fix(extractors): recover exports dropped by tree-sitter's bare export+newline misparse

[carlos-alm]

Summary

tree-sitter-javascript/typescript misparses export followed by a newline before const/let/var/class/function/interface/type as a standalone (expression_statement (identifier)) node instead of a single export_statement (export default/{/* ARE handled correctly across a newline — the grammar's ASI-like heuristic special-cases only those). The declaration that follows is then extracted as an ordinary, internal (non-exported) top-level statement — the export relationship never exists in the parse tree, so no Export entry is ever created, in either engine.

Per the ECMAScript grammar, export Declaration has no [no LineTerminator here] restriction — unlike return/throw, ASI does not apply — so this is valid, correctly-exported JS/TS in every real engine.

Fix

export is a reserved word, so a real identifier node whose text is exactly "export" can only ever occur via this misparse — never a legitimate variable reference — making recovery unambiguous:

• TS walk path (src/extractors/javascript.ts): handleExpressionStmt detects the bare-export shape and calls the new recoverBareExportMisparse, which walks forward past any comment siblings to the next declaration and feeds it through collectExportedDeclarations — the same function a correctly-parsed export_statement's declaration already goes through.
• TS query path: new capture (expression_statement (identifier) @bare_export_kw) @bare_export_stmt added to COMMON_QUERY_PATTERNS in both src/domain/parser.ts and src/domain/wasm-worker-entry.ts (this codebase duplicates that list across the two build contexts). Filtered by text (=== 'export') in dispatchQueryMatch rather than a query predicate, since this codebase has no existing predicate usage.
• Rust native engine (crates/codegraph-core/src/extractors/javascript.rs): handle_expr_stmt gained the same check, calling the new recover_bare_export_misparse, mirroring the TS walk path and reusing handle_export_declaration.

Restricted to direct children of program in both engines — export is not valid syntax anywhere else a bare single-identifier expression statement could appear.

Out-of-scope finding filed separately

While investigating, found that enum_declaration is entirely absent from collectExportedDeclarations/handle_export_declaration in both engines — export enum Foo {} never produces an Export record even when correctly parsed (a real, pre-existing, unrelated bug). Filed as #2560 rather than folding into this PR.

Test plan

• New tests in tests/parsers/javascript.test.ts (walk path): recovers const/class/function/TS interface/TS type-alias, skips an intervening comment, still exports same-line declarations normally, and doesn't misfire on an unrelated identifier (notExport;)
• New tests in tests/engines/query-walk-parity.test.ts: a parity case proving query and walk paths agree, plus two dedicated query-path-only tests proving the query capture itself fires correctly (not just "agrees with walk")
• New tests in crates/codegraph-core/src/extractors/javascript.rs mirroring all of the above
• Revert-verified on both engines: temporarily disabled the fix (TS walk path, TS query dispatch, Rust) and confirmed all corresponding new tests fail with the exact expected pre-fix output ([] / empty exports); restored and confirmed green
• End-to-end verified against the real build pipeline (not just unit tests) — rebuilt the native .node addon, ran codegraph build --engine native and --engine wasm against a temp fixture with the exact repro from the issue, confirmed codegraph exports lists the recovered symbol as exported on both engines
• npm test (full suite) — 339 files, 5447+ passed
• cargo test --lib (full suite) — 1088 passed
• npx tsc --noEmit — clean
• npm run lint — clean

Closes #2459

[greptile-apps]

Greptile Summary

The PR recovers JavaScript and TypeScript exports lost when tree-sitter parses a newline-separated bare export as an identifier expression.

• Adds equivalent recovery logic to the TypeScript walk/query paths and native Rust extractor.
• Keeps the main-thread and WASM-worker query definitions synchronized.
• Adds walk, query, parity, native, TypeScript, comment-boundary, and negative-case coverage.

Confidence Score: 5/5

The PR appears safe to merge with no concrete changed-code failure identified.

The native, query, and walk implementations apply the same top-level reserved-word guard and reuse their normal export-declaration handlers, with focused parity and negative-case coverage.

Important Files Changed

Filename	Overview
src/extractors/javascript.ts	Adds guarded bare-export recovery to both query and walk extraction paths while reusing existing export classification.
crates/codegraph-core/src/extractors/javascript.rs	Mirrors the TypeScript recovery in the native extractor and adds corresponding native tests.
src/domain/parser.ts	Adds the bare-identifier expression capture required by main-thread query extraction.
src/domain/wasm-worker-entry.ts	Keeps the WASM worker's duplicated query patterns aligned with the main parser.
tests/engines/query-walk-parity.test.ts	Verifies query/walk parity and independently checks query-path recovery and filtering.
tests/parsers/javascript.test.ts	Covers recovered declaration kinds, intervening comments, normal exports, and unrelated identifiers.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Source["export + newline + declaration"] --> Parser["tree-sitter parse"]
  Parser --> Bare["bare export expression + declaration sibling"]
  Bare --> Query["TS query capture"]
  Bare --> Walk["TS walk recovery"]
  Bare --> Native["Rust native recovery"]
  Query --> Shared["normal exported-declaration collector"]
  Walk --> Shared
  Native --> NativeCollector["native export-declaration handler"]
  Shared --> Output["ExtractorOutput.exports"]
  NativeCollector --> Output

Loading

_{Reviews (1): Last reviewed commit: "fix(extractors): recover exports dropped..." | Re-trigger Greptile}

[github-actions]

Codegraph Impact Analysis

3 functions changed → 8 callers affected across 3 files

• dispatchQueryMatch in src/extractors/javascript.ts:371 (4 transitive callers)
• recoverBareExportMisparse in src/extractors/javascript.ts:2549 (6 transitive callers)
• handleExpressionStmt in src/extractors/javascript.ts:2580 (3 transitive callers)