v0.29.0 — recurring parser performance and compatibility cleanup
Highlights
v0.29.0 improves recurring parser performance and continues consolidating
compatibility behavior into auditable shared mechanisms.
- Recurring C# and CSS witnesses are 34–35% faster and allocate 81–82% fewer
bytes. - Across the selected six-language recurring family, wall time improves 3.18%
and allocated bytes improve 6.81% geometrically. - Full-parse allocated bytes fall 18.68% while the canonical benchmark trio
remains neutral. - Forest parsing reuses and promptly returns DFA token sources, small forest
indexes stay inline, GLR scratch is sized for the current input, and visible
alias targets are precomputed once per parser. - More compatibility and retry behavior moves from language-specific code to
shared tables and exact-blob runtime profiles.
Install with:
go get github.com/odvcencio/gotreesitter@v0.29.0See the v0.29.0 changelog for the complete change list.