refactor: consolidate Python function discovery to CST path only

[KRRT7]

Summary

• Remove the AST-based Python discovery path (find_functions_with_return_statement, _find_all_functions_in_python_file, function_has_return_statement, function_is_a_property, and associated constants)
• Route Python through the unified _find_all_functions_via_language_support like all other languages
• Fix FunctionVisitor to skip nested functions and exclude @property/@cached_property decorated functions
• Let parse errors propagate from PythonSupport.discover_functions so invalid files correctly return {}

Test plan

• uv run pytest tests/test_function_discovery.py -x -q — 23 passed
• uv run pytest tests/test_async_function_discovery.py -x -q — all passed
• uv run ruff check / ruff format — clean

[claude]

PR Review Summary

Prek Checks

All checks pass (ruff check, ruff format) — no issues found.

Mypy

All mypy errors on changed files are pre-existing (not introduced by this PR). Errors include missing type exports from codeflash.languages.base, missing type params on list/dict generics in support.py, and missing annotations on test functions — all present on main as well.

Code Review

Overall this is a clean refactor that removes ~100 lines of dead AST-based code and unifies Python discovery through the CST path.

Minor issue (non-blocking):

• find_all_functions_in_file (line 493-497) has a stale docstring that still references "routing to either the Python-specific implementation (for backward compatibility) or the language support abstraction for other languages". The Python-specific path was removed in this PR — the docstring should reflect that all languages now go through _find_all_functions_via_language_support.

No critical bugs, security issues, or breaking API changes found.

Key observations:

• Error propagation change is safe: discover_functions no longer catches exceptions, but _find_all_functions_via_language_support still wraps calls in try/except
• get_functions_within_lines uses FunctionVisitor directly with its own error handling — unaffected by the change
• Nested function skipping and @property/@cached_property exclusion logic is correct
• Dict key unification from str to Path is consistent across all modified functions
• Test updates properly reflect the new behavior (nested functions skipped, errors propagated)

Test Coverage

File	PR	Main	Delta
codeflash/discovery/functions_to_optimize.py	69% (506 stmts)	70% (549 stmts)	-1% (43 stmts removed)
codeflash/languages/python/support.py	47% (302 stmts)	49% (311 stmts)	-2% (9 stmts removed)
codeflash/lsp/beta.py	0%	0%	no change (pre-existing, no tests)
Overall	78%	78%	No change

• Coverage percentages are slightly lower due to dead code removal (fewer stmts in denominator), but absolute covered lines remain similar
• No coverage regression — overall stays at 78%
• lsp/beta.py has 0% coverage on both branches (pre-existing)
• 8 test failures in test_tracer.py are pre-existing on main

---

Last updated: 2026-02-27

[KRRT7]

@claude Test results: 2463 passed, 13 failed, 57 skipped. The 13 failures are pre-existing (same failures occur on main). what failures are those?

[KRRT7]

@claude is the JUnit XML parsing logic actually being used for JS / TS?

[codeflash-ai]

⚡️ Codeflash found optimizations for this PR

📄 175% (1.75x) speedup for closest_matching_file_function_name in codeflash/discovery/functions_to_optimize.py

⏱️ Runtime : 24.7 milliseconds → 8.99 milliseconds (best of 111 runs)

A dependent PR with the suggested changes has been created. Please review:

• ⚡️ Speed up function closest_matching_file_function_name by 175% in PR #1689 (consolidate-python-discovery) #1691

If you approve, it will be merged into this PR (branch consolidate-python-discovery).