Glyph reuse

[rsheeter]

Identify and exploit sets of shapes where for each set of shapes there exists one shape from which all the others in the set can be derived by application of a transform.

See existing code for this in nanoemoji:

• https://github.com/googlefonts/picosvg/blob/main/src/picosvg/svg_reuse.py, https://github.com/googlefonts/picosvg/blob/main/tests/svg_reuse_test.py
• Use precomputed part files for shape reuse nanoemoji#414 - incomplete upgrade, the pattern here is good starting point for a fontmake-rs implementation

Implementation in fontmake-rs:

• Emit reuse information per-glyph after each glyph completes
• Once all glyphs complete merge them and figure out what is reusable

[anthrotype]

can be derived by application of a transform

if here you mean glyf components, note that with variable fonts currently only the component x and y offset can be varied, not its 2x2 transform; that has to match across the designspace.

[rsheeter]

I had in mind to do a generic reuse detection and then choose whether to exploit what it found later, such as perhaps electing to not do it for glyf if the 2x2 varies. That will set us up well to compile COLR and to leverage variable components when they arrive courtesy of https://github.com/harfbuzz/boring-expansion-spec (optimistic hat firmly on).