))
+					{
+						keyboard_set_input_mode(&global_settings_page.keyboard, INPUT_NUMERIC);
+					}
+					ui_push_text("Threads");
+					
+				}
+				ui_block_end();
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					if (ui_push_hypertext_link(localize("copy_config_path")))
+					{
+						char buffer[PATH_MAX];
+						platform_set_clipboard(main_window, get_config_save_location(buffer));
+					}
+				}
+				ui_block_end();
+			}
+			else if (global_settings_page.selected_tab_index == 1)
+			{
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_text(localize("language"));
+				}
+				ui_block_end();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					if (ui_push_dropdown(&global_settings_page.dropdown_language, locale_get_name()))
+					{
+						for (s32 i = 0; i < global_localization.mo_files.length; i++)
+						{
+							mo_file *file = array_at(&global_localization.mo_files, i);
+							
+							if (ui_push_dropdown_item(file->icon, file->locale_full, i))
+							{
+								set_locale(file->locale_id);
+							}
+						}
+					}
+				}
+				ui_block_end();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_text(localize("double_click_action"));
+				}
+				ui_block_end();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					char* available_options[OPTION_RESULT+1] = {
+						localize("double_click_action_1"),
+						localize("double_click_action_2"),
+						localize("double_click_action_3"),
+						localize("double_click_action_4"),
+					};
+					
+					if (ui_push_dropdown(&global_settings_page.dropdown_doubleclick, available_options[global_settings_page.current_double_click_selection_option]))
+					{
+						for (s32 i = 0; i < OPTION_RESULT+1; i++)
+						{
+							if (ui_push_dropdown_item(0, available_options[i], i))
+							{
+								global_settings_page.current_double_click_selection_option = i;
+							}
+						}
+					}
+				}
+				ui_block_end();
+				
+#if 0
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					ui_push_text("Style");
+				}
+				ui_block_end();
+				
+				ui_block_begin(LAYOUT_HORIZONTAL);
+				{
+					if (ui_push_color_button("Light", global_ui_context.style.id == UI_STYLE_LIGHT, rgb(250, 250, 250)))
+					{
+						ui_set_style(UI_STYLE_LIGHT);
+					}
+					if (ui_push_color_button("Dark", global_ui_context.style.id == UI_STYLE_DARK, rgb(50, 50, 50)))
+					{
+						ui_set_style(UI_STYLE_DARK);
+					}
+				}
+				ui_block_end();
+#endif
+			}
+			
+			global_ui_context.layout.offset_y = global_settings_page.window.height - 33;
+			
+			ui_block_begin(LAYOUT_HORIZONTAL);
+			{
+				if (ui_push_button(&global_settings_page.btn_close, localize("close")))
+				{
+					global_settings_page.textbox_max_thread_count.buffer[0] = 0; 
+					global_settings_page.textbox_max_file_size.buffer[0] = 0; 
+					ui_set_style(global_settings_page.current_style);
+					global_settings_page.current_double_click_selection_option = global_settings_page.selected_double_click_selection_option;
+					global_settings_page.active = false;
+					set_locale(global_settings_page.current_locale_id);
+					settings_page_hide();
+					
+					return;
+				}
+				if (ui_push_button(&global_settings_page.btn_close, localize("save")))
+				{
+					global_settings_page.current_style = global_ui_context.style.id;
+					global_settings_page.max_thread_count = string_to_s32(global_settings_page.textbox_max_thread_count.buffer);
+					if (global_settings_page.max_thread_count < 1)
+						global_settings_page.max_thread_count = DEFAULT_THREAD_COUNT;
+					
+					global_settings_page.max_file_size = string_to_s32(global_settings_page.textbox_max_file_size.buffer);
+					
+					global_settings_page.textbox_max_thread_count.buffer[0] = 0; 
+					global_settings_page.textbox_max_file_size.buffer[0] = 0;
+					global_settings_page.selected_double_click_selection_option = global_settings_page.current_double_click_selection_option;
+					
+					global_settings_page.active = false;
+					settings_page_hide();
+					return;
+				}
+			}
+			ui_block_end();
+		}
+		ui_end();
+		
+		if (!global_settings_page.window.is_open)
+		{
+			global_settings_page.active = false;
+			settings_page_hide();
+			return;
+		}
+		
+		platform_window_swap_buffers(&global_settings_page.window);
+	}
+}
+
+void settings_page_show()
+{
+	if (platform_window_is_valid(&global_settings_page.window)) return;
+	
+	load_current_settings_into_ui();
+	
+	global_settings_page.window = platform_open_window(localize("text_search_settings"), 
+													   450, 280, 450, 280, 450, 280);
+	
+	settings_window = &global_settings_page.window;
+	
+	global_settings_page.active = true;
+	global_settings_page.selected_tab_index = 0;
+	global_settings_page.current_locale_id = locale_get_id();
+	global_settings_page.current_double_click_selection_option = global_settings_page.selected_double_click_selection_option;
+	
+	platform_set_icon(&global_settings_page.window, global_settings_page.logo_img);
+}
+
+void settings_page_hide()
+{
+	if (platform_window_is_valid(&global_settings_page.window))
+	{
+		settings_window = 0;
+		
+		platform_destroy_window(&global_settings_page.window);
+		
+		global_settings_page.btn_close.state = false;
+		global_settings_page.btn_save.state = false;
+		global_settings_page.active = false;
+		
+		global_settings_page.mouse.x = -1;
+		global_settings_page.mouse.y = -1;
+	}
+}
+
+void settings_page_hide_without_save()
+{
+	if (platform_window_is_valid(&global_settings_page.window))
+	{
+		global_settings_page.textbox_max_thread_count.buffer[0] = 0; 
+		global_settings_page.textbox_max_file_size.buffer[0] = 0; 
+		
+		global_settings_page.active = false;
+		set_locale(global_settings_page.current_locale_id);
+		settings_page_hide();
+	}
+}
+
+void settings_page_destroy()
+{
+	keyboard_input_destroy(&global_settings_page.keyboard);
+	ui_destroy_textbox(&global_settings_page.textbox_max_file_size);
+	ui_destroy_textbox(&global_settings_page.textbox_max_thread_count);
+	assets_destroy_font(global_settings_page.font_small);
+	assets_destroy_image(global_settings_page.logo_img);
+	
+	if (platform_window_is_valid(&global_settings_page.window))
+		platform_destroy_window(&global_settings_page.window);
+}
diff --git a/src/settings.h b/src/settings.h
new file mode 100644
index 0000000..11ab307
--- /dev/null
+++ b/src/settings.h
@@ -0,0 +1,60 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_SETTINGS
+#define INCLUDE_SETTINGS
+
+typedef enum t_double_click_option
+{
+	OPTION_PATH,
+	OPTION_PATH_LINE,
+	OPTION_PATH_LINE_FILTER,
+	OPTION_RESULT,
+} double_click_option;
+
+typedef struct t_settings_page
+{
+	platform_window window;
+	keyboard_input keyboard;
+	mouse_input mouse;
+    camera camera;
+	bool active;
+	
+	font *font_small;
+	image *logo_img;
+	
+	button_state btn_close;
+	button_state btn_save;
+	dropdown_state dropdown_language;
+	dropdown_state dropdown_doubleclick;
+	textbox_state textbox_max_file_size;
+	textbox_state textbox_max_thread_count;
+	checkbox_state checkbox_parallelize_search;
+	s32 selected_tab_index;
+	
+	char *current_locale_id;
+	s32 max_thread_count;
+	s32 max_file_size;
+	u16 current_style;
+	u16 selected_double_click_selection_option; // saved state
+	u16 current_double_click_selection_option; // unsaved state 
+} settings_page;
+
+#define DEFAULT_THREAD_COUNT 10
+#define DEFAULT_MAX_FILE_SIZE 0
+#define DEFAULT_RECURSIVE_STATE 1
+#define DEFAULT_STYLE 1
+
+settings_page global_settings_page;
+
+void settings_page_create();
+void settings_page_hide_without_save();
+void settings_page_update_render();
+void settings_page_show();
+void settings_page_hide();
+void settings_page_destroy();
+
+#endif
\ No newline at end of file
diff --git a/src/settings_config.c b/src/settings_config.c
new file mode 100644
index 0000000..d2bc018
--- /dev/null
+++ b/src/settings_config.c
@@ -0,0 +1,240 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+void settings_config_write_to_file(settings_config *config, char *path)
+{
+	// @hardcoded
+	s32 len = kilobytes(20);
+	char *buffer = mem_alloc(len);
+	buffer[0] = 0;
+	
+	for (s32 i = 0; i < config->settings.length; i++)
+	{
+		config_setting *setting = array_at(&config->settings, i);
+		
+		char entry_buf[MAX_INPUT_LENGTH];
+		snprintf(entry_buf, MAX_INPUT_LENGTH, "%s = \"%s\"\n", setting->name, setting->value);
+		string_appendn(buffer, entry_buf, MAX_INPUT_LENGTH);
+	}
+	
+	set_active_directory(binary_path);
+	platform_write_file_content(path, "w+", buffer, strlen(buffer));
+	mem_free(buffer);
+}
+
+static void get_config_from_string(settings_config *config, char *string)
+{
+	config_setting current_entry;
+	current_entry.name = 0;
+	current_entry.value = 0;
+	
+	s32 len = 0;
+	bool in_literal = false;
+	char *original_string = string;
+	
+	while(*string)
+	{
+		
+		// property name
+		if (*string == ' ' && !current_entry.name)
+		{
+			current_entry.name = mem_alloc(len+1);
+			string_copyn(current_entry.name, string-len, len);
+			current_entry.name[len] = 0;
+			string_trim(current_entry.name);
+		}
+		
+		// property value
+		if (*string == '"' && (*(string+1) == 0 || !in_literal))
+		{
+			in_literal = !in_literal;
+			
+			if (in_literal)
+			{
+				len = -1;
+			}
+			else
+			{
+				current_entry.value = mem_alloc(len+1);
+				string_copyn(current_entry.value, string-len, len);
+				current_entry.value[len] = 0;
+				string_trim(current_entry.value);
+			}
+		}
+		
+		++len;
+		++string;
+	}
+	
+	array_push(&config->settings, &current_entry);
+}
+
+static void convert_crlf_to_lf(char *buffer)
+{
+	char *buffer_original = buffer;
+	
+	int write_offset = 0;
+	int read_offset = 0;
+	
+	while(buffer[read_offset])
+	{
+		if (buffer[read_offset] != 0x0D)
+		{
+			buffer_original[write_offset] = buffer[read_offset];
+			
+			++write_offset;
+		}
+		
+		++read_offset;
+	}
+}
+
+settings_config settings_config_load_from_file(char *path)
+{
+	settings_config config;
+	config.settings = array_create(sizeof(config_setting));
+	
+	set_active_directory(binary_path);
+	
+	file_content content = platform_read_file_content(path, "rb");
+	
+	if (!content.content || content.file_error)
+	{
+		platform_destroy_file_content(&content);
+		return config;
+	}
+	
+	convert_crlf_to_lf(content.content);
+	
+	s32 token_offset = 0;
+	for (s32 i = 0; i < content.content_length; i++)
+	{
+		char ch = ((char*)content.content)[i];
+		char prev_ch = i-1 > 0 ? ((char*)content.content)[i-1] : 255;
+		
+		// end of line [lf]
+		if (ch == 0x0A && prev_ch != 0x0D)
+		{
+			char line[MAX_INPUT_LENGTH];
+			
+			s32 line_len = i - token_offset;
+			snprintf(line, MAX_INPUT_LENGTH, "%.*s", line_len, (char*)content.content+token_offset);
+			token_offset = i + 1;
+			
+			get_config_from_string(&config, line);
+		}
+	}
+	
+	platform_destroy_file_content(&content);
+	
+	return config;
+}
+
+config_setting* settings_config_get_setting(settings_config *config, char *name)
+{
+	for (s32 i = 0; i < config->settings.length; i++)
+	{
+		config_setting *setting = array_at(&config->settings, i);
+		if (setting && setting->name && name && strcmp(setting->name, name) == 0)
+		{
+			return setting;
+		}
+	}
+	return 0;
+}
+
+char* settings_config_get_string(settings_config *config, char *name)
+{
+	config_setting* setting = settings_config_get_setting(config, name);
+	if (setting)
+		return setting->value;
+	else
+		return 0;
+}
+
+s64 settings_config_get_number(settings_config *config, char *name)
+{
+	config_setting* setting = settings_config_get_setting(config, name);
+	if (setting && setting->value)
+		return string_to_u64(setting->value);
+	else
+		return 0;
+}
+
+void settings_config_set_string(settings_config *config, char *name, char *value)
+{
+	config_setting* setting = settings_config_get_setting(config, name);
+	if (setting)
+	{
+		s32 len = strlen(value);
+		mem_free(setting->value);
+		setting->value = mem_alloc(len+1);
+		string_copyn(setting->value, value, len+1);
+	}
+	else
+	{
+		config_setting new_entry;
+		new_entry.name = 0;
+		
+		// name
+		s32 len = strlen(name);
+		new_entry.name = mem_alloc(len+1);
+		string_copyn(new_entry.name, name, len+1);
+		
+		// value
+		len = strlen(value);
+		new_entry.value = mem_alloc(len+1);
+		string_copyn(new_entry.value, value, len+1);
+		
+		array_push(&config->settings, &new_entry);
+	}
+}
+
+void settings_config_set_number(settings_config *config, char *name, s64 value)
+{
+	config_setting* setting = settings_config_get_setting(config, name);
+	if (setting)
+	{
+		char num_buf[20];
+		snprintf(num_buf, 20, "%"PRId64"", value);
+		
+		s32 len = strlen(num_buf);
+		mem_free(setting->value);
+		setting->value = mem_alloc(len+1);
+		string_copyn(setting->value, num_buf, len+1);
+	}
+	else
+	{
+		config_setting new_entry;
+		
+		// name
+		s32 len = strlen(name);
+		new_entry.name = mem_alloc(len+1);
+		string_copyn(new_entry.name, name, len+1);
+		
+		// value
+		char num_buf[20];
+		snprintf(num_buf, 20, "%"PRId64"", value);
+		
+		len = strlen(num_buf);
+		new_entry.value = mem_alloc(len+1);
+		string_copyn(new_entry.value, num_buf, len+1);
+		array_push(&config->settings, &new_entry);
+	}
+}
+
+void settings_config_destroy(settings_config *config)
+{
+	for (s32 i = 0; i < config->settings.length; i++)
+	{
+		config_setting *entry = array_at(&config->settings, i);
+		
+		mem_free(entry->name);
+		mem_free(entry->value);
+	}
+	
+	array_destroy(&config->settings);
+}
\ No newline at end of file
diff --git a/src/settings_config.h b/src/settings_config.h
new file mode 100644
index 0000000..1b931e1
--- /dev/null
+++ b/src/settings_config.h
@@ -0,0 +1,40 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_SETTINGS_CONFIG
+#define INCLUDE_SETTINGS_CONFIG
+
+typedef struct t_config_setting
+{
+	char *name;
+	char *value;
+} config_setting;
+
+typedef struct t_settings_config
+{
+	array settings;
+} settings_config;
+
+/* Example of file:
+*  NAME = "Aldrik Ramaekers"
+*  AGE = "69"
+*  NUMBER = "15"
+*/
+
+settings_config settings_config_load_from_file(char *path);
+void settings_config_write_to_file(settings_config *config, char *path);
+void settings_config_destroy(settings_config *config);
+
+config_setting* settings_config_get_setting(settings_config *config, char *name);
+char* settings_config_get_string(settings_config *config, char *name);
+s64 settings_config_get_number(settings_config *config, char *name);
+
+void settings_config_set_string(settings_config *config, char *name, char *value);
+void settings_config_set_number(settings_config *config, char *name, s64 value);
+
+
+
+#endif
\ No newline at end of file
diff --git a/src/string_utils.c b/src/string_utils.c
new file mode 100644
index 0000000..5c8aac9
--- /dev/null
+++ b/src/string_utils.c
@@ -0,0 +1,551 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+bool string_match(char *first, char *second)
+{
+	// If we reach at the end of both strings, we are done 
+    if (*first == '\0' && *second == '\0') 
+        return true; 
+	
+    // Make sure that the characters after '*' are present 
+    // in second string. This function assumes that the first 
+    // string will not contain two consecutive '*' 
+    if (*first == '*' && *(first+1) != '\0' && *second == '\0') 
+        return false;
+	
+    // If the first string contains '?', or current characters 
+    // of both strings string_match
+    if (*first == '?' || *first == *second) 
+        return string_match(first+1, second+1);
+	
+    // If there is *, then there are two possibilities 
+    // a) We consider current character of second string 
+    // b) We ignore current character of second string. 
