feat(parser): resolve typed member calls with import evidence

[tirth8205]

Summary

• Resolves static class and typed receiver calls only when same-file or repository-local import evidence identifies the target.
• Supports Python, Kotlin, Java, JavaScript, TypeScript, and TSX while preserving function, class, and block scope.
• Keeps uncertain receiver calls as bare CALLS edges instead of deleting or guessing them.
• Records receiver type and resolution evidence on qualified edges.

Reconciliation

This ports the safe intent from #330 and #338 and preserves attribution to Gideon Zenz in the commit trailer.

It deliberately does not port:

• feat(postprocess): fuzzy class-name disambiguation by directory prefix #331 directory-proximity guessing, because proximity is not semantic evidence and ties must remain unresolved.
• feat(parser): filter instance-method CALLS noise #337 the global method-name blocklist, because it deletes legitimate get, find, write, remove, and update calls.
• feat(parser): enrich CALLS via function-reference args and returns #339 function-reference CALLS changes, because current main already models callback values as REFERENCES and the remaining patch leaks class and lexical scope.

The branch is rebased onto current main commit 31feb78 and includes the Julia parser reconciliation from #647. The parser conflict was resolved by preserving the Julia lexical/qualified-call resolver unchanged and applying typed-receiver handling only to its six supported non-Julia languages.

Verification

• 37 focused typed-receiver and Julia reconciliation tests passed.
• 565 parser, language, notebook, Laravel, Julia, and focused receiver tests passed.
• 1 existing process-pool sandbox test was deselected locally; 2 existing tests xpassed.
• Ruff lint passed for production and test changes.
• Ruff format check passed for the new test file.
• Python syntax compilation and git diff checks passed.
• Hosted CI passed lint, review, type-check, schema, security, Python 3.10 through 3.13, GitGuardian, and Windows daemon/file-handle jobs.

A full local run also reproduced the already-tracked macOS iCloud ConfigWatcher native crash in crg-799; hosted CI is the authoritative full-suite run.

[github-actions]

code-review-graph review

Overall risk: 0.40 (MEDIUM) — 29 changed function(s)/class(es), 0 affected flow(s), 19 test gap(s)

Risk-scored changes

Risk	Level	Symbol	Location	Tested
0.40	medium	code_review_graph/parser.py::CodeParser	code_review_graph/parser.py:983	no
0.40	medium	code_review_graph/parser.py::CodeParser._resolve_call_targets	code_review_graph/parser.py:2448	no
0.40	medium	code_review_graph/parser.py::CodeParser.walk	code_review_graph/parser.py:2653	no
0.40	medium	code_review_graph/parser.py::CodeParser._typed_bindings_from_node	code_review_graph/parser.py:2814	no
0.40	medium	code_review_graph/parser.py::CodeParser._store_typed_binding	code_review_graph/parser.py:2898	no
0.40	medium	code_review_graph/parser.py::CodeParser._base_type_name	code_review_graph/parser.py:2909	no
0.40	medium	code_review_graph/parser.py::CodeParser._extract_calls	code_review_graph/parser.py:6022	no
0.40	medium	code_review_graph/parser.py::CodeParser._get_member_call_receiver_method	code_review_graph/parser.py:6201	no
0.40	medium	code_review_graph/parser.py::CodeParser._collect_file_scope	code_review_graph/parser.py:7304	no
0.40	medium	code_review_graph/parser.py::CodeParser._collect_import_names	code_review_graph/parser.py:7638	no

Test gaps

• code_review_graph/parser.py::CodeParser (code_review_graph/parser.py:983)
• code_review_graph/parser.py::CodeParser.parse_bytes (code_review_graph/parser.py:1149)
• code_review_graph/parser.py::CodeParser._resolve_call_targets (code_review_graph/parser.py:2448)
• code_review_graph/parser.py::CodeParser._collect_typed_call_targets (code_review_graph/parser.py:2623)
• code_review_graph/parser.py::CodeParser.walk (code_review_graph/parser.py:2653)
• ...and 14 more without direct tests

Token savings: this graph-backed report used ~93,399 fewer tokens (~94%) than reading every changed file in full (estimated, chars/4 approximation).

---

Powered by code-review-graph — local-first analysis; no code leaves the CI runner.