Universal Shaping Engine

[devongovett]

This starts the implementation of the Universal Shaping Engine. It borrows heavily from the implementation in Harfbuzz. Uses a state-machine compiler to generate a DFA that finds syllables to be reordered as defined by the USE spec.

Still a bunch to do, but it passes the Balinese test cases in text-rendering-tests. 😎

TODO:

• Implement rphf and pref feature processing
• Implement topographical feature processing using the existing arabic shaper
• Make mark advance zeroing configurable by the shaper
• Clean up code
• Add tests

[devongovett]

@behdad do you know of any open source fonts that use the rphf/pref features for USE scripts? Couldn't find any in Noto.

Also, are there actually any scripts with arabic-style joining that aren't already handled by the arabic shaper? From looking at ArabicShaping.txt, it seems not. If this is the case, I may not need to implement that part of the spec for now.

[behdad]

@behdad do you know of any open source fonts that use the rphf/pref features for USE scripts? Couldn't find any in Noto.

I don't know of. We tested with Windows fonts when implementing. Noto fonts have not been updated to USE yet; though, Khmer from noto-source might be:

https://github.com/googlei18n/noto-source/tree/master/src

I haven't built that one yet. @jungshik

Also, are there actually any scripts with arabic-style joining that aren't already handled by the arabic shaper? From looking at ArabicShaping.txt, it seems not. If this is the case, I may not need to implement that part of the spec for now.

MS routes N'Ko, and probably Mongolian and Phags Pa, through USE. We don't.

[devongovett]

Implemented rphf and pref feature behavior, though I don't have a font to test with at the moment. Made mark advance zeroing configurable by the shaper. Not implementing joining behavior because all scripts that need that are already handled by the Arabic shaper. I think this is pretty much ready to go!