perf: remove redundant project_root_path.resolve() from hot path by KRRT7 · Pull Request #1541 · codeflash-ai/codeflash

KRRT7 · 2026-02-19T03:48:11Z

Summary

Resolve project_root_path and tests_project_rootdir once at construction time (TestConfig.__post_init__) instead of on every call to module_name_from_file_path
Remove the redundant .resolve() syscall from module_name_from_file_path — eliminating ~776 filesystem syscalls from the hot path
Resolve early in FunctionOptimizer.__init__ and filter_functions entry to ensure consistency with symlink-canonical paths

Test plan

pytest tests/test_code_utils.py -k module_name_from_file_path — 4 pass
pytest tests/test_async_function_discovery.py::test_get_functions_to_optimize_with_async — pass
pytest tests/test_formatter.py::test_formatting_file_with_many_diffs — pass
pytest tests/test_function_discovery.py::test_filter_functions_tests_root_overlaps_source — pass

StdlibIntegration monkeypatches subprocess.Popen, adding span/breadcrumb overhead to every subprocess call. Profiling showed ~342k samples (~8.9% of total runtime) in sentry_sdk.integrations.stdlib and sentry_sdk.utils. Disable subprocess instrumentation and set traces/profiles sample rates to 0 since codeflash is a CLI tool that doesn't need performance tracing. Error capturing (capture_exception, capture_message, LoggingIntegration) remains fully functional.

…arse_test_output Replace lazy `.*?` quantifiers in matches_re_start/matches_re_end with negated character classes (`[^:]`, `[^#]`, `[^.:]`) to eliminate quadratic backtracking. Replace per-line regex search for the pytest FAILURES header with a simple `"= FAILURES =" in line` string check. Add tests for the regex patterns and failure header detection.

Resolve project root paths once at construction time (TestConfig.__post_init__, FunctionOptimizer.__init__, filter_functions entry) instead of on every call to module_name_from_file_path — eliminating ~776 redundant filesystem syscalls.

StdlibIntegration in sentry-sdk 2.x doesn't accept subprocess_instrumentation as a parameter. Use disabled_integrations instead, which also avoids httplib patching overhead.

codeflash/telemetry/sentry.py

Drop .resolve() from ImportResolver, TestsCache, init_javascript, create_pr, and filter_functions where callers already pass resolved paths via CLI init or TestConfig.__post_init__.

…/codeflash-ai/codeflash into narrow-sentry-stdlib-integration

codeflash/code_utils/code_utils.py

Drop .resolve() from ImportResolver, TestsCache, init_javascript, create_pr, and filter_functions where callers already pass resolved paths via CLI init or TestConfig.__post_init__. Also exclude test and fixture dirs from mypy to match ruff/ty config.

claude · 2026-02-19T04:10:58Z

PR Review Summary

Prek Checks

✅ Passed — Prek found 1 unused import (import os in code_context_extractor.py), already auto-fixed in commit 5efa1eeb.

Mypy

⚠️ Pre-existing issues only — No new mypy errors introduced by this PR. The disabled_integrations=[StdlibIntegration] type mismatch in sentry.py is a sentry SDK type-stub limitation (runtime behavior is correct).

Code Review

✅ No critical issues found in this re-review.

Changes reviewed:

TestConfig.__post_init__: Resolves project_root_path and tests_project_rootdir once at construction — correct, dataclass is not frozen
sentry.py: Disables StdlibIntegration, sets traces/profiles sample rate to 0 — reduces unnecessary overhead from Sentry's stdlib patching
parse_test_output.py: Regex refactored from greedy (.*?) to precise ([^:]*) patterns with documentation; FAILURES header detection simplified from regex to string in check — both correct
code_context_extractor.py: is_project_path now uses resolve().relative_to() instead of fragile str.startswith(str + os.sep); definition_path guarded against None from Jedi — correct
function_optimizer.py: self.project_root resolved once in __init__; path.resolve() in reformat_code_and_helpers ensures consistency — correct
create_pr.py: Removed redundant .resolve() since tests_project_rootdir is now resolved at construction — correct
New test file: tests/test_parse_test_output_regex.py (189 lines) — comprehensive coverage of the regex refactoring

All 3 existing inline review comments already resolved (✅ Fixed).

Test Coverage

File	Main	PR	Delta
`cli_cmds/init_javascript.py`	41%	41%	—
`code_utils/instrument_existing_tests.py`	92%	92%	—
`discovery/discover_unit_tests.py`	74%	74%	—
`discovery/functions_to_optimize.py`	70%	70%	—
`languages/javascript/import_resolver.py`	72%	72%	—
`languages/python/context/code_context_extractor.py`	91%	92%	+1%
`optimization/function_optimizer.py`	18%	18%	—
`result/create_pr.py`	67%	67%	—
`telemetry/sentry.py`	—	—	—
`verification/parse_test_output.py`	58%	58%	—
`verification/verification_utils.py`	60%	61%	+1%
`tests/test_parse_test_output_regex.py` (new)	—	99%	✅

Overall coverage: 78.59% → 78.61% (+0.02pp) — No regression.

New file tests/test_parse_test_output_regex.py has 99% coverage ✅

Note: 8 pre-existing test failures on both branches (7 in test_tracer.py, 1 in test_code_context_extractor.py — the latter is fixed on the remote by commit a726b464 but wasn't in local checkout).

Last updated: 2026-02-19

codeflash/verification/verification_utils.py

Paths with .. segments (e.g. Path(__file__) / "../foo") don't match resolved file_path in module_name_from_file_path. Resolve once at the public entry point and pass down to the async variant. Also exclude test dirs from mypy config.