+    if (*first == '*') 
+        return string_match(first+1, second) || string_match(first, second+1); 
+    return false; 
+}
+
+bool string_is_asteriks(char *text)
+{
+	utf8_int32_t ch;
+	while((text = utf8codepoint(text, &ch)) && ch)
+	{
+		if (ch != '*') return false;
+	}
+	return true;
+}
+
+bool string_contains_ex(char *text_to_search, char *text_to_find, array *text_matches, bool *cancel_search)
+{
+	bool final_result = false;
+	bool is_asteriks_only = false;
+	
+	// * wildcard at the start of text to find is not needed
+	if (string_is_asteriks(text_to_find))
+	{
+		is_asteriks_only = true;
+		text_to_find += strlen(text_to_find);
+	}
+	
+	// remove all asteriks from start
+	utf8_int32_t br;
+	while(utf8codepoint(text_to_find, &br) && br == '*')
+	{
+		text_to_find = utf8codepoint(text_to_find, &br);
+	}
+	
+	char *text_to_find_original = text_to_find;
+	bool save_info = (text_matches != 0);
+	
+	utf8_int32_t text_to_search_ch = 0;
+	utf8_int32_t text_to_find_ch = 0;
+	size_t text_to_find_char_len = utf8len(text_to_find);
+	
+	s32 line_nr_val = 1;
+	s32 word_offset_val = 0;
+	s32 word_match_len_val = 0;
+	char* line_start_ptr = text_to_search;
+	
+	s32 index = 0;
+	while((text_to_search = utf8codepoint(text_to_search, &text_to_search_ch)) 
+		  && text_to_search_ch)
+	{
+		if (cancel_search && *cancel_search) goto set_info_and_return_failure;
+		
+		word_offset_val++;
+		if (text_to_search_ch == '\n') 
+		{
+			line_nr_val++;
+			word_offset_val = 0;
+			line_start_ptr = text_to_search;
+		}
+		
+		char *text_to_search_current_attempt = text_to_search;
+		utf8_int32_t text_to_search_current_attempt_ch = 0;
+		
+		bool in_wildcard = false;
+		
+		text_to_find = utf8codepoint(text_to_find, &text_to_find_ch);
+		text_to_search_current_attempt = utf8codepoint(text_to_search_current_attempt,
+													   &text_to_search_current_attempt_ch);
+		
+		word_match_len_val = 0;
+		while(text_to_search_current_attempt_ch)
+		{
+			if (cancel_search && *cancel_search) goto set_info_and_return_failure;
+			
+			// wildcard, accept any character in text to search
+			if (text_to_find_ch == '?')
+				goto continue_search;
+			
+			// character matches, 
+			if (text_to_find_ch == text_to_search_current_attempt_ch && in_wildcard)
+				in_wildcard = false;
+			
+			// wildcard, accept any characters in text to search untill next char is found
+			if (text_to_find_ch == '*')
+			{
+				text_to_find = utf8codepoint(text_to_find, &text_to_find_ch);
+				in_wildcard = true;
+			}
+			
+			// text to find has reached 0byte, word has been found
+			if (text_to_find_ch == 0)
+			{
+				if (save_info)
+				{
+					text_match new_match;
+					new_match.line_nr = line_nr_val;
+					new_match.word_offset = word_offset_val;
+					new_match.word_match_len = word_match_len_val;
+					new_match.line_start = line_start_ptr;
+					new_match.line_info = 0;
+					array_push(text_matches, &new_match);
+				}
+				
+				final_result = true;
+				
+				if (is_asteriks_only)
+				{
+					return final_result;
+				}
+				
+				break;
+			}
+			
+			// character does not match, continue search
+			if (text_to_find_ch != text_to_search_current_attempt_ch && !in_wildcard)
+				break;
+			
+			continue_search:
+			if (!in_wildcard)
+				text_to_find = utf8codepoint(text_to_find, &text_to_find_ch);
+			
+			text_to_search_current_attempt = utf8codepoint(
+				text_to_search_current_attempt,
+				&text_to_search_current_attempt_ch);
+			
+			word_match_len_val++;
+		}
+		
+		text_to_find = text_to_find_original;
+		index++;
+	}
+	
+	return final_result;
+	
+	set_info_and_return_failure:
+	return false;
+}
+
+static char *ltrim(char *str, const char *seps)
+{
+	size_t totrim;
+	if (seps == NULL) {
+		seps = "\t\n\v\f\r ";
+	}
+	totrim = strspn(str, seps);
+	if (totrim > 0) {
+		size_t len = strlen(str);
+		if (totrim == len) {
+			str[0] = '\0';
+		}
+		else {
+			memmove(str, str + totrim, len + 1 - totrim);
+		}
+	}
+	return str;
+}
+
+static char *rtrim(char *str, const char *seps)
+{
+	int i;
+	if (seps == NULL) {
+		seps = "\t\n\v\f\r ";
+	}
+	i = strlen(str) - 1;
+	while (i >= 0 && strchr(seps, str[i]) != NULL) {
+		str[i] = '\0';
+		i--;
+	}
+	return str;
+}
+
+inline void string_trim(char *string)
+{
+	ltrim(rtrim(string, 0), 0);
+}
+
+inline bool string_equals(char *first, char *second)
+{
+	return (strcmp(first, second) == 0);
+}
+
+// replaces " with \" for file formats
+void string_appendf(char *buffer, char *text)
+{
+	u32 len = strlen(buffer);
+	while(*text)
+	{
+		if (*text < 32)
+		{
+			buffer[len] = ' ';
+			len++;
+			text++;
+			continue;
+		}
+		
+		if (*text == '"')
+		{
+			buffer[len] = '\\';
+			len++;
+		}
+		if (*text == '\\')
+		{
+			buffer[len] = '\\';
+			len++;
+		}
+		
+		buffer[len] = *text;
+		len++;
+		text++;
+	}
+}
+
+void string_copyn(char *buffer, char *text, s32 bufferlen)
+{
+	u32 len = 0;
+	while(*text && len < bufferlen)
+	{
+		buffer[len] = *text;
+		len++;
+		text++;
+	}
+	buffer[len] = 0;
+}
+
+void string_appendn(char *buffer, char *text, s32 bufferlen)
+{
+	u32 len = strlen(buffer);
+	while(*text && len < bufferlen)
+	{
+		buffer[len] = *text;
+		len++;
+		text++;
+	}
+	buffer[len] = 0;
+}
+
+void string_append(char *buffer, char *text)
+{
+	u32 len = strlen(buffer);
+	while(*text)
+	{
+		buffer[len] = *text;
+		len++;
+		text++;
+	}
+	buffer[len] = 0;
+}
+
+bool string_remove(char **buffer, char *text)
+{
+	s32 len = strlen(text);
+	char tmp[200];
+	memcpy(tmp, *buffer, len);
+	memset(tmp+len, 0, 1);
+	
+	if (string_equals(tmp, text))
+	{
+		*buffer += len;
+		return true;
+	}
+	
+	return false;
+}
+
+char* string_get_json_literal(char **buffer, char *tmp)
+{
+	char *buf_start = *buffer;
+	char *buf = *buffer;
+	s32 len = 0;
+	while(*buf)
+	{
+		if ((*buf == ',' || *buf == '}') && (len > 0 && *(buf-1) == '"') && (len > 1 && *(buf-2) != '\\'))
+		{
+			memcpy(tmp, buf_start, len);
+			memset(tmp+len-1, 0, 1);
+			*buffer += len-1;
+			return tmp;
+		}
+		
+		len++;
+		buf++;
+	}
+	
+	return tmp;
+}
+
+s32 string_get_json_ulong_number(char **buffer)
+{
+	char tmp[20];
+	char *buf_start = *buffer;
+	char *buf = *buffer;
+	s32 len = 0;
+	while(*buf)
+	{
+		if (*buf == ',' || *buf == '}')
+		{
+			memcpy(tmp, buf_start, len);
+			memset(tmp+len, 0, 1);
+			*buffer += len;
+			return string_to_u64(tmp);
+		}
+		
+		len++;
+		buf++;
+	}
+	
+	return 0;
+}
+
+s32 string_get_json_number(char **buffer)
+{
+	char tmp[20];
+	char *buf_start = *buffer;
+	char *buf = *buffer;
+	s32 len = 0;
+	while(*buf)
+	{
+		if (*buf == ',' || *buf == '}')
+		{
+			memcpy(tmp, buf_start, len);
+			memset(tmp+len, 0, 1);
+			*buffer += len;
+			return string_to_s32(tmp);
+		}
+		
+		len++;
+		buf++;
+	}
+	
+	return 0;
+}
+
+void utf8_str_remove_range(char *str, s32 from, s32 to)
+{
+	char *orig_str = str;
+	s32 i = 0;
+	utf8_int32_t ch = 0;
+	s32 total_len = strlen(str)+1+4;
+	char *replacement = calloc(total_len,1);
+	char *rep_off = replacement;
+	replacement[0] = 0;
+	
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (i < from || i >= to)
+		{
+			rep_off = utf8catcodepoint(rep_off, ch, 5);
+		}
+		
+		++i;
+	}
+	*rep_off = 0;
+	
+	string_copyn(orig_str, replacement, MAX_INPUT_LENGTH);
+	mem_free(replacement);
+}
+
+void utf8_str_remove_at(char *str, s32 at)
+{
+	char *orig_str = str;
+	s32 i = 0;
+	utf8_int32_t ch = 0;
+	s32 total_len = strlen(str)+1+4;
+	char *replacement = calloc(total_len,1);
+	char *rep_off = replacement;
+	replacement[0] = 0;
+	
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (at != i)
+		{
+			rep_off = utf8catcodepoint(rep_off, ch, 5);
+		}
+		
+		++i;
+	}
+	*rep_off = 0;
+	
+	string_copyn(orig_str, replacement, MAX_INPUT_LENGTH);
+	mem_free(replacement);
+}
+
+void utf8_str_insert_utf8str(char *str, s32 at, char *toinsert)
+{
+	s32 index = 0;
+	utf8_int32_t ch;
+	while((toinsert = utf8codepoint(toinsert, &ch)) && ch)
+	{
+		utf8_str_insert_at(str, at+index, ch);
+		index++;
+	}
+}
+
+void utf8_str_insert_at(char *str, s32 at, utf8_int32_t newval)
+{
+	char *orig_str = str;
+	s32 i = 0;
+	utf8_int32_t ch = 0;
+	s32 total_len = strlen(str)+1+4;
+	char *replacement = calloc(total_len,1);
+	char *rep_off = replacement;
+	replacement[0] = 0;
+	
+	while((str = utf8codepoint(str, &ch)))
+	{
+		if (at == i)
+		{
+			rep_off = utf8catcodepoint(rep_off, newval, 5);
+		}
+		
+		rep_off = utf8catcodepoint(rep_off, ch, 5);
+		
+		++i;
+		
+		if (!ch) break;
+	}
+	*rep_off = 0;
+	
+	string_copyn(orig_str, replacement, MAX_INPUT_LENGTH);
+	mem_free(replacement);
+}
+
+char *utf8_str_copy_upto(char *str, s32 roof, char *buffer)
+{
+	utf8_int32_t ch = 0;
+	s32 index = 0;
+	char *orig_buffer = buffer;
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (index == roof) break;
+		buffer = utf8catcodepoint(buffer, ch, 5);
+		index++;
+	}
+	buffer = utf8catcodepoint(buffer, 0, 5);
+	
+	return orig_buffer;
+}
+
+char *utf8_str_copy_range(char *str, s32 floor, s32 roof, char *buffer)
+{
+	utf8_int32_t ch = 0;
+	s32 index = 0;
+	char *orig_buffer = buffer;
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (index == roof) break;
+		if (index >= floor)
+			buffer = utf8catcodepoint(buffer, ch, 5);
+		index++;
+	}
+	buffer = utf8catcodepoint(buffer, 0, 5);
+	
+	return orig_buffer;
+}
+
+void utf8_str_replace_at(char *str, s32 at, utf8_int32_t newval)
+{
+	char *orig_str = str;
+	s32 i = 0;
+	utf8_int32_t ch = 0;
+	s32 total_len = strlen(str)+1+4;
+	char *replacement = calloc(total_len,1);
+	char *rep_off = replacement;
+	replacement[0] = 0;
+	
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (at == i)
+		{
+			rep_off = utf8catcodepoint(rep_off, newval, 5);
+		}
+		else
+		{
+			rep_off = utf8catcodepoint(rep_off, ch, 5);
+		}
+		++i;
+	}
+	*rep_off = 0;
+	
+	string_copyn(orig_str, replacement, MAX_INPUT_LENGTH);
+	mem_free(replacement);
+}
+
+char* utf8_str_upto(char *str, s32 index)
+{
+	s32 i = 0;
+	utf8_int32_t ch;
+	char *prev_str = str;
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (index == i) return prev_str;
+		prev_str = str;
+		++i;
+	}
+	
+	return str;
+}
+
+utf8_int32_t utf8_str_at(char *str, s32 index)
+{
+	s32 i = 0;
+	utf8_int32_t ch;
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (index == i) return ch;
+		
+		++i;
+	}
+	
+	return 0;
+}
+
+bool is_string_numeric(char *str)
+{
+	utf8_int32_t ch;
+	while((str = utf8codepoint(str, &ch)) && ch)
+	{
+		if (!(ch >= 48 && ch <= 57))
+		{
+			return false;
+		}
+	}
+	
+	return true;
+}
\ No newline at end of file
diff --git a/src/string_utils.h b/src/string_utils.h
new file mode 100644
index 0000000..2b5c85f
--- /dev/null
+++ b/src/string_utils.h
@@ -0,0 +1,87 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_STRING_UTILS
+#define INCLUDE_STRING_UTILS
+
+#if 0
+printf("[lll][llll*] : %d\n", string_contains("lll", "llll*"));
+printf("[lllll][l*lop] : %d\n", string_contains("lllll", "l*lop"));
+printf("[22lllll][l*l] : %d\n", string_contains("22lllll", "l*l"));
+
+printf("[hello world][h?lo] : %d\n", string_contains("hello world", "h?lo"));
+printf("[\"hello sailor\"][sailor] : %d\n", string_contains(" wsdf asd \"hello sailor\" asdf asdf ", "sailor"));
+printf("[\"hello sailor\"][*sailor] : %d\n", string_contains(" wsdf asd \"hello sailor\" asdf asdf ", "*sailor"));
+printf("\n");
+
+printf("[\"hello sailor\"][*sailor\"] : %d\n", string_contains(" wsdf asd \"hello sailor\" asdf asdf ", "*sailor\""));
+printf("[\"hello sailor\"][*sailor*] : %d\n", string_contains(" wsdf asd \"hello sailor\" asdf asdf ", "*sailor*"));
+printf("[\"hello sailor\"][sailor*] : %d\n", string_contains(" wsdf asd \"hello sailor\" asdf asdf ", "sailor*"));
+printf("[22lllll pi23hjp rbksje LSKJDh l][LS*] : %d\n",
+	   string_contains("22lllll pi23hjp rbksje LSKJDh l", "LS*"));
+printf("[22lllll lal][l*l] : %d\n", string_contains("22lllll lal", "l*l"));
+printf("[22lllll][*l*l] : %d\n", string_contains("lllll", "*l*l"));
+printf("[hello world][hello] : %d\n", string_contains("hello world", "hello"));
+printf("[hello world][h?llo] : %d\n", string_contains("hello world", "h?llo"));
+printf("[hello world][h????] : %d\n", string_contains("hello world", "h????"));
+printf("[hello world][h*lo] : %d\n", string_contains("hello world", "h*lo"));
+printf("[hello world][*] : %d\n", string_contains("hello world", "*"));
+printf("[hello world][h*] : %d\n", string_contains("hello world", "h*"));
+printf("[hello world][*o] : %d\n", string_contains("hello world", "*o"));
+printf("[hello world][h*o] : %d\n", string_contains("hello world", "h*o"));
+printf("[hello world][*lo] : %d\n", string_contains("hello world", "*lo"));
+printf("[hello world][hel*lo] : %d\n", string_contains("hello world", "hel*lo"));
+
+printf("[lllll][l*l] : %d\n", string_contains("lllll", "l*l"));
+printf("[llllllll][l*llll] : %d\n", string_contains("lllll", "l*llll"));
+printf("[llllllll][l*lll] : %d\n", string_contains("lllll", "l*lll"));
+printf("[llllllll][llll*l] : %d\n", string_contains("lllll", "llll*l"));
+printf("[llllllll][*] : %d\n", string_contains("lllll", "*"));
+printf("[lllll][l?lll] : %d\n", string_contains("lllll", "l?lll"));
+
+printf("[lllll][lllll] : %d\n", string_contains("lllll", "lllll"));
+printf("[lllll][*llll] : %d\n", string_contains("lllll", "*llll"));
+printf("[lllll][llll*] : %d\n", string_contains("lllll", "llll*"));
+printf("[lllll][*llll*] : %d\n", string_contains("lllll", "*llll*"));
+printf("[lllll][*lllll*] : %d\n", string_contains("lllll", "*lllll*"));
+printf("[lllll][*ll*] : %d\n", string_contains("lllll", "*ll*"));
+#endif
+
+typedef struct t_text_match
+{
+	u32 line_nr;
+	s32 word_offset;
+	s32 word_match_len;
+	char *line_start;
+	char *line_info;
+} text_match;
+
+#define string_contains(big, small) string_contains_ex(big, small, 0, 0)
+bool string_match(char *first, char *second);
+bool string_contains_ex(char *big, char *small, array *text_matches, bool *cancel_search);
+void string_trim(char *string);
+bool string_equals(char *first, char *second);
+void string_append(char *buffer, char *text);
+void string_copyn(char *buffer, char *text, s32 bufferlen);
+void string_appendn(char *buffer, char *text, s32 bufferlen);
+void string_appendf(char *buffer, char *text);
+bool string_remove(char **buffer, char *text);
+char* string_get_json_literal(char **buffer, char *tmp);
