[layout] hb_ot_layout_lookup_get_glyph_alternates, a new API

[ebraminio]

Fixes #673

As #673 (comment) proposal, let's see how it goes.

Used the following to find a font having alternates,

#include "src/harfbuzz.cc"

int main(int argc, char **argv) {
  hb_blob_t *blob = hb_blob_create_from_file (argv[1]);
  unsigned count = hb_face_count (blob);
  for (unsigned i = 0; i < count; ++i) {
    hb_face_t *face = hb_face_create (blob, 0);
    unsigned glyph_count = hb_face_get_glyph_count (face);
    unsigned lookup_count = hb_ot_layout_table_get_lookup_count (face, HB_OT_TAG_GSUB);
    for (unsigned lookup = 0; lookup < lookup_count; ++lookup)
      for (unsigned gid = 0; gid < glyph_count; ++gid)
	if (hb_ot_layout_lookup_get_alternates (face, lookup, gid, 0, nullptr, nullptr))
	{
	  printf ("%s\n", argv[1]);
	  goto done;
	}
    done:
    hb_face_destroy (face);
  }
  hb_blob_destroy (blob);
  return 0;
}

[ebraminio]

just to mention, it works with Rana Kufi also very well as it was the font I was using for testing initially.

[behdad]

Thanks @ebraminio. This is very nice!

Generally we push most of the parameters from the API down to where they are populated. Instead you pull out the internal array and copy in the public API. Any particular reason you did it that way? If not, I prefer it if you do it the way other methods are done. Thanks again!

[ebraminio]

PTAL

[behdad]

Quick show of hands if name should be get_alternates, get_glyph_alternates, or get_alternate_glyphs?

I feel like get_alternates is wrong because it implies it returns alternates to a lookup.

The get_glyph_alternates rightly specifies that this is a lookup method that returns certain data about a glyph.

The get_alternate_glyphs doesn't specify the input, but the output.

As of now I prefer the second one, get_glyph_alternates.

[jfkthame]

Quick show of hands if name should be get_alternates, get_glyph_alternates, or get_alternate_glyphs?

I feel like get_alternates is wrong because it implies it returns alternates to a lookup.

The get_glyph_alternates rightly specifies that this is a lookup method that returns certain data about a glyph.

The get_alternate_glyphs doesn't specify the input, but the output.

As of now I prefer the second one, get_glyph_alternates.

Yeah, get_glyph_alternates sounds OK to me.

Does this handle alternates referenced via an Extension subtable?

[behdad]

Does this handle alternates referenced via an Extension subtable?

Doesn't. I'll address that in my review.

[behdad]

Continuing in #2473