…istency On Windows, Path.resolve() can return 8.3 short names (e.g. RUNNER~1) that differ from long-name forms. Jedi returns resolved paths that may not match the project root's form, causing relative_to() failures. - Normalize Jedi definition paths through the project root via safe_relative_to to ensure consistent dict keys - Resolve both sides in is_project_path and module_name_from_file_path - Restore resolve in TestsCache for consistent cache keys

KRRT7 · 2026-02-19T05:04:46Z

@claude fix the windows unit tests, fix the root cause

…istency TestConfig.__post_init__ resolves project_root_path to long names, but FunctionToOptimize.file_path could retain Windows 8.3 short names (e.g., RUNNER~1 vs runneradmin), causing relative_to() failures. Co-authored-by: Kevin Turcios <KRRT7@users.noreply.github.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

KRRT7 · 2026-02-19T05:12:17Z

@claude fix the unit tests failures and push

…olution FunctionToOptimize is a frozen Pydantic dataclass, so directly assigning to file_path raises FrozenInstanceError. Use dataclasses.replace() to create a new instance with the resolved path instead. Co-authored-by: Kevin Turcios <KRRT7@users.noreply.github.com>

KRRT7 · 2026-02-19T05:22:39Z

@claude fix the failing unit tests

claude · 2026-02-19T05:23:17Z

Claude finished @KRRT7's task in 7m 1s —— View job

Fixing Failing Unit Tests

Identify failing tests — 9 failures in tests/test_unused_helper_revert.py (Windows only)
Analyze root causes
Fix the failing tests
Push changes

Root cause: Path mismatch in revert_unused_helper_functions (unused_definition_remover.py:597).

On Windows, tempfile.mkdtemp() returns paths with 8.3 short names (e.g., C:\Users\RUNNER~1\...). The original_helper_code dict is keyed with these unresolved paths, but unused_helpers[*].file_path (from Jedi discovery) are resolved to long-form paths. The dict lookup file_path in original_helper_code fails silently, so the unused helpers are never reverted.

Fix (commit 8f65c9f7): Resolve both sides of the path comparison in revert_unused_helper_functions:

Resolve original_helper_code keys via {p.resolve(): code for p, code in original_helper_code.items()}
Resolve helper.file_path when grouping by file

This ensures consistent path comparison regardless of Windows 8.3 name transformations.

|

…name consistency On Windows, tempfile.mkdtemp() returns paths with 8.3 short names while Jedi resolves to long-form paths. The dict lookup for original_helper_code failed silently, causing unused helpers to never be reverted. Co-authored-by: Kevin Turcios <KRRT7@users.noreply.github.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ration perf: remove redundant project_root_path.resolve() from hot path

KRRT7 and others added 5 commits February 18, 2026 21:57

style: auto-fix ruff formatting in parse_test_output.py

377083a

fix: use disabled_integrations to exclude StdlibIntegration

486f6b8

StdlibIntegration in sentry-sdk 2.x doesn't accept subprocess_instrumentation as a parameter. Use disabled_integrations instead, which also avoids httplib patching overhead.

claude bot reviewed Feb 19, 2026

View reviewed changes

codeflash/telemetry/sentry.py Show resolved Hide resolved

github-actions bot and others added 3 commits February 19, 2026 03:56

style: add return type annotations to test methods

8995279

perf: remove remaining redundant .resolve() calls on pre-resolved paths

4742585

Drop .resolve() from ImportResolver, TestsCache, init_javascript, create_pr, and filter_functions where callers already pass resolved paths via CLI init or TestConfig.__post_init__.

Merge branch 'narrow-sentry-stdlib-integration' of https://github.com…

27937d4

…/codeflash-ai/codeflash into narrow-sentry-stdlib-integration

claude bot reviewed Feb 19, 2026

View reviewed changes

codeflash/code_utils/code_utils.py Outdated Show resolved Hide resolved

KRRT7 and others added 2 commits February 18, 2026 23:05

style: auto-fix line length formatting in functions_to_optimize.py

ac096d9

claude bot reviewed Feb 19, 2026

View reviewed changes

codeflash/verification/verification_utils.py Show resolved Hide resolved

KRRT7 and others added 4 commits February 18, 2026 23:14

fix: guard against None definition_path from Jedi

a726b46

style: remove unused os import in code_context_extractor

5efa1ee

KRRT7 merged commit 401d24c into main Feb 19, 2026
27 of 28 checks passed

KRRT7 deleted the narrow-sentry-stdlib-integration branch February 19, 2026 05:59

KRRT7 added a commit that referenced this pull request Feb 19, 2026

Merge pull request #1541 from codeflash-ai/narrow-sentry-stdlib-integ…

7e2d8c0

…ration perf: remove redundant project_root_path.resolve() from hot path

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comments

perf: remove redundant project_root_path.resolve() from hot path#1541

perf: remove redundant project_root_path.resolve() from hot path#1541
KRRT7 merged 17 commits intomainfrom
narrow-sentry-stdlib-integration

KRRT7 commented Feb 19, 2026

Uh oh!

Uh oh!

Uh oh!

claude bot commented Feb 19, 2026 •

edited

Loading

Uh oh!

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

claude bot commented Feb 19, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Comments

Conversation

KRRT7 commented Feb 19, 2026

Summary

Test plan

Uh oh!

Uh oh!

Uh oh!

claude bot commented Feb 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Review Summary

Prek Checks

Mypy

Code Review

Test Coverage

Uh oh!

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

KRRT7 commented Feb 19, 2026

Uh oh!

claude bot commented Feb 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Fixing Failing Unit Tests

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

claude bot commented Feb 19, 2026 •

edited

Loading

claude bot commented Feb 19, 2026 •

edited

Loading