+s32 string_get_json_number(char **buffer);
+s32 string_get_json_ulong_number(char **buffer);
+
+utf8_int32_t utf8_str_at(char *str, s32 index);
+void utf8_str_remove_at(char *str, s32 at);
+void utf8_str_remove_range(char *str, s32 from, s32 to);
+void utf8_str_insert_at(char *str, s32 at, utf8_int32_t newval);
+void utf8_str_insert_utf8str(char *str, s32 at, char *toinsert);
+void utf8_str_replace_at(char *str, s32 at, utf8_int32_t newval);
+char* utf8_str_upto(char *str, s32 index);
+char *utf8_str_copy_upto(char *str, s32 roof, char *buffer);
+char *utf8_str_copy_range(char *str, s32 floor, s32 roof, char *buffer);
+bool is_string_numeric(char *str);
+
+#endif
\ No newline at end of file
diff --git a/src/thread.h b/src/thread.h
new file mode 100644
index 0000000..b019fdb
--- /dev/null
+++ b/src/thread.h
@@ -0,0 +1,67 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_THREAD
+#define INCLUDE_THREAD
+
+#ifdef OS_LINUX
+#include <pthread.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/syscall.h>
+
+struct t_thread
+{
+	pthread_t thread;
+	bool valid;
+};
+
+struct t_mutex
+{
+	pthread_mutex_t mutex;
+};
+#endif
+
+#ifdef OS_WIN
+#include <windows.h>
+#include <process.h>    /* _beginthread, _endthread */
+#include <stddef.h>
+#include <stdlib.h>
+#include <conio.h>
+
+struct t_thread
+{
+	HANDLE thread;
+	bool valid;
+};
+
+struct t_mutex
+{
+	HANDLE mutex;
+};
+#endif
+
+typedef struct t_thread thread;
+typedef struct t_mutex mutex;
+
+thread thread_start(void *(*start_routine) (void *), void *arg);
+void thread_join(thread *thread);
+bool thread_tryjoin(thread *thread);
+void thread_detach(thread *thread);
+void thread_stop(thread *thread);
+u32 thread_get_id();
+void thread_sleep(u64 microseconds);
+
+mutex mutex_create_recursive();
+mutex mutex_create();
+
+void mutex_lock(mutex *mutex);
+bool mutex_trylock(mutex *mutex);
+void mutex_unlock(mutex *mutex);
+
+void mutex_destroy(mutex *mutex);
+
+#endif
\ No newline at end of file
diff --git a/src/ui.c b/src/ui.c
new file mode 100644
index 0000000..e18052f
--- /dev/null
+++ b/src/ui.c
@@ -0,0 +1,1481 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+inline void ui_begin(s32 id)
+{
+	global_ui_context.item_hovered = false;
+	global_ui_context.next_id = id * 100;
+	global_ui_context.layout.offset_x = 0;
+	global_ui_context.layout.offset_y = 0;
+	global_ui_context.layout.width = global_ui_context.layout.active_window->width;
+	global_ui_context.layout.height = global_ui_context.layout.active_window->height;
+	global_ui_context.layout.active_dropdown_state = 0;
+}
+
+inline void ui_end()
+{
+	
+}
+
+inline checkbox_state ui_create_checkbox(bool selected)
+{
+	checkbox_state state;
+	state.state = selected;
+	
+	return state;
+}
+
+inline textbox_state ui_create_textbox(u16 max_len)
+{
+	assert(max_len > 0);
+	assert(max_len <= MAX_INPUT_LENGTH);
+	
+	textbox_state state;
+	state.max_len = max_len;
+	state.buffer = mem_alloc(max_len+1);
+	state.buffer[0] = 0;
+	state.state = false;
+	state.text_offset_x = 0;
+	state.history = array_create(sizeof(textbox_history_entry));
+	state.future = array_create(sizeof(textbox_history_entry));
+	array_reserve(&state.history, 100);
+	state.history.reserve_jump = 100;
+	array_reserve(&state.future, 100);
+	state.future.reserve_jump = 100;
+	state.selection_start_index = 0;
+	state.double_clicked_to_select = false;
+	state.double_clicked_to_select_cursor_index = 0;
+	state.diff = 0;
+	state.last_click_cursor_index = -1;
+	state.attempting_to_select = false;
+	state.deselect_on_enter = true;
+	
+	return state;
+}
+
+void ui_destroy_textbox(textbox_state *state)
+{
+	for (s32 i = 0; i < state->history.length; i++)
+	{
+		char **history_entry = array_at(&state->history, i);
+		mem_free(*history_entry);
+	}
+	array_destroy(&state->history);
+	array_destroy(&state->future);
+	
+	mem_free(state->buffer);
+}
+
+inline button_state ui_create_button()
+{
+	button_state state;
+	state.state = 0;
+	
+	return state;
+}
+
+inline scroll_state ui_create_scroll(s32 scroll)
+{
+	scroll_state state;
+	state.scroll = 0;
+	state.height = scroll;
+	
+	return state;
+}
+
+inline dropdown_state ui_create_dropdown()
+{
+	dropdown_state state;
+	state.state = 0;
+	state.selected_index = 0;
+	return state;
+}
+
+void ui_set_style(u16 style)
+{
+	global_ui_context.style.id = style;
+	if (style == UI_STYLE_LIGHT)
+	{
+		global_ui_context.style.hypertext_foreground = rgb(66, 134, 244);
+		global_ui_context.style.hypertext_hover_foreground = rgb(221, 93, 202);
+		global_ui_context.style.image_outline_tint = rgb(200,200,200);
+		global_ui_context.style.scrollbar_handle_background = rgb(225,225,225);
+		global_ui_context.style.info_bar_background = rgb(225,225,225);
+		global_ui_context.style.error_foreground = rgb(224,79,95);
+		global_ui_context.style.item_hover_background = rgb(240,220,220);
+		global_ui_context.style.scrollbar_background = rgb(255,255,255);
+		global_ui_context.style.background = rgb(255,255,255);
+		global_ui_context.style.menu_hover_background = rgb(200,200,200);
+		global_ui_context.style.menu_background = rgb(225,225,225);
+		global_ui_context.style.widget_hover_background = rgb(200,200,200);
+		global_ui_context.style.widget_background = rgb(225,225,225);
+		global_ui_context.style.border = rgb(180,180,180);
+		global_ui_context.style.foreground = rgb(10, 10, 10);
+		global_ui_context.style.textbox_background = rgb(240,240,240);
+		global_ui_context.style.textbox_foreground = rgb(10,10,10);
+		global_ui_context.style.textbox_placeholder_foreground = rgb(80,80,80);
+		global_ui_context.style.textbox_active_border = rgb(66, 134, 244);
+	}
+	if (style == UI_STYLE_DARK)
+	{
+		global_ui_context.style.hypertext_foreground = rgb(66, 134, 244);
+		global_ui_context.style.hypertext_hover_foreground = rgb(221, 93, 202);
+		global_ui_context.style.scrollbar_handle_background = rgb(50,50,50);
+		global_ui_context.style.menu_hover_background = rgb(60,60,60);
+		global_ui_context.style.item_hover_background = rgb(80,60,60);
+		global_ui_context.style.image_outline_tint = rgb(200,200,200);
+		global_ui_context.style.error_foreground = rgb(224,79,95);
+		global_ui_context.style.scrollbar_background = rgb(80,80,80);
+		global_ui_context.style.widget_hover_background = rgb(50,50,50);
+		global_ui_context.style.widget_background = rgb(65,65,65);
+		global_ui_context.style.info_bar_background = rgb(65,65,65);
+		global_ui_context.style.menu_background = rgb(65,65,65);
+		global_ui_context.style.background = rgb(80, 80, 80);
+		global_ui_context.style.border = rgb(60,60,60);
+		global_ui_context.style.foreground = rgb(240,240,240);
+		global_ui_context.style.textbox_background = rgb(65,65,65);
+		global_ui_context.style.textbox_foreground = rgb(240, 240,240);
+		global_ui_context.style.textbox_active_border = rgb(66, 134, 244);
+	}
+}
+
+static scroll_state empty_scroll;
+inline void ui_create(platform_window *window, keyboard_input *keyboard, mouse_input *mouse, camera *camera, font *font_small)
+{
+	ui_set_style(UI_STYLE_LIGHT);
+	
+	global_ui_context.layout.layout_direction = LAYOUT_VERTICAL;
+	global_ui_context.layout.offset_x = 0;
+	global_ui_context.layout.offset_y = 0;
+	global_ui_context.layout.active_window = window;
+	global_ui_context.layout.width = global_ui_context.layout.active_window->width;
+	empty_scroll = ui_create_scroll(1);
+	global_ui_context.layout.scroll = &empty_scroll;
+	
+	global_ui_context.keyboard = keyboard;
+	global_ui_context.mouse = mouse;
+	global_ui_context.font_small = font_small;
+	global_ui_context.active_menus = array_create(sizeof(s32));
+	global_ui_context.menu_item_count = 0;
+	global_ui_context.camera = camera;
+	
+	array_reserve(&global_ui_context.active_menus, 100);
+}
+
+static void ui_pop_scissor()
+{
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		render_set_scissor(global_ui_context.layout.active_window,
+						   global_ui_context.layout.offset_x,
+						   global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING + 1, 
+						   global_ui_context.layout.width, 
+						   global_ui_context.layout.scroll->height + WIDGET_PADDING - 3);
+	}
+	else
+	{
+		render_reset_scissor();
+	}
+}
+
+inline void ui_block_begin(layout_direction direction)
+{
+	global_ui_context.layout.layout_direction = direction;
+	global_ui_context.layout.block_height = 0;
+	global_ui_context.layout.start_offset_y = global_ui_context.layout.offset_y;
+	global_ui_context.layout.start_offset_x = global_ui_context.layout.offset_x;
+	
+	ui_pop_scissor();
+}
+
+inline void ui_block_end()
+{
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+	{
+		global_ui_context.layout.offset_y += global_ui_context.layout.block_height + WIDGET_PADDING;
+	}
+	global_ui_context.layout.offset_x = global_ui_context.layout.start_offset_x;
+}
+
+inline void ui_set_active_window(platform_window *window)
+{
+	global_ui_context.layout.active_window = window;
+}
+
+inline void ui_begin_menu_bar()
+{
+	s32 w = global_ui_context.layout.width;
+	s32 h = global_ui_context.layout.active_window->height;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y;
+	
+	global_ui_context.layout.offset_x = 0;
+	global_ui_context.layout.layout_direction = LAYOUT_HORIZONTAL;
+	
+	render_rectangle(0, y, w, MENU_BAR_HEIGHT, global_ui_context.style.menu_background);
+	render_rectangle(0, y, w, 1, global_ui_context.style.border);
+	global_ui_context.layout.menu_offset_y = 0;
+}
+
+inline bool ui_is_menu_active(u32 id)
+{
+	for (int i = 0; i < global_ui_context.active_menus.length; i++)
+	{
+		s32 *iid = array_at(&global_ui_context.active_menus, i);
+		if (*iid == id) return true;
+	}
+	return false;
+}
+
+inline u32 ui_get_id()
+{
+	return global_ui_context.next_id++;
+}
+
+inline void ui_push_separator()
+{
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y;
+	s32 w = global_ui_context.layout.width;
+	
+	render_rectangle(x, y, w, 1, global_ui_context.style.border);
+	global_ui_context.layout.offset_y += 1 + WIDGET_PADDING;
+}
+
+void ui_push_vertical_dragbar()
+{
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x + global_ui_context.layout.width;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y - WIDGET_PADDING;
+	s32 h = global_ui_context.layout.height;
+	
+	render_rectangle(x, y, 5, h, global_ui_context.style.border);
+}
+
+inline void ui_push_menu_item_separator()
+{
+	global_ui_context.layout.menu_offset_y += 1;
+}
+
+static s32 ui_get_scroll()
+{
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		return global_ui_context.layout.scroll->scroll;
+	}
+	
+	return 0;
+}
+
+bool ui_push_color_button(char *text, bool selected, color c)
+{
+	bool result = false;
+	
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 text_x = x + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 text_y = y + (BUTTON_HEIGHT/2) - (global_ui_context.font_small->px_h/2);
+	s32 total_w =
+		BUTTON_HORIZONTAL_TEXT_PADDING + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	s32 h = BUTTON_HEIGHT;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	color bg_color = c;
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + h;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y >= virt_top && mouse_y < virt_bottom && !global_ui_context.item_hovered)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		bg_color.r-=20;
+		bg_color.g-=20;
+		bg_color.b-=20;
+		
+		if (is_left_clicked(global_ui_context.mouse)) 
+		{
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			result = true;
+		}
+	}
+	
+	if (selected)
+	{
+		render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+		render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 4, global_ui_context.style.border);
+	}
+	else
+	{
+		render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+		render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 1, global_ui_context.style.border);
+	}
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+	
+	return result;
+}
+
+bool ui_push_dropdown_item(image *icon, char *title, s32 index)
+{
+	bool result = false;
+	
+	set_render_depth(30);
+	
+	u32 id = ui_get_id();
+	global_ui_context.layout.dropdown_item_count++;
+	s32 h = BUTTON_HEIGHT;
+	s32 x = global_ui_context.layout.dropdown_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() + ((global_ui_context.layout.dropdown_item_count)*h-(1*global_ui_context.layout.dropdown_item_count));
+	s32 text_x = x + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 text_y = y + (BUTTON_HEIGHT/2) - (global_ui_context.font_small->px_h / 2);
+	s32 total_w = DROPDOWN_ITEM_WIDTH 
+		+ BUTTON_HORIZONTAL_TEXT_PADDING + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	color bg_color = global_ui_context.style.widget_background;
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y > y && mouse_y < y + h)
+	{
+		global_ui_context.item_hovered = true;
+		
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			global_ui_context.layout.active_dropdown_state->selected_index = index;
+			result = true;
+		}
+		
+		bg_color = global_ui_context.style.widget_hover_background;
+	}
+	
+	
+	render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+	render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 1, global_ui_context.style.border);
+	if (icon)
+	{
+		render_image(icon, x+(BUTTON_HORIZONTAL_TEXT_PADDING/2), 
+					 y + (h - (h-10))/2, h-10, h-10);
+		text_x += h-10;
+	}
+	render_text(global_ui_context.font_small, text_x+(BUTTON_HORIZONTAL_TEXT_PADDING/2)-5, text_y, title, global_ui_context.style.foreground);
+	
+	
+#if 0
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+#endif
+	set_render_depth(1);
+	
+	return result;
+}
+
+bool ui_push_dropdown(dropdown_state *state, char *title)
+{
+	bool result = false;
+	
+	global_ui_context.layout.active_dropdown_state = state;
+	
+	u32 id = ui_get_id();
+	global_ui_context.layout.dropdown_item_count = 0;
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 text_x = x + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 text_y = y + (BUTTON_HEIGHT/2) - (global_ui_context.font_small->px_h/2);
+	s32 total_w = DROPDOWN_WIDTH + BUTTON_HORIZONTAL_TEXT_PADDING + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	s32 h = BUTTON_HEIGHT;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	color bg_color = global_ui_context.style.widget_background;
