text feature requests about vkvg HOT 27 CLOSED

I like just calling it vkvg_text_run_create_with_length.

from vkvg.

2. I think the easy thing would be to just store a pointer to the raw font data instead of the path in `vkvg_font_identity_t`, and then just skip the `fopen` (stbtt) or use `FT_New_Memory_Face`.

This sounded a so good and straightforward idea that I've implemented it . #89
I've quickly test it, tell me if it fits your needs.

from vkvg.

3. I think this would just be a public API for `glyphs` in `vkvg_text_run_t`

I've implemented vkvg_text_run_get_glyph_positions in the same branch, the struc returned will be different with hb or not, but it's a compile time feature...

from vkvg.

Or maybe, to keep it simple, just a non-variadic function which always requires the length parameter, in which case for 0-terminated strings a strlen(text) would suffice?

Ping @jpbruyere.

from vkvg.

I've implemented vkvg_text_run_get_glyph_positions in the same branch, the struc returned will be different with hb or not, but it's a compile time feature...

Thanks. It's unfortunate that hb_glyph_position_t is partially opaque with an undefined length.

One option is to define vkvg_glyph_info_t in the public API and internally just copy the 5 fields you need from hb_glyph_position_t over.

from vkvg.

So rasterized for smaller sizes? Yeah, that would probably be ideal.

from vkvg.

stb fonts are not working because of this line:

assert(stbtt_InitFont(&font->stbInfo, font->fontBuffer, 0) && "stbtt_initFont failed");

assert() gets compiled out in release code, the font never gets initialized, and it segfaults.

from vkvg.

In _font_cache_create_text_run, strlen still gets called on all strings.

from vkvg.

I was assuming that assert would not discard the expression in release mode, but it looks like it's also discarded with gcc. I've pushed the fix on master, thanks for the report.

from vkvg.

In _font_cache_create_text_run, strlen still gets called on all strings.

I've pushed the removal on master (311d063).

from vkvg.

return array of glyph positions. useful for selecting text with a mouse.

I think that would be useful in pangovkvg, too.

from vkvg.

I've implemented your first point by adding vkvg_text_run_create2 with a length parameter (glyph count). ( #88 )
We should find a better name for the method, and also perform tests. This point was quiet straightforward to implement...

from vkvg.

What about making vkvg_text_run_create a variadic function, with an optional length parameter, like the open system call? So you don't have to maintain two functions.

from vkvg.

What about making vkvg_text_run_create a variadic function, with an optional length parameter, like the open system call? So you don't have to maintain two functions.

Bof, higher level wrappers may have problems to handle variadic func, and it add complexity to the code for a low gain. (in that case)

from vkvg.

Or maybe, to keep it simple, just a non-variadic function which always requires the length parameter, in which case for 0-terminated strings a strlen(text) would suffice?
harfbuzz compute length for null terminated string, also utf8 string is expected with the character count , not the byte length.

from vkvg.

Oh ok

from vkvg.

1. support for text strings that are not null terminated.  just a convenience for rendering long/multi-line text.

So I've quickly implemented it, but not yet tested.

2. load font from memory

For now, each font size has it's own face instance. I have to study this feature combined with 'single face shared among sizes', this is some work.

3. return array of glyph positions.  useful for selecting text with a mouse.

For harfbuzz, this is straightforward, I have to study how to do without.

4. font kerning without harfbuzz

I've never played with kerning by myself, maybe that's not too much work. But without harfbuzz, text shaping support will be minimal.

from vkvg.

1. Thank you.
2. I think the easy thing would be to just store a pointer to the raw font data instead of the path in vkvg_font_identity_t, and then just skip the fopen (stbtt) or use FT_New_Memory_Face.
3. I think this would just be a public API for glyphs in vkvg_text_run_t
4. Freetype has FT_Get_Kerning and stbtt has stbtt_GetGlyphKernAdvance.

from vkvg.

This sounded a so good and straightforward idea that I've implemented it . #89 I've quickly test it, tell me if it fits your needs.

Awesome. I just checked, this is basically what nanovg does. Although it only has one font struct. It autoscales that font size with the transformation matrix. This looks better better when zooming in, but probably uses a lot more memory. The ideal solution is probably using mipmapping with the requested size and a set of arbitrarily chosen larger sizes scaled down.

from vkvg.

Ah yes, an opaque pointer, so I would have had to implement some query functions on it. What about returning individual position elements by glyph index, I've pushed the fix.

from vkvg.

This sounded a so good and straightforward idea that I've implemented it . #89 I've quickly test it, tell me if it fits your needs.

Awesome. I just checked, this is basically what nanovg does. Although it only has one font struct. It autoscales that font size with the transformation matrix. This looks better better when zooming in, but probably uses a lot more memory. The ideal solution is probably using mipmapping with the requested size and a set of arbitrarily chosen larger sizes scaled down.

Due to font subpixel filtering, mipmapping could hardly improve, unless mipmaps are constructed with font rasteriser, but it could be tricky to implement. I plan to implement sdf fonts for adaptative scaling...

from vkvg.

vkvg don't use the freetype font matrix transform, cairo does, at least in some cases. This transforms also the subpixel filter result (I need to double check, but if not, I can't see another advantage to have a font matrix), which gives better result. The vkvg's font cache store untransformed glyphs for each sizes. The context matrix transformation gives poor results on fonts due to the existing subpixel filter on cached glyphs.

from vkvg.

another option for better adaptative scaling, is uploading vectorized glyphs instead of rasterized bitmaps per sizes...

from vkvg.

normally, the method should be called in release without the assert check. Do you experienced that the complete line is discarded? even the stbtt_InitFont call?

from vkvg.

Correct, from assert.h:

ifdef NDEBUG
#define assert(x) (void)0
#else
#define assert(x) ((void)((x) || (__assert_fail(#x, __FILE__, __LINE__, __func__),0)))
#endif

from vkvg.

I've confirmed the above fix works.

from vkvg.

3. return array of glyph positions. useful for selecting text with a mouse.

How would you implement text selection?

from vkvg.