Extract kerning and shaping tables from fonts?

[mikemorris]

tl;dr Can we use ICU or HarfBuzz to extract shaping and kerning tables from a font?

From discussion with @nickidlugash on adding kerning support:

FreeType only supports kerning via the (rather simple) ‘kern’ table. For the interpretation of kerning data in the (highly sophisticated) ‘GPOS’ table you need a higher-level library like ICU or HarfBuzz since it can be context dependent (this is, the kerning may vary depending on the position within a text string, for example). http://freetype.org/freetype2/docs/glyphs/glyphs-4.html

• FreeType has limited support for reading kerning tables and running FT_Get_Kerning for all possible glyph pairs seems like a terrible idea.
• We'll need to integrate HarfBuzz (and a JavaScript equivalent?) into our rendering engines at some point for complex text shaping, could we handle kerning at this point too?
• Is it possible to use ICU or HarfBuzz to extract shaping and kerning tables from a font at upload time to be stashed on S3 and used later for rendering, rather than having to load the full font file into HarfBuzz/ICU?

[springmeyer]

Good questions, I'll add this tidbit:

FreeType can now use the HarfBuzz library to greatly improve the
      auto-hinting of  fonts that  use OpenType features:  Many glyphs
      that are part  of such features but don't have  cmap entries are
      now handled  properly, for  example small caps  or superscripts.
      Define the configuration  macro FT_CONFIG_OPTION_USE_HARFBUZZ to
      activate HarfBuzz support

http://sourceforge.net/projects/freetype/files/freetype2/2.5.3/

[mattdesl]

Opentype.js has also been working on basic GPOS parsing: https://github.com/nodebox/opentype.js/issues/19

[mikemorris]

Current thinking is to implement something like Harfbuzz's reference_table method to extract and dump the binary SFNT table data while node-fontnik has the font open in FreeType, and write a custom backend for HarfBuzz to read this data by itself rather than needing to open up a font file at shaping time. @lbud is digging into how this will work.

I started working on porting pomax/A-binary-parser-generator to Node.js with Browserify today to try extracting and parsing the shaping tables in JavaScript, but I think that approach will end up being slower, more difficult and result in less useful data (JavaScript objects instead of binary data HarfBuzz expects) than just doing this through FreeType.

[felipesanches]

I just posted a few comments here while I actually intended to post them to issue #95. I have just deleted them from here and posted them to the correct issue. Sorry.