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y >= y && mouse_y < y + h && !global_ui_context.item_hovered)
+	{
+		global_ui_context.item_hovered = true;
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			state->state = !state->state;
+		}
+		
+		bg_color = global_ui_context.style.widget_hover_background;
+	}
+	else if (is_left_down(global_ui_context.mouse) && state->state)
+	{
+		state->state = false;
+		// render dropdown this frame so item can be selected
+		result = true;
+	}
+	
+	render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+	render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 1, global_ui_context.style.border);
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	
+	render_triangle(x+total_w - h, y+(h-(h-12))/2, h-12, h-12, global_ui_context.style.border);
+	global_ui_context.layout.dropdown_x = global_ui_context.layout.offset_x;
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+	
+	return result || state->state;
+}
+
+bool ui_push_menu(char *title)
+{
+	bool result = false;
+	
+	global_ui_context.layout.menu_offset_y = 0;
+	global_ui_context.menu_item_count = 0;
+	u32 id = ui_get_id();
+	
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 w = calculate_text_width(global_ui_context.font_small, title) +
+		(MENU_HORIZONTAL_PADDING*2);
+	s32 text_h = global_ui_context.font_small->px_h;
+	s32 h = MENU_BAR_HEIGHT-1;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y+1;
+	s32 text_y = global_ui_context.layout.offset_y - (text_h / 2) + (h / 2) + global_ui_context.camera->y;
+	s32 text_x = x + MENU_HORIZONTAL_PADDING;
+	
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	color bg_color = global_ui_context.style.menu_background;
+	
+	bool is_open = ui_is_menu_active(id);
+	result = is_open;
+	
+	if (mouse_x >= x && mouse_x < x + w && mouse_y >= y && mouse_y < y + h)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			if (is_open)
+				array_remove_by(&global_ui_context.active_menus, &id);
+			else
+				array_push(&global_ui_context.active_menus, &id);
+			
+			result = !is_open;
+			is_open = result;
+		}
+		
+		bg_color = global_ui_context.style.menu_hover_background;
+	}
+	else if (is_left_down(global_ui_context.mouse))
+	{
+		if (is_open)
+			array_remove_by(&global_ui_context.active_menus, &id);
+		is_open = false;
+	}
+	if (!global_ui_context.layout.active_window->has_focus && is_open)
+	{
+		array_remove_by(&global_ui_context.active_menus, &id);
+		is_open = false;
+	}
+	
+	render_rectangle(x, y, w, h, bg_color);
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	
+	global_ui_context.layout.prev_offset_x = global_ui_context.layout.offset_x;
+	global_ui_context.layout.offset_x += w;
+	
+	return result;
+}
+
+static void ui_set_active_textbox(textbox_state *state)
+{
+	if (global_ui_context.current_active_textbox && global_ui_context.current_active_textbox != state)
+	{
+		global_ui_context.current_active_textbox->state = false;
+	}
+	global_ui_context.current_active_textbox = state;
+}
+
+void set_active_textbox(textbox_state *textbox)
+{
+	ui_set_active_textbox(textbox);
+	keyboard_set_input_text(global_ui_context.keyboard, textbox->buffer);
+	textbox->state = true;
+	global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+	global_ui_context.keyboard->take_input = textbox->state;
+}
+
+bool ui_push_textbox(textbox_state *state, char *placeholder)
+{
+	bool result = false;
+	static u64 cursor_tick = 0;
+	static u64 last_cursor_pos = -1;
+	
+	if (!global_ui_context.layout.active_window->has_focus)
+		state->state = false;
+	
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 text_x = x + 5;
+	s32 text_y = y + (TEXTBOX_HEIGHT/2) - (global_ui_context.font_small->px_h/2);
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	if (global_ui_context.layout.block_height < TEXTBOX_HEIGHT)
+		global_ui_context.layout.block_height = TEXTBOX_HEIGHT;
+	
+	bool has_text = state->buffer[0] != 0;
+	
+	if (!state->state)
+	{
+		render_rectangle(x, y, TEXTBOX_WIDTH, TEXTBOX_HEIGHT, global_ui_context.style.textbox_background);
+		render_rectangle_outline(x, y, TEXTBOX_WIDTH, TEXTBOX_HEIGHT, 1, global_ui_context.style.border);
+	}
+	else
+	{
+		cursor_tick++;
+		render_rectangle(x, y, TEXTBOX_WIDTH, TEXTBOX_HEIGHT, global_ui_context.style.textbox_background);
+		render_rectangle_outline(x, y, TEXTBOX_WIDTH, TEXTBOX_HEIGHT, 1, global_ui_context.style.textbox_active_border);
+	}
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + TEXTBOX_HEIGHT;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	/////////////////////////////////////////////
+	/////////////////////////////////////////////
+	/////////////////////////////////////////////
+	
+	bool is_selecting = false;
+	bool clicked_to_select = false;
+	bool double_clicked_to_select_first = false;
+	bool clicked_to_set_cursor = false;
+	if (mouse_x >= x && mouse_x < x + TEXTBOX_WIDTH && mouse_y >= virt_top && mouse_y < virt_bottom)
+	{
+		if (is_left_double_clicked(global_ui_context.mouse) && has_text)
+		{
+			ui_set_active_textbox(state);
+			
+			global_ui_context.keyboard->selection_begin_offset = 0;
+			global_ui_context.keyboard->selection_length = utf8len(global_ui_context.keyboard->input_text);
+			global_ui_context.keyboard->has_selection = true;
+			state->selection_start_index = 0;
+			
+			global_ui_context.mouse->left_state &= ~MOUSE_DOUBLE_CLICK;
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			
+			state->double_clicked_to_select = true;
+			double_clicked_to_select_first = true;
+		}
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			ui_set_active_textbox(state);
+			
+			keyboard_set_input_text(global_ui_context.keyboard, state->buffer);
+			cursor_tick = 0;
+			
+			if (global_ui_context.keyboard->has_selection)
+			{
+				global_ui_context.keyboard->has_selection = false;
+			}
+			
+			clicked_to_set_cursor = true;
+			
+			state->state = true;
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			result = true;
+			
+			global_ui_context.keyboard->take_input = state->state;
+		}
+	}
+	else if (is_left_clicked(global_ui_context.mouse))
+	{
+		if (state->state)
+		{
+			global_ui_context.keyboard->has_selection = false;
+		}
+		
+		state->state = false;
+	}
+	
+	if (is_left_released(global_ui_context.mouse))
+	{
+		state->attempting_to_select = false;
+	}
+	
+	if (state->state && global_ui_context.keyboard->has_selection && is_left_down(global_ui_context.mouse))
+		is_selecting = true;
+	
+	if (keyboard_is_key_pressed(global_ui_context.keyboard, KEY_ENTER) && state->deselect_on_enter)
+	{
+		global_ui_context.keyboard->has_selection = false;
+		state->state = false;
+	}
+	
+	// calculate scissor rectangle
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		vec4 v = render_get_scissor();
+		s32 scissor_x = v.x + WIDGET_PADDING + 5;
+		s32 scissor_y = global_ui_context.layout.scroll->scroll_start_offset_y;
+		s32 scissor_w = v.w;
+		s32 scissor_h = global_ui_context.layout.scroll->height - 2;
+		
+		render_set_scissor(global_ui_context.layout.active_window, scissor_x,
+						   scissor_y, scissor_w, scissor_h);
+	}
+	else
+	{
+		s32 scissor_x = x - global_ui_context.camera->x+3;
+		s32 scissor_y = y - global_ui_context.camera->y;
+		s32 scissor_w = TEXTBOX_WIDTH - 5;
+		s32 scissor_h = TEXTBOX_HEIGHT;
+		
+		render_set_scissor(global_ui_context.layout.active_window, 
+						   scissor_x, scissor_y, scissor_w, scissor_h);
+	}
+	
+	s32 cursor_text_w;
+	s32 cursor_x = 0;
+	
+	//if (!global_ui_context.keyboard->has_selection)
+	//state->diff = 0;
+	
+	// select first character on click
+	if (clicked_to_set_cursor)
+	{
+		global_ui_context.keyboard->cursor = calculate_cursor_position(global_ui_context.font_small, 
+																	   state->buffer, mouse_x + state->diff - text_x);
+		
+		state->last_click_cursor_index = global_ui_context.keyboard->cursor;
+		state->attempting_to_select = true;
+		
+		global_ui_context.keyboard->selection_begin_offset = global_ui_context.keyboard->cursor;
+		
+#if 0
+		global_ui_context.keyboard->has_selection = true;
+		global_ui_context.keyboard->selection_begin_offset = calculate_cursor_position(global_ui_context.font_small, 
+																					   state->buffer, mouse_x + state->diff - text_x);
+		global_ui_context.keyboard->selection_length = 1;
+		state->selection_start_index = global_ui_context.keyboard->selection_begin_offset;
+#endif
+	}
+	
+	if (state->state)
+	{
+		s32 len = utf8len(global_ui_context.keyboard->input_text);
+		s32 old_len = utf8len(state->buffer);
+		
+		// check if text changes, add to history if true
+		bool is_lctrl_down = global_ui_context.keyboard->keys[KEY_LEFT_CONTROL];
+		
+		// go to previous state
+		if (is_lctrl_down && keyboard_is_key_pressed(global_ui_context.keyboard, KEY_Z) && state->history.length)
+		{
+			textbox_history_entry history_entry;
+			history_entry.text = mem_alloc(strlen(state->buffer)+1);
+			history_entry.cursor_offset = last_cursor_pos;
+			string_copyn(history_entry.text, state->buffer, MAX_INPUT_LENGTH);
+			array_push(&state->future, &history_entry);
+			
+			global_ui_context.keyboard->text_changed = true;
+			
+			textbox_history_entry *old_text = array_at(&state->history, state->history.length-1);
+			string_copyn(state->buffer, old_text->text, MAX_INPUT_LENGTH);
+			keyboard_set_input_text(global_ui_context.keyboard, state->buffer);
+			
+			mem_free(old_text->text);
+			array_remove_at(&state->history, state->history.length-1);
+			
+			global_ui_context.keyboard->cursor = old_text->cursor_offset;
+		}
+		else if (is_lctrl_down && 
+				 keyboard_is_key_pressed(global_ui_context.keyboard, KEY_Y) && state->future.length)
+		{
+			textbox_history_entry history_entry;
+			history_entry.text = mem_alloc(strlen(state->buffer)+1);
+			history_entry.cursor_offset = last_cursor_pos;
+			string_copyn(history_entry.text, state->buffer, MAX_INPUT_LENGTH);
+			array_push(&state->history, &history_entry);
+			
+			global_ui_context.keyboard->text_changed = true;
+			
+			textbox_history_entry *old_text = array_at(&state->future, state->future.length-1);
+			string_copyn(state->buffer, old_text->text, MAX_INPUT_LENGTH);
+			keyboard_set_input_text(global_ui_context.keyboard, state->buffer);
+			
+			mem_free(old_text->text);
+			array_remove_at(&state->future, state->future.length-1);
+			
+			global_ui_context.keyboard->cursor = old_text->cursor_offset;
+		}
+		else
+		{
+			if (global_ui_context.keyboard->text_changed)
+			{
+				if (last_cursor_pos != -1)
+				{
+					textbox_history_entry history_entry;
+					history_entry.text = mem_alloc(strlen(state->buffer)+1);
+					history_entry.cursor_offset = last_cursor_pos;
+					string_copyn(history_entry.text, state->buffer, MAX_INPUT_LENGTH);
+					array_push(&state->history, &history_entry);
+				}
+			}
+			
+			string_copyn(state->buffer, global_ui_context.keyboard->input_text, state->max_len);
+			if (global_ui_context.keyboard->cursor > state->max_len)
+			{
+				global_ui_context.keyboard->cursor = state->max_len;
+				utf8_str_replace_at(global_ui_context.keyboard->input_text,global_ui_context.keyboard->cursor, 0);
+			}
+		}
+		
+		// cursor ticking after text change
+		if (last_cursor_pos != global_ui_context.keyboard->cursor || global_ui_context.keyboard->text_changed)
+			cursor_tick = 0;
+		
+		// draw cursor
+		cursor_text_w = calculate_text_width_upto(global_ui_context.font_small, 
+												  state->buffer, global_ui_context.keyboard->cursor);
+		
+		s32 text_w = calculate_text_width(global_ui_context.font_small, state->buffer);
+		
+		cursor_x = text_x + cursor_text_w - state->diff;
+		
+		// change offset after cursor position change
+		if (!is_selecting && !global_ui_context.keyboard->has_selection)
+		{
+			if (cursor_text_w < state->diff)
+			{
+				state->diff = cursor_text_w;
+			}
+			if (cursor_text_w - state->diff > TEXTBOX_WIDTH - 10)
+			{
+				state->diff = (cursor_text_w) - (TEXTBOX_WIDTH - 10);
+			}
+		}
+		
+		// make sure offset is recalculated when text changes or a portion of text is changed so the textbox doesnt end up half empty
+#if 1
+		if (!clicked_to_select && !clicked_to_set_cursor && !is_selecting && !global_ui_context.keyboard->has_selection && global_ui_context.keyboard->text_changed)
+		{
+			if ((text_w > TEXTBOX_WIDTH -10) && (global_ui_context.keyboard->text_changed || global_ui_context.keyboard->cursor != last_cursor_pos))
+			{
+				state->diff = text_w - TEXTBOX_WIDTH + 10;
+			}
+			else if ((text_w <= TEXTBOX_WIDTH -10))
+			{
+				state->diff = 0;
+			}
+		}
+#endif
+	}
+	
+	s32 curr_index = calculate_cursor_position(global_ui_context.font_small, 
+											   state->buffer, mouse_x + state->diff - text_x);
+	
+	//////////////////////////////////
+	{
+		if (curr_index != state->last_click_cursor_index && state->attempting_to_select)
+		{
+			clicked_to_select = true;
+			state->attempting_to_select = false;
+		}
+	}
+	
+	
+	// select first character on click
+	if (clicked_to_select)
+	{
+#if 1
+		global_ui_context.keyboard->has_selection = true;
+		//global_ui_context.keyboard->selection_begin_offset = calculate_cursor_position(global_ui_context.font_small, 
+		//state->buffer, mouse_x + state->diff - text_x);
+		global_ui_context.keyboard->selection_length = 0;
+		state->selection_start_index = global_ui_context.keyboard->selection_begin_offset;
+		
+		state->selection_start_index--;
+#endif
+	}
+	
+	
+	if (is_selecting)
+	{
+		// move text offset x when selecting so we can select more text than available on screen.
+		if (global_ui_context.mouse->x < x + 10)
+		{
+			s32 text_w = calculate_text_width(global_ui_context.font_small, state->buffer);
+			if (text_w > TEXTBOX_WIDTH-10)
+			{
+				state->diff -= TEXTBOX_SCROLL_X_SPEED;
+				if (state->diff < 0) state->diff = 0;
+			}
+		}
+		if (global_ui_context.mouse->x > x + TEXTBOX_WIDTH - 10)
+		{
+			s32 text_w = calculate_text_width(global_ui_context.font_small, state->buffer);
+			s32 diff = text_w - TEXTBOX_WIDTH + 10;
+			
+			if (text_w > TEXTBOX_WIDTH-10)
+			{
+				state->diff += TEXTBOX_SCROLL_X_SPEED;
+				if (state->diff > diff)
+					state->diff = diff;
+			}
+		}
+		///////////////////////////////////////////////////////////
+	}
+	
+	// change selection area based on cursor position.
