Add API to get glyph alternates

[behdad]

Such that apps can do fancy GUI.

[Tavmjong]

I've been working on a Font Features dialog for Inkscape and would love to have this implemented. Using hb_ot_layout_lookup_collect_glyphs, one can git a list of input and output glyphs but cannot tell absolutely which output glyph corresponds to which input glyph if there is not a one-to-one correspondence as with a type 3 lookup table.

[ebraminio]

AAT's Zapf also provides info related to this.

[behdad]

AAT's Zapf also provides info related to this.

That table is pretty big and very few fonts use it. I'm hesitant to expose it, at least not before we expose OpenType GSUB properly first.

[behdad]

Any volunteers for this?

[ebraminio]

Any volunteers for this?

Please explain what sort of API should be exposed for this.

[behdad]

For starters this would do:

HB_EXTERN hb_bool_t
hb_ot_layout_lookup_get_alternates (hb_face_t            *face,
                                    unsigned int          lookup_index,
                                    hb_codepoint_t    glyph,
                                    unsigned int         start_offset,
                                    unsigned int         alternate_count,
                                    hb_codepoint_t   *alternate_glyphs)
                                     ;

But maybe we can rename get_substitutes and make it work for single, alternate, and multiple substitutes and return a type enum.

In that case maybe can also take glyphstring and also work for ligatures. That's very similar to would_substitute API already. I think I like the latter.

[behdad]

Khaled, can you please work on this? I'm surprised you went the tour to parsing GSUB yourself instead.

https://twitter.com/KhaledGhetas/status/1212830043602337792?s=19

[khaledhosny]

I looked into implementing the proposed API above first, but got overwhelmed by HarfBuzz internals, and it would have taken me too much time than it took me to write the 200 or so JavaScript lines I needed to parse the few lookup types I needed for this project (I control the font, so I can do partial implementations and unsafe assumptions).

[behdad]

Fair enough. Can you please comment on the API based on your learning? I can give this a try. Should be a dozen lines of code at it's core.

[khaledhosny]

I think the would_substitute-like API is fine. I used would_substitute and was quite happy with it.

One issue I had with it is ligatures, to reliably get all ligature substitutions involving the current glyph, I need to try it with all subsequent glyphs adding one at a time, i.e. if I have the string abcdef and I want to check ligatures involving the a glyph, I need to try ab, abc, abcd, abcde, and abcdef as input since I don’t know if a is involved in any ligatures at all and if yes what is the maximum length of such ligatures. When reading the GSUB lookups directly I can optimize this a bit by checking if a is in the lookup coverage at all, the max componentCount, etc.