Merge names from FEA into final table #1133

cmyr · 2024-11-19T23:54:30Z

Previously we were just dropping them on the floor.

This is one of a number of cases where it is possible for FEA to generate a table that is also generated (or could be generated) through the main compilation process.

This introduces an overall strategy for handling this: when FEA generates these extra tables we now stash them in the main context, where we can reference them when building the final font.

This patch also uses this mechanism to merge name records generated from FEA into the main name table; this looks to get us at least a +7 on crater.

As a bonus this includes a commit that fixes a little bug in ttx_diff where we were accidentally sorting name ids as if they were lookup ids, and which very briefly had me doubting my sanity.

fixes GSUB stylistic set names not being added to name table #1130

fea-rs/src/compile/tables/name.rs

rsheeter · 2024-11-20T03:30:53Z

fontbe/src/features.rs

@@ -56,6 +58,8 @@ pub use marks::create_mark_work;

 const DFLT_SCRIPT: Tag = Tag::new(b"DFLT");
 const DFLT_LANG: Tag = Tag::new(b"dflt");
+// https://learn.microsoft.com/en-us/typography/opentype/spec/name#name-ids
+const FIRST_NON_RESERVED_NAME_ID: u16 = 256;


I think I've seen this twice, I suggest putting it in fontdrasil which both fea-rs and fontbe are entitled to reference

fontbe/src/name.rs

resources/testdata/WghtVar-Bold.ufo/features.fea

rsheeter · 2024-11-20T03:41:26Z

resources/scripts/ttx_diff.py

Nit: this would have been nicer as it's own PR

As a compromise it's in its own commit 😬

fea-rs/src/compile/output.rs

resources/testdata/WghtVar-Bold.ufo/features.fea

Previously we were dropping these on the floor. The strategy here is simple enough: if FEA compilation generates any extra tables we now stash these in the context, and then we can merge them (so far, name only) into the final binary at the end.