+	// if double clicked to select the entire textbox we should only 
+	// do this when the mouse has moved enough to select a new character
+#if 1
+	if (is_selecting)
+	{
+		s32 index = calculate_cursor_position(global_ui_context.font_small, 
+											  state->buffer, mouse_x + state->diff - text_x)+1;
+		
+		if (double_clicked_to_select_first) 
+			state->double_clicked_to_select_cursor_index = index;
+		
+		if (!state->double_clicked_to_select || (state->double_clicked_to_select && index != state->double_clicked_to_select_cursor_index))
+		{
+			if (index <= state->selection_start_index+1)
+			{
+				global_ui_context.keyboard->selection_begin_offset = index - 1;
+				global_ui_context.keyboard->selection_length = state->selection_start_index - index + 2;
+			}
+			else if (index > state->selection_start_index)
+			{
+				global_ui_context.keyboard->selection_begin_offset = state->selection_start_index+1;
+				global_ui_context.keyboard->selection_length = index - state->selection_start_index-2;
+			}
+			
+			state->double_clicked_to_select = false;
+		}
+	}
+	
+	// render cursor
+	if (state->state)
+	{
+		last_cursor_pos = global_ui_context.keyboard->cursor;
+		
+		s32 cursor_y = y + 4;
+		s32 cursor_h = TEXTBOX_HEIGHT - 8;
+		s32 cursor_w = 2;
+		
+		if (cursor_tick % 40 < 20 && !global_ui_context.keyboard->has_selection)
+			render_rectangle(cursor_x, cursor_y, cursor_w, cursor_h, global_ui_context.style.textbox_foreground);
+	}
+	
+	
+	// render selection area
+	if (global_ui_context.keyboard->has_selection && state->state)
+	{
+		char tmp_buffer[MAX_INPUT_LENGTH];
+		utf8_str_copy_upto(state->buffer,
+						   global_ui_context.keyboard->selection_begin_offset,
+						   tmp_buffer);
+		
+		s32 selection_start_x = calculate_text_width(global_ui_context.font_small, 
+													 tmp_buffer);
+		
+		utf8_str_copy_upto(
+			utf8_str_upto(
+			state->buffer,
+			global_ui_context.keyboard->selection_begin_offset),
+			global_ui_context.keyboard->selection_length,
+			tmp_buffer);
+		
+		s32 selection_width = calculate_text_width(global_ui_context.font_small, 
+												   tmp_buffer);
+		
+		render_rectangle(text_x - state->diff + selection_start_x, y+4, selection_width, TEXTBOX_HEIGHT-8, global_ui_context.style.textbox_active_border);
+	}
+#endif
+	
+	if (!has_text)
+	{
+		render_text(global_ui_context.font_small, text_x - state->diff, text_y, 
+					placeholder, global_ui_context.style.textbox_placeholder_foreground);
+	}
+	else
+	{
+		render_text(global_ui_context.font_small, text_x - state->diff, text_y, 
+					state->buffer, global_ui_context.style.foreground);
+	}
+	
+	ui_pop_scissor();
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += TEXTBOX_WIDTH + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += TEXTBOX_HEIGHT + WIDGET_PADDING;
+	
+	return result || state->state;
+}
+
+bool ui_push_hypertext_link(char *text)
+{
+	bool result = false;
+	
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + WIDGET_PADDING;
+	s32 text_h = global_ui_context.font_small->px_h;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (global_ui_context.font_small->px_h/2) + spacing_y;
+	s32 total_w = calculate_text_width(global_ui_context.font_small, text) +
+		WIDGET_PADDING + WIDGET_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	if (global_ui_context.layout.block_height < global_ui_context.font_small->px_h)
+		global_ui_context.layout.block_height = global_ui_context.font_small->px_h;
+	
+	color bg_color = global_ui_context.style.hypertext_foreground;
+	if (mouse_x >= text_x && mouse_x < text_x + total_w && mouse_y >= text_y && mouse_y < text_y+text_h && !global_ui_context.item_hovered)
+	{
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			result = true;
+		}
+		bg_color = global_ui_context.style.hypertext_hover_foreground;
+	}
+	
+	s32 text_width = render_text(global_ui_context.font_small, text_x, text_y, text, bg_color);
+	
+	if (result)
+		render_rectangle(text_x, text_y + text_h-1, text_width, 1, bg_color);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+	
+	return result;
+}
+
+void ui_push_textf(font *f, char *text)
+{
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + WIDGET_PADDING;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (f->px_h/2) + spacing_y;
+	s32 total_w = calculate_text_width(f, text) +
+		WIDGET_PADDING + WIDGET_PADDING;
+	
+	if (global_ui_context.layout.block_height < f->px_h)
+		global_ui_context.layout.block_height = f->px_h+5;
+	
+	render_text(f, text_x, text_y, text, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+}
+
+
+void ui_push_textf_width(font *f, char *text, s32 maxw)
+{
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + WIDGET_PADDING;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (f->px_h/2) + spacing_y;
+	s32 total_w = maxw +
+		WIDGET_PADDING + WIDGET_PADDING;
+	
+	if (global_ui_context.layout.block_height < f->px_h)
+		global_ui_context.layout.block_height = f->px_h+5;
+	
+	render_text_ellipsed(f, text_x, text_y, maxw, text, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+}
+
+
+void ui_push_text(char *text)
+{
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + WIDGET_PADDING;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (global_ui_context.font_small->px_h/2) + spacing_y;
+	s32 total_w = calculate_text_width(global_ui_context.font_small, text) +
+		WIDGET_PADDING + WIDGET_PADDING;
+	
+	if (global_ui_context.layout.block_height < global_ui_context.font_small->px_h)
+		global_ui_context.layout.block_height = global_ui_context.font_small->px_h+5;
+	
+	render_text(global_ui_context.font_small, text_x, text_y, text, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+}
+
+void ui_push_text_width(char *text, s32 maxw)
+{
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + WIDGET_PADDING;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (global_ui_context.font_small->px_h/2) + spacing_y;
+	s32 total_w = maxw +
+		WIDGET_PADDING + WIDGET_PADDING;
+	
+	if (global_ui_context.layout.block_height < global_ui_context.font_small->px_h)
+		global_ui_context.layout.block_height = global_ui_context.font_small->px_h+5;
+	
+	render_text_ellipsed(global_ui_context.font_small, text_x, text_y, maxw, text, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+}
+
+bool ui_push_checkbox(checkbox_state *state, char *title)
+{
+	bool result = false;
+	
+	s32 spacing_y = (BLOCK_HEIGHT - CHECKBOX_SIZE)/2;
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll() - spacing_y;
+	s32 text_x = x + CHECKBOX_SIZE + WIDGET_PADDING;
+	s32 text_y = y + (BLOCK_HEIGHT/2) - (global_ui_context.font_small->px_h/2) + spacing_y;
+	s32 total_w = calculate_text_width(global_ui_context.font_small, title) +
+		CHECKBOX_SIZE + WIDGET_PADDING + WIDGET_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	if (global_ui_context.layout.block_height < CHECKBOX_SIZE)
+		global_ui_context.layout.block_height = CHECKBOX_SIZE;
+	
+	render_rectangle_outline(x, y, CHECKBOX_SIZE, CHECKBOX_SIZE, 1, global_ui_context.style.border);
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + CHECKBOX_SIZE;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	if (mouse_x >= x && mouse_x < x + CHECKBOX_SIZE && mouse_y >= virt_top && mouse_y < virt_bottom && !global_ui_context.item_hovered)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		if (is_left_clicked(global_ui_context.mouse))
+		{
+			state->state = !state->state;
+			result = true;
+		}
+	}
+	
+	if (state->state)
+	{
+		s32 spacing = 2;
+		render_rectangle(x+spacing, y+spacing, CHECKBOX_SIZE-(spacing*2), CHECKBOX_SIZE-(spacing*2), global_ui_context.style.border);
+	}
+	
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w;
+	else
+		global_ui_context.layout.offset_y += CHECKBOX_SIZE + WIDGET_PADDING;
+	
+	return result;
+}
+
+inline bool is_shortcut_down(s32 shortcut_keys[2])
+{
+	return keyboard_is_key_down(global_ui_context.keyboard, shortcut_keys[0]) &&
+		keyboard_is_key_pressed(global_ui_context.keyboard, shortcut_keys[1]);
+}
+
+bool ui_push_menu_item(char *title, char *shortcut)
+{
+	bool result = false;
+	
+	set_render_depth(30);
+	
+	global_ui_context.menu_item_count++;
+	
+	s32 x = global_ui_context.layout.prev_offset_x + global_ui_context.camera->x;
+	s32 w = MENU_ITEM_WIDTH;
+	s32 text_h = global_ui_context.font_small->px_h;
+	s32 h = MENU_BAR_HEIGHT;
+	s32 y = global_ui_context.layout.offset_y + (global_ui_context.menu_item_count * h)+1 +
+		global_ui_context.layout.menu_offset_y + global_ui_context.camera->y;
+	s32 text_y = y - (text_h / 2) + (h / 2);
+	s32 text_x = x + MENU_HORIZONTAL_PADDING;
+	s32 text_2_x = x + w - MENU_HORIZONTAL_PADDING
+		- calculate_text_width(global_ui_context.font_small, shortcut);
+	
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	color bg_color = global_ui_context.style.menu_background;
+	
+	if ((mouse_x >= x && mouse_x < x + w && mouse_y >= y && mouse_y < y + h))
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		bg_color = global_ui_context.style.menu_hover_background;
+		global_ui_context.item_hovered = true;
+		
+		if (is_left_clicked(global_ui_context.mouse)) 
+		{
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			result = true;
+		}
+	}
+	
+	render_rectangle(x, y, w, h, bg_color);
+	render_rectangle(x, y+MENU_BAR_HEIGHT, w, 1, global_ui_context.style.border);
+	
+	// borders
+	render_rectangle(x, y, w, 1,  bg_color);
+	render_rectangle(x, y, 1, MENU_BAR_HEIGHT, global_ui_context.style.border);
+	render_rectangle(x+w, y, 1, MENU_BAR_HEIGHT+1, global_ui_context.style.border);
+	
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	render_text(global_ui_context.font_small, text_2_x, text_y, shortcut, global_ui_context.style.foreground);
+	
+	set_render_depth(1);
+	
+	return result;
+}
+
+bool ui_push_image(image *img, s32 w, s32 h, s32 outline, color tint)
+{
+	bool result = false;
+	
+	if (!img->loaded) return result;
+	
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 total_w = w;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + h;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y >= virt_top && mouse_y < virt_bottom && !global_ui_context.item_hovered)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		
+		if (is_left_clicked(global_ui_context.mouse)) 
+		{
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+		}
+		if (is_left_released(global_ui_context.mouse))
+		{
+			result = true;
+		}
+	}
+	
+	render_image_tint(img,x,y,w,h,global_ui_context.style.image_outline_tint);
+	render_image_tint(img,x+outline,y+outline,w-(outline*2),h-(outline*2),tint);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+	
+	return result;
+}
+
+bool ui_push_button(button_state *state, char *title)
+{
+	bool result = false;
+	
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 text_x = x + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 text_y = y + (BUTTON_HEIGHT/2) - (global_ui_context.font_small->px_h/2);
+	s32 total_w = calculate_text_width(global_ui_context.font_small, title) +
+		BUTTON_HORIZONTAL_TEXT_PADDING + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	s32 h = BUTTON_HEIGHT;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	color bg_color = global_ui_context.style.widget_background;
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + h;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y >= virt_top && mouse_y < virt_bottom && !global_ui_context.item_hovered)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		bg_color = global_ui_context.style.widget_hover_background;
+		
+		if (is_left_clicked(global_ui_context.mouse)) 
+		{
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			state->state = 1;
+		}
+		if (is_left_released(global_ui_context.mouse) && state->state)
+		{
+			state->state = 0;
+			result = true;
+		}
+	}
+	if (is_left_released(global_ui_context.mouse))
+	{
+		//state->state = 0;
+	}
+	
+	render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+	render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 1, global_ui_context.style.border);
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+	
+	return result;
+}
+
+
+bool ui_push_button_image(button_state *state, char *title, image *img)
+{
+	bool result = false;
+	
+	s32 x = global_ui_context.layout.offset_x + WIDGET_PADDING + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y + ui_get_scroll();
+	s32 text_x = x + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 text_y = y + (BUTTON_HEIGHT/2) - (global_ui_context.font_small->px_h/2);
+	s32 text_w = calculate_text_width(global_ui_context.font_small, title);
+	s32 total_w = text_w + BUTTON_HORIZONTAL_TEXT_PADDING;
+	s32 mouse_x = global_ui_context.mouse->x + global_ui_context.camera->x;
+	s32 mouse_y = global_ui_context.mouse->y + global_ui_context.camera->y;
+	s32 h = BUTTON_HEIGHT;
+	
+	if (title[0] == 0)
+	{
+		total_w = 0;
+	}
+	
+	color bg_color = global_ui_context.style.widget_background;
+	
+	if (global_ui_context.layout.block_height < h)
+		global_ui_context.layout.block_height = h;
+	
+	int icon_w = 1;
+	int icon_h = 1;
+	if (img->loaded)
+	{
+		float max_icon_size = BUTTON_HEIGHT - (BUTTON_IMAGE_PADDING*2);
+		float scale = 1.0f;
+		if (img->width >= img->height)
+		{
+			scale = img->width / max_icon_size;
+			
+			icon_w = img->width / scale;
+			icon_h = icon_w;
+		}
+		else if (img->height >= img->width)
+		{
+			scale = img->height / max_icon_size;
+			
+			icon_h = img->height / scale;
+			icon_w = icon_h;
+		}
+		
+		total_w += icon_w + (BUTTON_IMAGE_SPACING*2);
+	}
+	
+	s32 virt_top = y;
+	s32 virt_bottom = y + h;
+	if (global_ui_context.layout.scroll->in_scroll)
+	{
+		s32 bottom = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+		if (bottom < virt_bottom)
+			virt_bottom = bottom;
+		s32 top = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		if (top > virt_top)
+			virt_top = top;
+	}
+	
+	if (mouse_x >= x && mouse_x < x + total_w && mouse_y >= virt_top && mouse_y < virt_bottom && !global_ui_context.item_hovered)
+	{
+		platform_set_cursor(global_ui_context.layout.active_window, CURSOR_POINTER);
+		bg_color = global_ui_context.style.widget_hover_background;
+		
+		if (is_left_clicked(global_ui_context.mouse)) 
+		{
+			global_ui_context.mouse->left_state &= ~MOUSE_CLICK;
+			state->state = 1;
+		}
+		if (is_left_released(global_ui_context.mouse) && state->state)
+		{
+			state->state = 0;
+			result = true;
+		}
+	}
+	if (is_left_released(global_ui_context.mouse))
+	{
+		state->state = 0;
+	}
+	
+	render_rectangle(x, y, total_w, BUTTON_HEIGHT, bg_color);
+	render_rectangle_outline(x, y, total_w, BUTTON_HEIGHT, 1, global_ui_context.style.border);
+	render_text(global_ui_context.font_small, text_x, text_y, title, global_ui_context.style.foreground);
+	render_image(img, x + total_w - icon_w - BUTTON_IMAGE_SPACING, y + BUTTON_IMAGE_PADDING, img->width, img->height);
+	
+	if (global_ui_context.layout.layout_direction == LAYOUT_HORIZONTAL)
+		global_ui_context.layout.offset_x += total_w + WIDGET_PADDING;
+	else
+		global_ui_context.layout.offset_y += BUTTON_HEIGHT + WIDGET_PADDING;
+	
+	return result;
+}
+
+inline void ui_end_menu_bar()
+{
+	global_ui_context.layout.layout_direction = LAYOUT_VERTICAL;
+	global_ui_context.layout.offset_x = 0;
+	global_ui_context.layout.offset_y += MENU_BAR_HEIGHT;
+	
+	ui_push_separator();
+}
+
+inline void ui_destroy()
+{
+	array_destroy(&global_ui_context.active_menus);
+}
+
+void ui_scroll_begin(scroll_state *state)
+{
+	global_ui_context.layout.scroll = state;
+	global_ui_context.layout.scroll->in_scroll = true;
+	//global_ui_context.layout.scroll->height = height;
+	
+	s32 w = global_ui_context.layout.width;
+	s32 h = state->height;
+	s32 x = global_ui_context.layout.offset_x + global_ui_context.camera->x;
+	s32 y = global_ui_context.layout.offset_y + global_ui_context.camera->y - WIDGET_PADDING;
+	
+	//global_ui_context.layout.offset_x += WIDGET_PADDING;
+	global_ui_context.layout.start_offset_x = global_ui_context.layout.offset_x;
