Skip to content

TODO.md

Latest commit

 

History

History
104 lines (82 loc) · 5.31 KB

TODO.md

File metadata and controls

104 lines (82 loc) · 5.31 KB

Last updated: 2026-03-01

codegraph-monorepo TODO

High Priority

14. LSP is_public doesn't read parser visibility property

  • All 13 parsers store visibility as a string ("public"/"protected"/"private")
  • LSP server's get_symbol_info only checks boolean is_public or exported, defaults to true
  • Result: Python _protected_function and __private_function both show is_public: true
  • Fix: LSP should fall back to visibility string when is_public bool is absent
  • Affects: engine.rs lines 405-409, 989-991, and node_to_symbol_info

Medium Priority

6. Expand test coverage for recent features

  • cpp: system include distinction (is_system property) — added in 1256a12
  • ruby: require_relative tracking — added in 1256a12
  • typescript: triple-slash reference directives — added in 1256a12
  • php: dynamic include/require resolution — added in 25c4dba
  • python: is_external fix — added in 25c4dba
  • rust: mod declaration import edges — added in d92e2ae
  • Integration tests exist but could use edge-case coverage

7. codegraph-tcl grammar robustness

  • tree-sitter-tcl uses a vendored ABI v15→v14 patched grammar
  • Position-0 splitting is worked around with sibling-stitching in visitor
  • Consider: upstream grammar fix, or switching to a maintained grammar fork
  • Collapsed word_list and fragmented body issues are handled but fragile

15. Remaining parser capability gaps

  • TypeScript: arrow functions stored as arrow_function instead of variable name (e.g., fetchData)
  • C#, Java, Kotlin, Ruby, Swift, PHP: no parameter extraction, no return type extraction
  • C#, Java, PHP, Ruby, Swift: no enum-as-class extraction
  • Kotlin, Ruby, Swift: no ABC/trait detection from class hierarchy
  • Use Rust parser as gold standard reference for feature parity

Future

8. Additional language parsers

  • No Lua, Perl, Scala, Dart, or Zig parsers yet
  • Add based on demand / stellarion user requests

9. Cross-language import resolution

  • Current parsers track imports within their own language
  • No cross-language linking (e.g., Python calling C extensions, TypeScript importing WASM)

10. Graph algorithm extensions

  • Current: BFS, DFS, cycle detection
  • Potential: strongly connected components, topological sort, PageRank for code importance

11. Storage backend alternatives

  • Current: RocksDB (production), BTreeMap (tests)
  • Potential: SQLite for zero-config, PostgreSQL for multi-user

Completed

13. Align parser capabilities across TypeScript, Python, Go, and C++

  • TypeScript: enum extraction (as ClassEntity), arrow function async detection (4384d7d)
  • Python: enum detection from base classes, ABC/Protocol → TraitEntity, visibility from name prefixes
  • Go: parameter extraction (incl. variadic), return type, receiver type, visibility, test detection
  • C++: C++20 coroutine detection via co_await/co_return/co_yield and return type patterns
  • Versions bumped: TS 0.4.2, Python 0.4.3, Go 0.1.6, C++ 0.2.2

12. Missing Calls edge extraction in Rust and Go parsers

  • Rust: added call_expression handling in visitor, including macro invocation bodies (78fad1d, be9a66f)
  • Go: added call extraction (78fad1d)
  • C: was already fixed separately

1. Fix codegraph-c test_parse_syntax_error test

  • Fixed: switched parse_source() from tolerant mode to strict extract(), matching other parsers
  • All 144 codegraph-c tests now pass

2. Add import relationships in codegraph-python builder

  • Fixed: parser_impl.rs already had import mapping; added matching implementation to builder.rs
  • Module nodes created with is_external flag, Imports edges with alias/symbols/wildcard properties

3. Complete async and decorator capture in codegraph-python

  • Fixed: parser_impl.rs stored is_async/is_static as strings instead of bools; now uses proper types
  • Added is_test and attributes (decorators) properties to function nodes
  • Extractor now assigns decorators to func.attributes
  • Test assertions completed for both async detection and decorator extraction

4. Consistent error handling for syntax errors across all parsers

  • Added has_error() check to cpp, csharp, java, kotlin, php, ruby, swift extractors (f34b749)
  • TCL intentionally excluded — vendored grammar produces ERROR nodes for keywords, resolved by sibling-stitching
  • Removed misleading "tolerant parsing" comment from PHP extractor
  • 12 of 13 parsers now enforce strict syntax error checking; TCL documented as exception
  • Added test_syntax_error integration tests to all 7 parsers

5. Dead #[allow(dead_code)] across visitor modules

  • Removed unused config: ParserConfig field from 11 visitors (all except Rust) (9e396b1)
  • Removed debug print_tree/dump_ast helpers + their test callers from Kotlin, Ruby, Tcl
  • Removed dead Python VisitorContext, type_to_string, default_to_string
  • Updated C extract_with_options signature (removed config param) and doctest
  • 3 intentional annotations remain: EdaCommand enum (write-only fields), C Ifdef/Ifndef fields
  • 25 files changed, -249 lines net

Workspace Health

  • Total tests: 1098 passing, 0 failing, 4 ignored
  • Crates: 15 (core + parser-api + 13 language parsers)
  • Clippy: clean
  • Recent features: cyclomatic complexity (all 13 parsers), TypeReference IR, cross-file type resolution