Hamza is a Header-Only, Fast and Portable C99 Unicode/OpenType shaping and rendering library. It's designed to be a small, portable and optimized shaper that's easy to integrate into any existing project. Below is an image of a short string of Arabic shaped with this library using a fairly complex font, random colors are assigned to each glyph.
Hamza includes the single-file programs update_ucd_ftp
and generate_ucd_headers
. The first pulls the necessary UCD files from the FTP server at ftp.unicode.org and requires curl. The second generates optimized C headers from those UCD files. Both of these programs make use of the POSIX regex library for filtering and parsing.
Download the UCD txt, this might take a few minutes so only do if UCD headers are out of date:
./build/update_ucd_ftp
Generate the header files for the UCD versions:
./build/generate_ucd_headers
To start using Hamza, define HZ_IMPLEMENTATION
before including hz.h
. You can optionally define HZ_NO_STDLIB
for . It's also necessary to include the header for the UCD for the version you require.
We will explain later how these are generated and how you can update them yourself.
#define HZ_IMPLEMENTATION
#include <hz/hz_ucd_15_0_0.h>
#include <hz/hz.h>
To initialize the library first fill a hz_config_t
struct and call hz_init
:
hz_config_t cfg = {
};
if (hz_init(&cfg) != HZ_OK) {
fprintf(stderr, "%s\n", "Failed to initialize Hamza!");
return -1;
}
Next, before you can shape any text you must provide font data. You want to load a font into a stbtt_fontinfo
struct. Hamza includes stb_truetype.h
which is intended to be used in reading fonts. To create a hz_font_t
from a stbtt font, write:
hz_font_t *font = hz_stbtt_font_create(&fontinfo);
Hamza aims to let the user manage the memory allocation and the data as much as possible. Before shaping the font data has to be parsed into a hz_font_data_t
struct. This holds all the OpenType table data required for shaping with a specific font. The hz_font_data_init
function takes as argument how much memory will be allocated to hold that font's data:
hz_font_data_t font_data;
hz_font_data_init(&font_data, 1024*1024); // 1MiB
hz_font_data_load(&font_data, font);
Create a shaper and initialize it:
hz_shaper_t shaper;
hz_shaper_init(&shaper);
Set the shaper's required parameters:
hz_shaper_set_direction(&shaper, HZ_DIRECTION_RTL);
hz_shaper_set_script(&shaper, HZ_SCRIPT_ARABIC);
hz_shaper_set_language(&shaper, HZ_LANGUAGE_ARABIC);
Set the shaper's typography features:
hz_feature_t features[] = {
HZ_FEATURE_ISOL,
HZ_FEATURE_INIT,
HZ_FEATURE_MEDI,
HZ_FEATURE_FINA,
HZ_FEATURE_RLIG,
HZ_FEATURE_LIGA,
};
hz_shaper_set_features(&shaper, features, sizeof(features)/sizeof(features[0]));
Create glyph buffer and shape!
hz_buffer_t buffer;
hz_buffer_init(&buffer);
hz_shape_sz1(&shaper, &font_data, HZ_ENCODING_UTF8, "السلام عليكم", &buffer);
After this, you can access the buffer's glyph data and render. After you are done with everything you have to deinitialize.
hz_buffer_release(&buffer);
hz_font_data_release(&font_data);
hz_font_destroy(font);
hz_deinit();
- GCC 10.3.0 x86_64-w64-wingw32
- GCC 10.3.0 x86_64-w64-wingw32 (mingw64)
- MSVC 19.35.32217.1
- MSVC 19.29.30148.0
- Clang 16.0.0 x86_64-pc-windows-msvc
- Joining script support and RTL writing
- Kerning
- Ligatures
- Support for new OpenType language tags (mixture of ISO 639-3 and ISO 639-2 codes)
- Vertical layout Support (mostly for CJK, Mongolian, etc...)
- Color Emojis
- Emoji Combinations
- Multi-Threading
- Unicode Normalization (NFC,NFD,NFKC,NFKD)
- Open
.aat
.woff
and.woff2
formats
Hamza is licensed under LGPLv3.