+	//global_ui_context.layout.offset_y += WIDGET_PADDING;
+	global_ui_context.layout.scroll->scroll_start_offset_y = global_ui_context.layout.offset_y;
+	
+	//render_rectangle_outline(x, y, w, h, 1, global_ui_context.style.border);
+	render_set_scissor(global_ui_context.layout.active_window, x, y, w, h);
+}
+
+void ui_scroll_end()
+{
+	s32 max_scroll = (global_ui_context.layout.scroll->scroll_start_offset_y -
+					  global_ui_context.layout.offset_y) + global_ui_context.layout.scroll->height;
+	
+	//global_ui_context.layout.offset_x -= WIDGET_PADDING;
+	global_ui_context.layout.offset_y = global_ui_context.layout.scroll->scroll_start_offset_y + global_ui_context.layout.scroll->height;
+	global_ui_context.layout.offset_y += WIDGET_PADDING;
+	
+	// draw scrollbar
+	if (max_scroll < 0)
+	{
+		s32 scroll_y = 0;
+		if (global_ui_context.mouse->scroll_state == SCROLL_UP)
+			scroll_y+=SCROLL_SPEED;
+		if (global_ui_context.mouse->scroll_state == SCROLL_DOWN)
+			scroll_y-=SCROLL_SPEED;
+		global_ui_context.layout.scroll->scroll += scroll_y;
+		if (global_ui_context.layout.scroll->scroll > 0)
+			global_ui_context.layout.scroll->scroll = 0;
+		if (global_ui_context.layout.scroll->scroll < max_scroll)
+			global_ui_context.layout.scroll->scroll = max_scroll;
+		
+		float percentage = global_ui_context.layout.scroll->scroll / 
+			(float)max_scroll;
+		
+		float scrollbar_height_percentage = -(max_scroll - global_ui_context.layout.scroll->height) / (float)global_ui_context.layout.scroll->height;
+		
+		s32 scrollbar_height = global_ui_context.layout.scroll->height / scrollbar_height_percentage;
+		s32 scrollbar_pos_y = global_ui_context.layout.scroll->scroll_start_offset_y - WIDGET_PADDING;
+		
+		s32 tw = global_ui_context.layout.width - WIDGET_PADDING*2;
+		s32 tx = global_ui_context.layout.offset_x + global_ui_context.camera->x + WIDGET_PADDING;
+		
+		s32 scrollbar_pos_x = tx + tw + WIDGET_PADDING - 10;
+		
+		scrollbar_pos_y += (global_ui_context.layout.scroll->height - 
+							scrollbar_height + WIDGET_PADDING - 2) * percentage;
+		
+		render_reset_scissor();
+		render_rectangle(scrollbar_pos_x, scrollbar_pos_y, 10, scrollbar_height, global_ui_context.style.scrollbar_handle_background);
+	}
+	render_reset_scissor();
+	global_ui_context.layout.scroll->in_scroll = false;
+}
diff --git a/src/ui.h b/src/ui.h
new file mode 100644
index 0000000..b9a1f0d
--- /dev/null
+++ b/src/ui.h
@@ -0,0 +1,205 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#ifndef INCLUDE_UI
+#define INCLUDE_UI
+
+#define BLOCK_HEIGHT 25
+#define MENU_BAR_HEIGHT 25
+#define MENU_HORIZONTAL_PADDING 10
+#define WIDGET_PADDING 8
+#define BUTTON_HORIZONTAL_TEXT_PADDING 15
+#define MENU_ITEM_WIDTH 220
+#define CHECKBOX_SIZE BLOCK_HEIGHT - 8
+#define TEXTBOX_WIDTH 280
+#define TEXTBOX_HEIGHT BLOCK_HEIGHT
+#define BUTTON_HEIGHT BLOCK_HEIGHT
+#define BUTTON_IMAGE_PADDING 5
+#define BUTTON_IMAGE_SPACING 8
+#define DROPDOWN_WIDTH 225
+#define DROPDOWN_ITEM_WIDTH 225
+#define TEXTBOX_SCROLL_X_SPEED 32
+
+typedef enum t_ui_style_type
+{
+	UI_STYLE_LIGHT = 1,
+	UI_STYLE_DARK = 2,
+} ui_style_type;
+
+typedef struct t_ui_style
+{
+	u16 id;
+	color foreground;
+	color background;
+	color border;
+	color textbox_background;
+	color textbox_active_border;
+	color textbox_foreground;
+	color image_outline_tint;
+	color scrollbar_handle_background;
+	color info_bar_background;
+	color error_foreground;
+	color item_hover_background;
+	color scrollbar_background;
+	color menu_background;
+	color menu_hover_background;
+	color menu_foreground;
+	color widget_hover_background;
+	color widget_background;
+	color hypertext_foreground;
+	color hypertext_hover_foreground;
+	color textbox_placeholder_foreground;
+} ui_style;
+
+typedef enum t_layout_direction
+{
+	LAYOUT_HORIZONTAL,
+	LAYOUT_VERTICAL,
+} layout_direction;
+
+typedef struct t_dropdown_state
+{
+	bool state;
+	int selected_index;
+} dropdown_state;
+
+typedef struct t_scroll_state
+{
+	s32 height;
+	s32 scroll;
+	s32 scroll_start_offset_y;
+	bool in_scroll;
+} scroll_state;
+
+typedef struct t_ui_layout
+{
+	s32 dropdown_item_count;
+	s32 dropdown_x;
+	s32 offset_x;
+	s32 offset_y;
+	platform_window *active_window;
+	layout_direction layout_direction;
+	s32 prev_offset_x;
+	s32 width;
+	s32 height;
+	s32 menu_offset_y;
+	s32 block_height;
+	s32 start_offset_y;
+	s32 start_offset_x;
+	scroll_state *scroll;
+	s32 padding;
+	dropdown_state *active_dropdown_state;
+} ui_layout;
+
+typedef struct t_textbox_history_entry
+{
+	char *text;
+	s32 cursor_offset;
+} textbox_history_entry;
+
+typedef struct t_textbox_state
+{
+	bool deselect_on_enter;
+	char *buffer;
+	s32 selection_start_index;
+	bool state;
+	s32 diff;
+	bool double_clicked_to_select;
+	s32 double_clicked_to_select_cursor_index;
+	s32 max_len;
+	s32 text_offset_x;
+	bool attempting_to_select;
+	array history;
+	array future;
+	s32 last_click_cursor_index;
+} textbox_state;
+
+typedef struct t_checkbox_state
+{
+	bool state;
+} checkbox_state;
+
+typedef struct t_button_state
+{
+	bool state;
+} button_state;
+
+typedef struct t_ui_context
+{
+	ui_style style;
+	ui_layout layout;
+	keyboard_input *keyboard;
+	mouse_input *mouse;
+	camera *camera;
+	font *font_small;
+	array active_menus;
+	u32 next_id;
+	s32 menu_item_count;
+	dropdown_state *active_dropdown;
+	textbox_state *current_active_textbox;
+	bool item_hovered;
+} ui_context;
+
+///// our global ui states ////
+checkbox_state checkbox_recursive;
+textbox_state textbox_search_text;
+textbox_state textbox_path;
+textbox_state textbox_file_filter;
+button_state button_select_directory;
+button_state button_find_text;
+button_state button_cancel;
+///////////////////////////////
+
+ui_context global_ui_context;
+
+u32 ui_get_id();
+void ui_create(platform_window *window, keyboard_input *keyboard, mouse_input *mouse, camera *camera, font *font_small);
+void ui_set_active_window(platform_window *window);
+void ui_destroy();
+void ui_begin();
+void ui_end();
+bool ui_is_menu_active(u32 id);
+char* name_of_day(s32 day);
+char* name_of_month(s32 month);
+void ui_set_style(u16 style);
+void set_active_textbox(textbox_state *textbox);
+
+// widget initialization
+checkbox_state ui_create_checkbox(bool selected);
+textbox_state ui_create_textbox(u16 max_len);
+button_state ui_create_button();
+scroll_state ui_create_scroll(s32 scroll);
+dropdown_state ui_create_dropdown();
+
+void ui_destroy_textbox(textbox_state *state);
+
+// widgets
+bool is_shortcut_down(s32 shortcut_keys[2]);
+void ui_begin_menu_bar();
+bool ui_push_menu(char *title);
+bool ui_push_menu_item(char *title, char *shortcut);
+void ui_push_menu_item_separator();
+bool ui_push_dropdown(dropdown_state *state, char *title);
+bool ui_push_dropdown_item(image *icon, char *title, s32 index);
+void ui_push_separator();
+void ui_push_vertical_dragbar();
+void ui_block_begin(layout_direction direction);
+void ui_block_end();
+void ui_end_menu_bar();
+void ui_push_text(char *text);
+void ui_push_textf(font *f, char *text);
+void ui_push_textf_width(font *f, char *text, s32 maxw);
+bool ui_push_hypertext_link(char *text);
+bool ui_push_color_button(char *text, bool selected, color color);
+bool ui_push_image(image *img, s32 w, s32 h, s32 outline, color tint);
+bool ui_push_checkbox(checkbox_state *state, char *title);
+bool ui_push_textbox(textbox_state *state, char *title);
+bool ui_push_button(button_state *button, char *title);
+bool ui_push_button_image(button_state *button, char *title, image *img);
+void ui_scroll_begin(scroll_state *state);
+void ui_scroll_end();
+
+#endif
\ No newline at end of file
diff --git a/src/windows/platform.c b/src/windows/platform.c
new file mode 100644
index 0000000..2187806
--- /dev/null
+++ b/src/windows/platform.c
@@ -0,0 +1,1267 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#include <locale.h>
+#include <windows.h>
+#include <GL/gl.h>
+#include <stdbool.h>
+#include <sysinfoapi.h>
+#include <wingdi.h>
+#include <errno.h>
+#include <shlwapi.h>
+#include <objbase.h>
+#include <shellapi.h>
+#include <gdiplus.h>
+#include <shlobj.h>
+#include "../external/LooplessSizeMove.c"
+
+struct t_platform_window
+{
+	HWND window_handle;
+	HDC hdc;
+	HGLRC gl_context;
+	WNDCLASS window_class;
+	
+    s32 min_width;
+	s32 min_height;
+	s32 max_width;
+	s32 max_height;
+	
+	// shared window properties
+	s32 width;
+	s32 height;
+	bool is_open;
+	bool has_focus;
+	cursor_type curr_cursor_type;
+	cursor_type next_cursor_type;
+};
+
+extern BOOL GetPhysicallyInstalledSystemMemory(PULONGLONG TotalMemoryInKilobytes);
+
+LARGE_INTEGER perf_frequency;
+static HINSTANCE instance;
+platform_window *current_window_to_handle;
+keyboard_input *current_keyboard_to_handle;
+mouse_input *current_mouse_to_handle;
+
+int cmd_show;
+
+bool platform_get_clipboard(platform_window *window, char *buffer)
+{
+	if (!OpenClipboard(NULL))
+		return false;
+	
+	if (!IsClipboardFormatAvailable(CF_UNICODETEXT)) 
+	{
+		CloseClipboard();
+		return false;
+	}
+	
+	wchar_t* clip_str = GetClipboardData(CF_UNICODETEXT);
+	if (!clip_str)
+	{
+		CloseClipboard();
+		return false;
+	}
+	
+	WideCharToMultiByte(CP_UTF8, 0, clip_str, -1, buffer, MAX_INPUT_LENGTH ,0,0);
+	
+	CloseClipboard();
+	return true;
+}
+
+bool platform_set_clipboard(platform_window *window, char *buffer)
+{
+	HANDLE clipboard_data;
+	
+	int char_num = MultiByteToWideChar(CP_UTF8, 0, buffer, -1, 0, 0);
+	wchar_t *convstr = mem_alloc(char_num*2);
+	int result = MultiByteToWideChar(CP_UTF8, 0, buffer, -1, convstr, char_num);
+	
+	size_t len = result;
+	size_t size = (len+1) * sizeof(wchar_t);
+	LPSTR dst;
+	
+	if (!OpenClipboard(NULL))
+		return false;
+	
+	clipboard_data = GlobalAlloc(GMEM_MOVEABLE, size);
+	if (clipboard_data)
+	{
+		dst = GlobalLock(clipboard_data);
+		memmove(dst, convstr, size);
+		dst[len*2] = 0;
+		GlobalUnlock(clipboard_data);
+		
+		SetClipboardData(CF_UNICODETEXT, clipboard_data);
+	}
+	else
+	{
+		CloseClipboard();
+		return false;
+	}
+	
+	CloseClipboard();
+	return true;
+}
+
+inline void platform_show_alert(char *title, char *message)
+{
+	// not implemented
+}
+
+inline void platform_destroy()
+{
+	assets_destroy();
+	
+#if defined(MODE_DEVELOPER)
+	memory_print_leaks();
+#endif
+}
+
+bool is_platform_in_darkmode()
+{
+	char *key = "HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Themes\\Personalize\\";
+	
+	HKEY result;
+	LSTATUS o = RegOpenKeyExA(HKEY_CURRENT_USER, key, 0, KEY_READ, &result);
+	
+	if (o == 0)
+	{
+		BYTE data;
+		DWORD len = 4;
+		RegQueryValueExA(result, "AppsUseLightTheme", NULL, NULL, &data, &len);
+		
+		if (data == 1) return true;
+	}
+	
+	return false;
+}
+
+inline void platform_set_cursor(platform_window *window, cursor_type type)
+{
+	if (window->next_cursor_type != type)
+	{
+		window->next_cursor_type = type;
+	}
+}
+
+bool platform_directory_exists(char *path)
+{
+	return PathFileExistsA(path) == TRUE;
+}
+
+static void create_key_tables()
+{
+	keycode_map[0x30] = KEY_0;
+	keycode_map[0x31] = KEY_1;
+	keycode_map[0x32] = KEY_2;
+	keycode_map[0x33] = KEY_3;
+	keycode_map[0x34] = KEY_4;
+	keycode_map[0x35] = KEY_5;
+	keycode_map[0x36] = KEY_6;
+	keycode_map[0x37] = KEY_7;
+	keycode_map[0x38] = KEY_8;
+	keycode_map[0x39] = KEY_9;
+	keycode_map[0x41] = KEY_A;
+	keycode_map[0x42] = KEY_B;
+	keycode_map[0x43] = KEY_C;
+	keycode_map[0x44] = KEY_D;
+	keycode_map[0x45] = KEY_E;
+	keycode_map[0x46] = KEY_F;
+	keycode_map[0x47] = KEY_G;
+	keycode_map[0x48] = KEY_H;
+	keycode_map[0x49] = KEY_I;
+	keycode_map[0x4A] = KEY_J;
+	keycode_map[0x4B] = KEY_K;
+	keycode_map[0x4C] = KEY_L;
+	keycode_map[0x4D] = KEY_M;
+	keycode_map[0x4E] = KEY_N;
+	keycode_map[0x4F] = KEY_O;
+	keycode_map[0x50] = KEY_P;
+	keycode_map[0x51] = KEY_Q;
+	keycode_map[0x52] = KEY_R;
+	keycode_map[0x53] = KEY_S;
+	keycode_map[0x54] = KEY_T;
+	keycode_map[0x55] = KEY_U;
+	keycode_map[0x56] = KEY_V;
+	keycode_map[0x57] = KEY_W;
+	keycode_map[0x58] = KEY_X;
+	keycode_map[0x59] = KEY_Y;
+	keycode_map[0x5A] = KEY_Z;
+	
+	keycode_map[VK_OEM_7] = KEY_APOSTROPHE;
+	keycode_map[VK_OEM_102] = KEY_BACKSLASH;
+	keycode_map[VK_OEM_COMMA] = KEY_COMMA;
+	keycode_map[VK_OEM_3] = KEY_GRAVE_ACCENT;
+	keycode_map[VK_OEM_4] = KEY_LEFT_BRACKET;
+	keycode_map[VK_OEM_MINUS] = KEY_MINUS;
+	keycode_map[VK_OEM_PERIOD] = KEY_PERIOD;
+	
+	keycode_map[VK_BACK] = KEY_BACKSPACE;
+	keycode_map[VK_DELETE] = KEY_DELETE;
+	keycode_map[VK_END] = KEY_END;
+	keycode_map[VK_RETURN] = KEY_ENTER;
+	keycode_map[VK_ESCAPE] = KEY_ESCAPE;
+	keycode_map[VK_HOME] = KEY_HOME;
+	keycode_map[VK_INSERT] = KEY_INSERT;
+	keycode_map[VK_MENU] = KEY_MENU;
+	keycode_map[VK_NEXT] = KEY_PAGE_DOWN;
+	keycode_map[VK_PRIOR] = KEY_PAGE_UP;
+	keycode_map[VK_PAUSE] = KEY_PAUSE;
+	keycode_map[VK_TAB] = KEY_TAB;
+	keycode_map[VK_CAPITAL] = KEY_CAPS_LOCK;
+	keycode_map[VK_NUMLOCK] = KEY_NUM_LOCK;
+	keycode_map[VK_SCROLL] = KEY_SCROLL_LOCK;
+	keycode_map[0x70] = KEY_F1;
+	keycode_map[0x71] = KEY_F2;
+	keycode_map[0x72] = KEY_F3;
+	keycode_map[0x73] = KEY_F4;
+	keycode_map[0x74] = KEY_F5;
+	keycode_map[0x75] = KEY_F6;
+	keycode_map[0x76] = KEY_F7;
+	keycode_map[0x77] = KEY_F8;
+	keycode_map[0x78] = KEY_F9;
+	keycode_map[0x79] = KEY_F10;
+	keycode_map[0x7A] = KEY_F11;
+	keycode_map[0x7B] = KEY_F12;
+	keycode_map[0x7C] = KEY_F13;
+	keycode_map[0x7D] = KEY_F14;
+	keycode_map[0x7E] = KEY_F15;
+	keycode_map[0x7F] = KEY_F16;
+	keycode_map[0x80] = KEY_F17;
+	keycode_map[0x81] = KEY_F18;
+	keycode_map[0x82] = KEY_F19;
+	keycode_map[0x83] = KEY_F20;
+	keycode_map[0x84] = KEY_F21;
+	keycode_map[0x85] = KEY_F22;
+	keycode_map[0x86] = KEY_F23;
+	keycode_map[0x87] = KEY_F24;
+	keycode_map[0x88] = KEY_LEFT_ALT;
+	keycode_map[VK_CONTROL] = KEY_LEFT_CONTROL;
+	keycode_map[VK_LCONTROL] = KEY_LEFT_CONTROL;
+	keycode_map[VK_LSHIFT] = KEY_LEFT_SHIFT;
+	keycode_map[VK_LWIN] = KEY_LEFT_SUPER;
+	keycode_map[VK_SNAPSHOT] = KEY_PRINT_SCREEN;
+	keycode_map[VK_RMENU] = KEY_RIGHT_ALT;
+	keycode_map[VK_RCONTROL] = KEY_RIGHT_CONTROL;
+	keycode_map[VK_RSHIFT] = KEY_RIGHT_SHIFT;
+	keycode_map[VK_RWIN] = KEY_RIGHT_SUPER;
+	keycode_map[VK_DOWN] = KEY_DOWN;
+	keycode_map[VK_LEFT] = KEY_LEFT;
+	keycode_map[VK_RIGHT] = KEY_RIGHT;
+	keycode_map[VK_UP] = KEY_UP;
+	
+	keycode_map[VK_NUMPAD0] = KEY_KP_0;
+	keycode_map[VK_NUMPAD1] = KEY_KP_1;
+	keycode_map[VK_NUMPAD2] = KEY_KP_2;
+	keycode_map[VK_NUMPAD3] = KEY_KP_3;
+	keycode_map[VK_NUMPAD4] = KEY_KP_4;
+	keycode_map[VK_NUMPAD5] = KEY_KP_5;
+	keycode_map[VK_NUMPAD6] = KEY_KP_6;
+	keycode_map[VK_NUMPAD7] = KEY_KP_7;
+	keycode_map[VK_NUMPAD8] = KEY_KP_8;
+	keycode_map[VK_NUMPAD9] = KEY_KP_9;
+	keycode_map[VK_ADD] = KEY_KP_ADD;
+	keycode_map[VK_DECIMAL] = KEY_KP_DECIMAL;
+	keycode_map[VK_DIVIDE] = KEY_KP_DIVIDE;
+	keycode_map[VK_MULTIPLY] = KEY_KP_MULTIPLY;
+	keycode_map[VK_SUBTRACT] = KEY_KP_SUBTRACT;
+}
+
+bool platform_file_exists(char *path)
+{
+	DWORD dwAttrib = GetFileAttributes(path);
+	
+	return (dwAttrib != INVALID_FILE_ATTRIBUTES && 
+			!(dwAttrib & FILE_ATTRIBUTE_DIRECTORY));
+}
+
+void platform_create_config_directory()
+{
+	char tmp[PATH_MAX];
+	if(SUCCEEDED(SHGetFolderPathA(0, CSIDL_LOCAL_APPDATA|CSIDL_FLAG_CREATE, NULL, 0, tmp)))
+	{
+		string_appendn(tmp, "/text-search", PATH_MAX);
+	}
+	
+	
+	if (!platform_directory_exists(tmp))
+	{
+		CreateDirectoryA(tmp, NULL);
+	}
+}
+
+char* get_config_save_location(char *buffer)
+{
+	if(SUCCEEDED(SHGetFolderPathA(0, CSIDL_LOCAL_APPDATA|CSIDL_FLAG_CREATE, NULL, 0, buffer)))
+	{
+		string_appendn(buffer, "\\text-search\\config.txt", MAX_INPUT_LENGTH);
+		return buffer;
+	}
+	
+	return 0;
+}
+
+void platform_show_message(platform_window *window, char *message, char *title)
+{
+	HWND handle = window ? window->window_handle : NULL;
+	MessageBox(handle, message, title, MB_ICONINFORMATION | MB_OK);
+}
+
+LRESULT CALLBACK main_window_callback(HWND window, UINT message, WPARAM wparam, LPARAM lparam)
+{
+	LRESULT result = 0;
+	
+	if (message == WM_SIZE)
+	{
