[exKAZUu]

Why

Reduce parser storage and parsing cost with profiled compact tables and measured compiler settings. Compared with published 9.0.2: tree-sitter-java.wasm: 318,730 → 315,827 bytes (0.9% smaller).

Requirements

• required: Improve performance and binary size as much as possible in tree-sitter and our grammar repositories.
• required: Backward compatibility and the current tree structure need not be retained.
• required: For this PR, run review-booster, then complete-pr, then release.
• required: Preserve valid-input acceptance, accurate source coordinates, error reporting, and agreement between incremental and fresh parsing.
• chosen: Pin runtime 10.0.2 and regenerate with its matching CLI: it combines compact ABI 0x57420007 tables with lexical-continuation and incremental-recovery fixes.
• chosen: Use Wasm optimization z after comparing s/z/2/3 alternatives because it balances parsing throughput and module size.

Customer Summary

Use @willbooster/web-tree-sitter 10.0.2 or willbooster-tree-sitter 10.0.2 or later within 10.x. Parse-tree grouping, including recovery, may change.

Technical Summary

Regenerate profiled sparse parser tables with ABI 0x57420007, select the measured Wasm compiler setting, and update runtime dependencies and compatibility documentation.

Testing

• bun run verify passed.
• bun run build/ci passed with the matching v10.0.2 CLI from the core release workflow.
• All 75 unit tests passed; cargo test --locked passed.
• Corpus/example audit (example files up to 1,000,000 bytes each) against published 9.0.2: 8,104 inputs, no newly rejected inputs or valid tree changes. Full PR CI passed on the reviewed head.

Notes

Full-input Wasm comparison against published 9.0.2 measured throughput at parity (+0.04%) while reducing the module by 0.9%. Results describe the audited corpus/example workload.