Text shaping #21

This PR adds two new functions, fcft_text_run_rasterize(), and fcft_text_run_destroy().

fcft_text_run_rasterize() is like fcft_glyph_rasterize() and fcft_grapheme_rasterize(), but takes a whole string, instead of a single character, or grapheme cluster.

fcft does not cache the resulting set of glyphs - this can be done by the calling application.

TODO

• documentation (including example code? Or refer to the example program)
• changelog
• bump fcft version to at least 2.3.91, (削除) or maybe go straight to 2.4.0 (削除ここまで)
• update README

Closes #12

Current state:

Screenshot of text rendering

• First row is rendered using fcft_glyph_rasterize().
• Second row is rendered using fcft_grapheme_rasterize().
• Third row is rendered using fcft_text_run_rasterize().

fcft_grapheme_rasterize() doesn't handle RTL at all, and requires the calling application to do the grapheme clustering (using e.g. utf8proc).

fcft_text_run_rasterize() does everything for you.

TODO: while the arabic text looks right, at least to me (who speaks, or reads, no arabic at all), I'm not sure the spacing in the hebrew text is correct.

note: the first two rows looking wrong is expected.

I think I've fixed the glyphs' x-offsets now:

Screenshot of text rendering

This was done by adding the HarfBuzz provided glyph x offset to the offset from FreeType, instead of replacing it. Still need to find example code, or documentation, that backs this up.

And I still can't make heads or tails of the glyphs' y offset.

And I still can't make heads or tails of the glyphs' y offset.

Looks like it was a bad font. With other fonts it's very clear that the vertical offset must be applied in the exact same way the horizontal offset is.

With Noto:

Screenshot of text rendering

One limitation of the current implementation is that LTR/RTL changes are only detected when/where we fail to lookup a glyph. In other words, if the primary font has all the glyphs, we never detect LTR/RTL changes.

@sterni is this something you'd be interrested in testing before I merge? If not, just say no :)

There's an example program (configure meson with -Dexamples=true to build) you can use to render arbitrary strings using the three currently available methods: individual glyphs, individual grapheme clusters and text runs. It is what you see in the screenshots above.

This will soon be released as fcft-2.4.0, and I plan to use it in yambar, fuzzel and fnott, but not in foot (due to all the previously discussed issues with text shaping in a terminal).