+		u32 width = lparam&0xFFFF;
+		u32 height = lparam>>16;
+		
+		current_window_to_handle->width = width;
+		current_window_to_handle->height = height;
+	}
+	else if (message == WM_CHAR)
+	{
+		if (current_keyboard_to_handle->take_input)
+		{
+			char buf[5];
+			memset(buf, 0, 5);
+			char *ch = 0;
+			
+			wchar_t codep = wparam;
+			
+			WideCharToMultiByte(CP_UTF8, 0, &codep, 1, buf, 5 ,0,0);
+			
+			if (utf8len(buf) == 1)
+			{
+				char val = buf[0];
+				
+				if (current_keyboard_to_handle->input_mode == INPUT_NUMERIC)
+				{
+					if (!(val >= 48 && val <= 57))
+					{
+						ch = 0;
+					}
+					else
+					{
+						snprintf(buf, 2,  "%c", val);
+						ch = buf;
+					}
+				}
+				else if (val >= 32 && val <= 126)
+				{
+					snprintf(buf, 5, "%c", val);
+					ch = buf;
+				}
+			}
+			
+			if (ch != 0)
+				keyboard_handle_input_string(current_window_to_handle, current_keyboard_to_handle, ch);
+		}
+	}
+	else if (message == WM_MOUSELEAVE)
+	{
+		//current_mouse_to_handle->x = MOUSE_OFFSCREEN;
+		//current_mouse_to_handle->y = MOUSE_OFFSCREEN;
+	}
+	else if (message == WM_KILLFOCUS)
+	{
+		current_mouse_to_handle->x = MOUSE_OFFSCREEN;
+		current_mouse_to_handle->y = MOUSE_OFFSCREEN;
+		
+		current_window_to_handle->has_focus = false;
+		memset(current_keyboard_to_handle->keys, 0, MAX_KEYCODE);
+	}
+	else if (message == WM_SETFOCUS)
+	{
+		current_window_to_handle->has_focus = true;
+	}
+	else if (message == WM_KEYDOWN)
+	{
+		s32 key = wparam;
+		
+		current_keyboard_to_handle->keys[keycode_map[key]] = true;
+		current_keyboard_to_handle->input_keys[keycode_map[key]] = true;
+		
+		if (current_keyboard_to_handle->take_input)
+			keyboard_handle_input_string(current_window_to_handle, 
+										 current_keyboard_to_handle, 0);
+	}
+	else if (message == WM_KEYUP)
+	{
+		s32 key = wparam;
+		current_keyboard_to_handle->keys[keycode_map[key]] = false;
+		current_keyboard_to_handle->input_keys[keycode_map[key]] = false;
+	}
+	else if (message == WM_LBUTTONDOWN || 
+			 message == WM_RBUTTONDOWN ||
+			 message == WM_MBUTTONDOWN ||
+			 message == WM_MOUSEWHEEL)
+	{
+		bool is_left_down = wparam & MK_LBUTTON;
+		bool is_right_down = wparam & MK_RBUTTON;
+		bool is_middle_down = wparam & MK_MBUTTON;
+		
+		u64 ev_time = platform_get_time(TIME_FULL, TIME_MILI_S);
+		static u64  last_ev_time = 0;
+		
+		if (message == WM_MOUSEWHEEL)
+		{
+			s16 scroll_val = wparam>>16;
+			
+			if (scroll_val < 0)
+				current_mouse_to_handle->scroll_state = SCROLL_DOWN;
+			else
+				current_mouse_to_handle->scroll_state = SCROLL_UP;
+		}
+		
+		if (is_left_down)
+		{
+			if (ev_time - last_ev_time < 200)
+			{
+				current_mouse_to_handle->left_state |= MOUSE_DOUBLE_CLICK;
+			}
+			
+			current_mouse_to_handle->left_state |= MOUSE_DOWN;
+			current_mouse_to_handle->left_state |= MOUSE_CLICK;
+			
+			current_mouse_to_handle->total_move_x = 0;
+			current_mouse_to_handle->total_move_y = 0;
+			last_ev_time = ev_time;
+		}
+		if (is_right_down)
+		{
+			current_mouse_to_handle->right_state |= MOUSE_DOWN;
+			current_mouse_to_handle->right_state |= MOUSE_CLICK;
+		}
+	}
+	else if (message == WM_LBUTTONUP || 
+			 message == WM_RBUTTONUP ||
+			 message == WM_MBUTTONUP)
+	{
+		bool is_left_up = message == WM_LBUTTONUP;
+		bool is_right_up = message == WM_RBUTTONUP;
+		bool is_middle_up = message == WM_MBUTTONUP;
+		
+		if (is_left_up)
+		{
+			current_mouse_to_handle->left_state = MOUSE_RELEASE;
+		}
+		if (is_right_up)
+		{
+			current_mouse_to_handle->right_state = MOUSE_RELEASE;
+		}
+	}
+	else if (message == WM_MOUSEMOVE)
+	{
+		current_window_to_handle->curr_cursor_type = -999;
+		
+#if 0
+		s32 x = lparam&0xFFFF;
+		s32 y = lparam>>16;
+		
+		current_mouse_to_handle->x = x;
+		current_mouse_to_handle->y = y;
+#endif
+		
+		TRACKMOUSEEVENT track;
+		track.cbSize = sizeof(track);
+		track.dwFlags = TME_LEAVE;
+		track.hwndTrack = current_window_to_handle->window_handle;
+		TrackMouseEvent(&track);
+	}
+	else if (message == WM_GETMINMAXINFO)
+	{
+		MINMAXINFO *info = (MINMAXINFO*)lparam;
+		
+		info->ptMinTrackSize.x = current_window_to_handle->min_width;
+		info->ptMinTrackSize.y = current_window_to_handle->min_height;
+		
+		if (current_window_to_handle->max_width)
+			info->ptMaxTrackSize.x = current_window_to_handle->max_width;
+		if (current_window_to_handle->max_height)
+			info->ptMaxTrackSize.y = current_window_to_handle->max_height;
+	}
+	else if (message == WM_DESTROY)
+	{
+		current_window_to_handle->is_open = false;
+	}
+	else if (message == WM_CLOSE)
+	{
+		current_window_to_handle->is_open = false;
+	}
+	else
+	{
+		result = LSMProc(window, message, wparam, lparam);
+	}
+	
+	return result;
+}
+
+void platform_window_set_title(platform_window *window, char *name)
+{
+	SetWindowText(window->window_handle, name);
+}
+
+vec2 platform_get_window_size(platform_window *window)
+{
+	RECT rec;
+	GetWindowRect(window->window_handle, &rec);
+	vec2 res;
+	res.x = rec.right - rec.left;
+	res.y = rec.bottom - rec.top;
+	return res;
+}
+
+void platform_get_focus(platform_window *window)
+{
+	SetFocus(window->window_handle);
+}
+
+platform_window platform_open_window(char *name, u16 width, u16 height, u16 max_w, u16 max_h, u16 min_w, u16 min_h)
+{
+#if !defined(MODE_GDBDEBUG) && !defined(MODE_DEVELOPER)
+	ShowWindow(GetConsoleWindow(), SW_HIDE);
+#endif
+	
+	platform_window window;
+	window.has_focus = true;
+	window.window_handle = 0;
+	window.hdc = 0;
+	window.width = width;
+	window.height = height;
+	window.min_width = min_w;
+	window.min_height = min_h;
+	window.max_width = max_w;
+	window.max_height = max_h;
+	window.curr_cursor_type = -1;
+	window.next_cursor_type = CURSOR_DEFAULT;
+	
+	current_window_to_handle = &window;
+	
+	memset(&window.window_class, 0, sizeof(WNDCLASS));
+	window.window_class.style = CS_OWNDC;
+	window.window_class.lpfnWndProc = main_window_callback;
+	window.window_class.hInstance = instance;
+	window.window_class.lpszClassName = name;
+	window.window_class.hIcon = LoadIcon(NULL, IDI_WINLOGO);
+	//window.window_class.hCursor = LoadCursor(NULL, IDC_ARROW);
+	
+	if (RegisterClass(&window.window_class))
+	{
+		int style = WS_VISIBLE|WS_SYSMENU|WS_CAPTION|WS_MINIMIZEBOX;
+		
+		if (min_w != max_w && min_h != max_h)
+			style |= WS_SIZEBOX;
+		else
+			style |= WS_THICKFRAME;
+		
+		window.window_handle = CreateWindowEx(0,
+											  window.window_class.lpszClassName,
+											  name,
+											  style,
+											  CW_USEDEFAULT,
+											  CW_USEDEFAULT,
+											  width,
+											  height,
+											  0,
+											  0,
+											  instance,
+											  0);
+		
+		if (window.window_handle)
+		{
+			window.hdc = GetDC(window.window_handle);
+			
+			PIXELFORMATDESCRIPTOR format;
+			memset(&format, 0, sizeof(PIXELFORMATDESCRIPTOR));
+			format.nSize = sizeof(PIXELFORMATDESCRIPTOR);
+			format.nVersion = 1;
+			format.dwFlags = PFD_SUPPORT_OPENGL | PFD_DRAW_TO_WINDOW | PFD_DOUBLEBUFFER;
+			format.cColorBits = 24;
+			format.cAlphaBits = 8;
+			format.iLayerType = PFD_MAIN_PLANE; // PFD_TYPE_RGBA
+			s32 suggested_format_index = ChoosePixelFormat(window.hdc, &format);
+			
+			PIXELFORMATDESCRIPTOR actual_format;
+			DescribePixelFormat(window.hdc, suggested_format_index, sizeof(actual_format), &actual_format);
+			SetPixelFormat(window.hdc, suggested_format_index, &actual_format);
+			
+			window.gl_context = wglCreateContext(window.hdc);
+			
+			static HGLRC share_list = 0;
+			if (share_list == 0)
+			{
+				share_list = window.gl_context;
+			}
+			else
+			{
+				wglShareLists(share_list, window.gl_context);
+			}
+			
+			wglMakeCurrent(window.hdc, window.gl_context);
+			
+			ShowWindow(window.window_handle, cmd_show);
+			
+			// blending
+			glEnable(GL_DEPTH_TEST);
+			//glDepthMask(true);
+			//glClearDepth(50);
+			glDepthFunc(GL_LEQUAL);
+			
+			glEnable(GL_BLEND);
+			glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
+			
+			// setup multisampling
+#if 0
+			glEnable(GL_ALPHA_TEST);
+			glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE);
+			glEnable(GL_SAMPLE_ALPHA_TO_ONE);
+			glEnable(GL_MULTISAMPLE);
+			glHint(GL_MULTISAMPLE_FILTER_HINT_NV, GL_NICEST);
+#endif
+			
+			// https://stackoverflow.com/questions/5627229/sub-pixel-drawing-with-opengl
+			//glHint(GL_POINT_SMOOTH, GL_NICEST);
+			//glHint(GL_LINE_SMOOTH, GL_NICEST);
+			//glHint(GL_POLYGON_SMOOTH, GL_NICEST);
+			
+			//glEnable(GL_SMOOTH);
+			//glEnable(GL_POINT_SMOOTH);
+			//glEnable(GL_LINE_SMOOTH);
+			//glEnable(GL_POLYGON_SMOOTH);
+			//////////////////
+			
+			window.is_open = true;
+			
+			glMatrixMode(GL_PROJECTION);
+			glLoadIdentity();
+			glOrtho(0, width, height, 0, -1, 1);
+			
+			//GLint m_viewport[4];
+			//glGetIntegerv( GL_VIEWPORT, m_viewport );
+			//printf("%d %d %d %d\n", m_viewport[0], m_viewport[1], m_viewport[2], m_viewport[3]);
+			
+			glMatrixMode(GL_MODELVIEW);
+			
+			TRACKMOUSEEVENT track;
+			track.cbSize = sizeof(track);
+			track.dwFlags = TME_LEAVE;
+			track.hwndTrack = window.window_handle;
+			TrackMouseEvent(&track);
+		}
+		else
+		{
+			platform_show_message(0, "An error occured within Windows, please restart the program.", "Error");
+			abort();
+		}
+	}
+	else
+	{
+		platform_show_message(0, "An error occured within Windows, please restart the program.", "Error");
+		abort();
+	}
+	
+	platform_get_focus(&window);
+	
+	return window;
+}
+
+void platform_window_set_size(platform_window *window, u16 width, u16 height)
+{
+	s32 style = GetWindowLong(window->window_handle, GWL_STYLE);
+	BOOL menu = FALSE;
+	
+	RECT rec;
+	rec.left = 0;
+	rec.top = 0;
+	rec.right = width;
+	rec.bottom = height;
+	AdjustWindowRectEx(&rec, style, menu, 0);
+	SetWindowPos(window->window_handle,NULL,rec.left,rec.top,rec.right,rec.bottom,SWP_NOMOVE|SWP_NOZORDER);
+}
+
+void platform_window_set_position(platform_window *window, u16 x, u16 y)
+{
+	RECT rec;
+	GetWindowRect(window->window_handle, &rec);
+	MoveWindow(window->window_handle, x, y, rec.right-rec.left, rec.bottom-rec.top, FALSE);
+}
+
+bool platform_window_is_valid(platform_window *window)
+{
+	return window->hdc && window->window_handle;
+}
+
+void platform_destroy_window(platform_window *window)
+{
+	wglMakeCurrent(NULL, NULL);
+	wglDeleteContext(window->gl_context);
+	
+	ReleaseDC(window->window_handle, window->hdc);
+	CloseWindow(window->window_handle);
+	DestroyWindow(window->window_handle);
+	UnregisterClassA(window->window_class.lpszClassName, instance);
+	window->hdc = 0;
+	window->window_handle = 0;
+}
+
+void platform_handle_events(platform_window *window, mouse_input *mouse, keyboard_input *keyboard)
+{
+	current_window_to_handle = window;
+	current_keyboard_to_handle = keyboard;
+	current_mouse_to_handle = mouse;
+	
+	mouse->left_state &= ~MOUSE_CLICK;
+	mouse->right_state &= ~MOUSE_CLICK;
+	mouse->left_state &= ~MOUSE_DOUBLE_CLICK;
+	mouse->right_state &= ~MOUSE_DOUBLE_CLICK;
+	mouse->left_state &= ~MOUSE_RELEASE;
+	mouse->right_state &= ~MOUSE_RELEASE;
+	memset(keyboard->input_keys, 0, MAX_KEYCODE);
+	mouse->move_x = 0;
+	mouse->move_y = 0;
+	mouse->scroll_state = 0;
+	keyboard->text_changed = false;
+	
+	// mouse position (including outside of window)
+	current_window_to_handle->has_focus = GetFocus() == current_window_to_handle->window_handle;
+	
+	if (current_window_to_handle->has_focus)
+	{
+		if((GetKeyState(VK_LBUTTON) & 0x100) == 0)
+		{
+			current_mouse_to_handle->left_state = MOUSE_RELEASE;
+		}
+		
+		RECT rec;
+		GetWindowRect(window->window_handle, &rec);
+		POINT p;
+		GetCursorPos(&p);
+		mouse->x = p.x - rec.left - GetSystemMetrics(SM_CYSIZEFRAME);
+		mouse->y = p.y - rec.top - GetSystemMetrics(SM_CYSIZE) - GetSystemMetrics(SM_CYFRAME);
+		//printf("%d %d\n",GetSystemMetrics(SM_CYSIZE), GetSystemMetrics(SM_CYFRAME));
+	}
+	
+	MSG message;
+	while(PeekMessageA(&message, window->window_handle, 0, 0, TRUE))
+	{
+		TranslateMessage(&message); 
+		SizingCheck(&message);
+		DispatchMessage(&message); 
+	}
+	
+	glViewport(0, 0, window->width, window->height);
+}
+
+void platform_window_swap_buffers(platform_window *window)
+{
+	// set cursor if changed
+	if (window->curr_cursor_type != window->next_cursor_type)
+	{
+		char *cursor_shape = 0;
+		switch(window->next_cursor_type)
+		{
+			case CURSOR_DEFAULT: cursor_shape = IDC_ARROW; break;
+			case CURSOR_POINTER: cursor_shape = IDC_HAND; break;
+		}
+		
+		HCURSOR cursor = LoadCursorA(NULL, cursor_shape);
+		
+		window->curr_cursor_type = window->next_cursor_type;
+		SetCursor(cursor);
+	}
+	
+	SwapBuffers(window->hdc);
+}
+
+file_content platform_read_file_content(char *path, const char *mode)
+{
+	file_content result;
+	result.content = 0;
+	result.content_length = 0;
+	result.file_error = 0;
+	
+	FILE *file = fopen(path, mode);
+	if (!file) 
+	{
+		if (errno == EMFILE)
+			result.file_error = FILE_ERROR_TOO_MANY_OPEN_FILES_PROCESS;
+		else if (errno == ENFILE)
+			result.file_error = FILE_ERROR_TOO_MANY_OPEN_FILES_SYSTEM;
+		else if (errno == EACCES)
+			result.file_error = FILE_ERROR_NO_ACCESS;
+		else if (errno == EPERM)
+			result.file_error = FILE_ERROR_NO_ACCESS;
+		else if (errno == ENOENT)
+			result.file_error = FILE_ERROR_NOT_FOUND;
+		else if (errno == ECONNABORTED)
+			result.file_error = FILE_ERROR_CONNECTION_ABORTED;
+		else if (errno == ECONNREFUSED)
+			result.file_error = FILE_ERROR_CONNECTION_REFUSED;
+		else if (errno == ENETDOWN)
+			result.file_error = FILE_ERROR_NETWORK_DOWN;
+		else
+		{
+			result.file_error = FILE_ERROR_GENERIC;
+			printf("ERROR: %d\n", errno);
+		}
+		
+		goto done_failure;
+	}
+	
+	fseek(file, 0 , SEEK_END);
+	int length = ftell(file);
+	fseek(file, 0, SEEK_SET);
+	
+	s32 length_to_alloc = length+1;
+	
+	result.content = mem_alloc(length_to_alloc);
+	if (!result.content) goto done;
+	
+	memset(result.content, 0, length);
+	s32 read_result = fread(result.content, 1, length, file);
+	if (read_result == 0 && length != 0)
+	{
+		mem_free(result.content);
+		result.content = 0;
+		return result;
+	}
+	
+	result.content_length = read_result;
+	
+	((char*)result.content)[length] = 0;
+	
+	done:
+	fclose(file);
+	done_failure:
+	return result;
+}
+
+bool platform_write_file_content(char *path, const char *mode, char *buffer, s32 len)
+{
+	bool result = false;
+	
+	FILE *file = fopen(path, mode);
+	
+	if (!file)
+	{
+		goto done_failure;
+	}
+	else
+	{
+		fprintf(file, buffer);
+	}
+	
+	//done:
+	fclose(file);
+	done_failure:
+	return result;
+}
+
+void platform_destroy_file_content(file_content *content)
+{
+	assert(content);
+	mem_free(content->content);
+}
+
+bool get_active_directory(char *buffer)
+{
+	return GetCurrentDirectory(MAX_INPUT_LENGTH, buffer);
+}
+
+bool set_active_directory(char *path)
+{
+	return SetCurrentDirectory(path);
+}
+
+void platform_list_files_block(array *list, char *start_dir, array filters, bool recursive, memory_bucket *bucket,  bool include_directories, bool *is_cancelled)
+{
+	assert(list);
+	s32 len = 0;
+	char *matched_filter = 0;
+	
+	char *subdirname_buf;
+	if (bucket)
+		subdirname_buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+	else
+		subdirname_buf = mem_alloc(MAX_INPUT_LENGTH);
+	
+	char *start_dir_fix;
+	if (bucket)
+		start_dir_fix = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+	else
+		start_dir_fix = mem_alloc(MAX_INPUT_LENGTH);
+	snprintf(start_dir_fix, MAX_INPUT_LENGTH, "%s*", start_dir);
+	
+	char *start_dir_clean;
+	if (bucket)
+		start_dir_clean = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+	else
+		start_dir_clean = mem_alloc(MAX_INPUT_LENGTH);
+	string_copyn(start_dir_clean, start_dir, MAX_INPUT_LENGTH);
+	
+	WIN32_FIND_DATAA file_info;
+	HWND handle = FindFirstFileA(start_dir_fix, &file_info);
+	
+	if (!bucket)
+		mem_free(start_dir_fix);
+	
+	if (handle == INVALID_HANDLE_VALUE)
+	{
+		return;
+	}
+	
+	do
+	{
+		if (*is_cancelled) break;
+		char *name = file_info.cFileName;
+		
+		// symbolic link is not allowed..
