language-tools

Summary Cache internal ASTs across invocations and use for incremental reparsing instead of on every call The language server maintains a per-file cache that is passed through to , enabling TypeScript to reuse unchanged AST subtrees Falls back to full parsing when no prior AST is available or content is unchanged Benchmark results Isolated svelte2tsx (80 sequential single-char edits, realistic ~100 line component): Full LSP round-trip (type char → hover): Median: 3.1ms — the cache saves ~0.2-0.5ms per request, with the TS language service being the dominant cost How it works Inside , there are two calls that parse `svelte2tsxts.updateSourceFile(oldAst, newText, changeRange)Svelte2TsxCachecomputeScriptChangeRange()ts.TextChangeRangegetOrUpdateSourceFile()ts.updateSourceFile()ts.createSourceFile()ts.updateSourceFile()` preserves parent node references [x] Verified fallback to full parse when cache is empty or content unchanged 🤖 Generated with Claude Code

Summary accounts for ~10% of wall-clock time Most consumers never need the source map on the hot path: Language server: , , don't use the map. The mapper is already lazily initialized, but the map object itself was eagerly computed. svelte-check incremental mode: only uses , never The property is now a lazy getter that calls on first access and caches the result via Consumers that do access (TypeScript plugin, sourcemap tests) see no regression — the map is generated once on first access Benchmarks Measured with realistic generated Svelte components of various sizes: Test plan [x] All 365 svelte2tsx tests pass (including 16 sourcemap tests) [x] All 665 language-server tests pass 🤖 Generated with Claude Code

Fixes #2956