+		if ((file_info.dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT))
+			continue;
+		
+		
+		if ((file_info.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY))
+		{
+			if ((strcmp(name, ".") == 0) || (strcmp(name, "..") == 0))
+				continue;
+			
+			if (include_directories)
+			{
+				if ((len = filter_matches(&filters, name, 
+										  &matched_filter)) && len != -1)
+				{
+					// is file
+					char *buf;
+					if (bucket)
+						buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+					else
+						buf = mem_alloc(MAX_INPUT_LENGTH);
+					snprintf(buf, MAX_INPUT_LENGTH, "%s%s",start_dir, name);
+					
+					found_file f;
+					f.path = buf;
+					
+					if (bucket)
+						f.matched_filter= memory_bucket_reserve(bucket, len+1);
+					else
+						f.matched_filter= mem_alloc(len+1);
+					
+					string_copyn(f.matched_filter, matched_filter, len+1);
+					
+					mutex_lock(&list->mutex);
+					array_push_size(list, &f, sizeof(found_file));
+					mutex_unlock(&list->mutex);
+				}
+			}
+			
+			if (recursive)
+			{
+				string_copyn(subdirname_buf, start_dir_clean, MAX_INPUT_LENGTH);
+				string_appendn(subdirname_buf, name, MAX_INPUT_LENGTH);
+				string_appendn(subdirname_buf, "\\", MAX_INPUT_LENGTH);
+				
+				// is directory
+				platform_list_files_block(list, subdirname_buf, filters, recursive, bucket, include_directories, is_cancelled);
+			}
+		}
+		else if ((file_info.dwFileAttributes & FILE_ATTRIBUTE_COMPRESSED) ||
+				 (file_info.dwFileAttributes & FILE_ATTRIBUTE_ENCRYPTED) ||
+				 (file_info.dwFileAttributes & FILE_ATTRIBUTE_HIDDEN) ||
+				 (file_info.dwFileAttributes & FILE_ATTRIBUTE_NORMAL) || 
+				 (file_info.dwFileAttributes & FILE_ATTRIBUTE_READONLY) ||
+				 (file_info.dwFileAttributes & FILE_ATTRIBUTE_ARCHIVE))
+		{
+			if ((len = filter_matches(&filters, name, 
+									  &matched_filter)) && len != -1)
+			{
+				// is file
+				char *buf;
+				if (bucket)
+					buf = memory_bucket_reserve(bucket, MAX_INPUT_LENGTH);
+				else
+					buf = mem_alloc(MAX_INPUT_LENGTH);
+				
+				snprintf(buf, MAX_INPUT_LENGTH, "%s%s",start_dir, name);
+				
+				found_file f;
+				f.path = buf;
+				
+				if (bucket)
+					f.matched_filter = memory_bucket_reserve(bucket, len+1);
+				else
+					f.matched_filter = mem_alloc(len+1);
+				
+				string_copyn(f.matched_filter, matched_filter, len+1);
+				
+				mutex_lock(&list->mutex);
+				array_push_size(list, &f, sizeof(found_file));
+				mutex_unlock(&list->mutex);
+			}
+		}
+	}
+	while (FindNextFile(handle, &file_info) != 0);
+	
+	if (!bucket)
+		mem_free(start_dir_clean);
+	
+	FindClose(handle);
+}
+
+static void* platform_open_file_dialog_implementation(void *data)
+{
+	struct open_dialog_args *args = data;
+	
+	OPENFILENAME info;
+	info.lStructSize = sizeof(OPENFILENAME);
+	info.hwndOwner = NULL;
+	info.hInstance = NULL;
+	
+	char filter[MAX_INPUT_LENGTH];
+	memset(filter, 0, MAX_INPUT_LENGTH);
+	
+	if (args->file_filter)
+	{
+		string_copyn(filter, args->file_filter, MAX_INPUT_LENGTH);
+		filter[strlen(filter)] = 0;
+		filter[strlen(filter)+1] = 0;
+		info.lpstrFilter = filter;
+	}
+	else
+	{
+		info.lpstrFilter = NULL;
+	}
+	
+	char szFile[MAX_INPUT_LENGTH];
+	memset(szFile, 0, MAX_INPUT_LENGTH);
+	
+	info.lpstrCustomFilter = NULL;
+	info.nMaxCustFilter = MAX_INPUT_LENGTH;
+	info.nFilterIndex = 0;
+	info.lpstrFile = (char*)szFile;
+	info.nMaxFile = MAX_INPUT_LENGTH;
+	
+	info.lpstrDefExt = args->default_save_file_extension;
+	
+	info.lpstrFileTitle = NULL;
+	info.lpstrInitialDir = args->start_path;
+	info.lpstrTitle = NULL;
+	
+	if (args->type == SAVE_FILE)
+	{
+		info.Flags = OFN_EXTENSIONDIFFERENT | OFN_OVERWRITEPROMPT;
+		GetSaveFileNameA(&info);
+		string_copyn(args->buffer, info.lpstrFile, MAX_INPUT_LENGTH);
+	}
+	else if (args->type == OPEN_DIRECTORY)
+	{
+		BROWSEINFOA inf;
+		PIDLIST_ABSOLUTE result = SHBrowseForFolderA(&inf);
+		if (!result) return 0;
+		
+		SHGetPathFromIDListA(result, args->buffer);
+	}
+	else if (args->type == OPEN_FILE)
+	{
+		info.Flags = OFN_FILEMUSTEXIST;
+		GetOpenFileNameA(&info);
+		string_copyn(args->buffer, info.lpstrFile, MAX_INPUT_LENGTH);
+	}
+	
+	return 0;
+}
+
+void *platform_open_file_dialog_block(void *arg)
+{
+	platform_open_file_dialog_implementation(arg);
+	mem_free(arg);
+	return 0;
+}
+
+char *platform_get_full_path(char *file)
+{
+	char *buf = mem_alloc(MAX_INPUT_LENGTH);
+	if (!GetFullPathNameA(file, MAX_INPUT_LENGTH, buf, NULL))
+	{
+		buf[0] = 0;
+	}
+	
+	return buf;
+}
+
+void platform_open_url(char *command)
+{
+	platform_run_command(command);
+}
+
+void platform_run_command(char *command)
+{
+	// might be start instead of open
+	ShellExecuteA(NULL, "open", command, NULL, NULL, SW_SHOWDEFAULT);
+}
+
+void platform_window_make_current(platform_window *window)
+{
+	wglMakeCurrent(window->hdc, window->gl_context);
+}
+
+void platform_init(int argc, char **argv)
+{
+	setlocale(LC_ALL, "en_US.UTF-8");
+	
+	QueryPerformanceFrequency(&perf_frequency);
+	CoInitialize(NULL);
+	create_key_tables();
+	
+	instance = GetModuleHandle(NULL);
+	cmd_show = argc;
+	
+	// get fullpath of the directory the exe is residing in
+	binary_path = platform_get_full_path(argv[0]);
+	
+	platform_create_config_directory();
+	
+	char buf[MAX_INPUT_LENGTH];
+	get_directory_from_path(buf, binary_path);
+	string_copyn(binary_path, buf, MAX_INPUT_LENGTH);
+	
+	assets_create();
+}
+
+void platform_set_icon(platform_window *window, image *img)
+{
+	BYTE *bmp;
+	s32 data_len = img->width * img->height * 4;
+	s32 total_len = data_len + 40 * 4;
+	
+	bmp = mem_alloc(total_len);
+	
+	struct {
+		int32_t header_size, width, geight;
+		int16_t color_plane, bits_per_pixel;
+		int32_t compression_mode, img_length, obsolete[4];
+	} bmp_header = {40, img->width, img->height * 2, 1, 32, BI_RGB, data_len, {0,0,0,0} };
+	
+	memcpy(bmp, &bmp_header, 40);
+	
+	s32 index = 0;
+	for (s32 y = img->height-1; y >= 0; y--)
+	{
+		for (s32 x = 0; x < img->width; x++)
+		{
+			s32 img_pixel = *(((s32*)img->data+(x+(y*img->width))));
+			
+			// 0xAABBGGRR
+			s32 a = (img_pixel>>24) & 0x000000FF;
+			s32 b = (img_pixel>>16) & 0x000000FF;
+			s32 g = (img_pixel>> 8) & 0x000000FF;
+			s32 r = (img_pixel>> 0) & 0x000000FF;
+			
+			//s32 c = (r << 24) | (g << 16) | (b << 8) | (a << 0);
+			s32 c = (a << 24) | (r << 16) | (g << 8) | (b << 0);
+			memcpy(bmp+40+(index*4), &c, 4);
+			
+			++index;
+		}
+	}
+	
+	HICON icon = CreateIconFromResourceEx(bmp, total_len, TRUE, 0x00030000, img->width, img->height, LR_DEFAULTCOLOR);
+	
+	SendMessage(window->window_handle, WM_SETICON, ICON_SMALL, (LPARAM)icon);
+	SendMessage(window->window_handle, WM_SETICON, ICON_BIG, (LPARAM)icon);
+	
+	mem_free(bmp);
+	
+	if (!icon)
+		printf("Failed to load icon, error code: %ld.\n", GetLastError());
+}
+
+u64 platform_get_time(time_type time_type, time_precision precision)
+{
+	LARGE_INTEGER counter;
+	QueryPerformanceCounter(&counter);
+	
+	double sec = counter.QuadPart / (double)(perf_frequency.QuadPart);
+	
+	//printf("%I64d %I64d %f\n", counter.QuadPart, perf_frequency.QuadPart, sec);
+	
+	double val = sec;
+	
+	if (precision == TIME_NS)
+	{
+		return val*1000000000;
+	}
+	if (precision == TIME_US)
+	{
+		return val*1000000;
+	}
+	if (precision == TIME_MILI_S)
+	{
+		return val*1000;
+	}
+	if (precision == TIME_S)
+	{
+		return val;
+	}
+	return val;
+}
+
+s32 platform_get_memory_size()
+{
+	u64 result;
+	GetPhysicallyInstalledSystemMemory(&result);
+	return result;
+}
+
+s32 platform_get_cpu_count()
+{
+	SYSTEM_INFO info;
+	GetSystemInfo(&info);
+	
+	return info.dwNumberOfProcessors;
+}
+
+u64 string_to_u64(char *str)
+{
+	return (u64)strtoull(str, 0, 10);
+}
+
+u32 string_to_u32(char *str)
+{
+	return (u32)strtoul(str, 0, 10);
+}
+
+u16 string_to_u16(char *str)
+{
+	return (u16)strtoul(str, 0, 10);
+}
+
+u8 string_to_u8(char *str)
+{
+	return (u8)strtoul(str, 0, 10);
+}
+
+s64 string_to_s64(char *str)
+{
+	return (s64)strtoull(str, 0, 10);
+}
+
+s32 string_to_s32(char *str)
+{
+	return (s32)strtoul(str, 0, 10);
+}
+
+s16 string_to_s16(char *str)
+{
+	return (s16)strtoul(str, 0, 10);
+}
+
+s8 string_to_s8(char *str)
+{
+	return (s8)strtoul(str, 0, 10);
+}
\ No newline at end of file
diff --git a/src/windows/thread.c b/src/windows/thread.c
new file mode 100644
index 0000000..9669158
--- /dev/null
+++ b/src/windows/thread.c
@@ -0,0 +1,102 @@
+/* 
+*  BSD 2-Clause “Simplified” License
+*  Copyright (c) 2019, Aldrik Ramaekers, aldrik.ramaekers@protonmail.com
+*  All rights reserved.
+*/
+
+#include <synchapi.h>
+
+thread thread_start(void *(*start_routine) (void *), void *arg)
+{
+	thread result;
+	result.valid = false;
+	
+	result.thread = CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE)start_routine, 
+								 arg, 0, NULL);
+	result.valid = true;
+	
+	return result;
+}
+
+void thread_join(thread *thread)
+{
+	if (thread->valid)
+	{
+		WaitForSingleObject(thread->thread, INFINITE);
+		CloseHandle(thread->thread);
+	}
+}
+
+bool thread_tryjoin(thread *thread)
+{
+	if (thread->valid)
+	{
+		s32 result = WaitForSingleObject(thread->thread, 0);
+		return result == WAIT_OBJECT_0;
+	}
+	return false;
+}
+
+void thread_detach(thread *thread)
+{
+	if (thread->valid)
+	{
+		CloseHandle(thread->thread);
+	}
+}
+
+void thread_stop(thread *thread)
+{
+	if (thread->valid)
+	{
+		SuspendThread(thread->thread);
+	}
+}
+
+u32 thread_get_id()
+{
+	return GetCurrentThreadId();
+}
+
+void thread_sleep(u64 microseconds)
+{
+	Sleep(microseconds/1000);
+}
+
+mutex mutex_create()
+{
+	mutex result;
+	result.mutex = CreateMutex( 
+        NULL,              // default security attributes
+        FALSE,             // initially not owned
+        NULL);             // unnamed mutex
+	
+	return result;
+}
+
+mutex mutex_create_recursive()
+{
+	return mutex_create();
+}
+
+void mutex_lock(mutex *mutex)
+{
+	WaitForSingleObject( 
+		mutex->mutex,    // handle to mutex
+		INFINITE);  // no time-out interval
+}
+
+bool mutex_trylock(mutex *mutex)
+{
+	return WaitForSingleObject(mutex->mutex, 1) == WAIT_OBJECT_0;
+}
+
+void mutex_unlock(mutex *mutex)
+{
+	ReleaseMutex(mutex->mutex);
+}
+
+void mutex_destroy(mutex *mutex)
+{
+	CloseHandle(mutex->mutex);
+}