LSP reduce no of candidates

[Saga4]

PR Type

Enhancement

---

Description

• Add LSP-aware dynamic config getters

• Reduce LSP candidates, tests, runtime

• Replace constants with getters across code

• Improve profiling step logging

---

Diagram Walkthrough

flowchart LR
  A["config_consts: LSP constants + getters"] -- "get_n_candidates/get_n_tests/get_total_time" --> B["git_utils: worktree count uses getter"]
  A -- "dynamic values" --> C["function_optimizer: tests, candidates, timing"]
  A -- "dynamic values" --> D["test_runner: runtime params via getter"]
  C -- "profiling UX" --> E["Add info log + console rule"]

Loading

File Walkthrough

Relevant files

Enhancement

config_consts.py Add LSP-specific constants and dynamic getters                      codeflash/code_utils/config_consts.py Add LSP-specific constants. Introduce getters for candidates, tests, time. Route values via LSP enabled check. +23/-0    git_utils.py Use dynamic candidate count for worktrees                                codeflash/code_utils/git_utils.py Import dynamic candidate getter. Use getter to size worktrees. +2/-2      function_optimizer.py Dynamic tests/candidates/time; better profiling logs          codeflash/optimization/function_optimizer.py Replace constants with getters for tests/time/candidates. Adjust executor and assertions to dynamic tests. Pass dynamic candidate count to optimizers. Improve line profiler logging and rule. +28/-19  test_runner.py Test runners use dynamic total runtime                                      codeflash/verification/test_runner.py Swap fixed runtime with dynamic getter. Apply to behavioral, profiling, benchmarking runners. +4/-4

---

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit 65a5b6a)

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Import Overhead Repeated intra-function imports of is_LSP_enabled in small getters may add overhead on hot paths; consider caching the flag or the computed values to avoid repeated checks. def get_n_candidates() -> int: from codeflash.lsp.helpers import is_LSP_enabled return N_CANDIDATES_LSP if is_LSP_enabled() else N_CANDIDATES def get_n_tests_to_generate() -> int: from codeflash.lsp.helpers import is_LSP_enabled return N_TESTS_TO_GENERATE_LSP if is_LSP_enabled() else N_TESTS_TO_GENERATE def get_total_looping_time() -> float: from codeflash.lsp.helpers import is_LSP_enabled return TOTAL_LOOPING_TIME_LSP if is_LSP_enabled() else TOTAL_LOOPING_TIME Default Arg Eval Using get_total_looping_time() as a function default value will bind the value at import time; if LSP mode can change at runtime, the default may become stale. Pass None and compute inside, or set at call sites. def run_and_parse_tests( self, testing_type: TestingMode, test_env: dict[str, str], test_files: TestFiles, optimization_iteration: int, testing_time: float = get_total_looping_time(), *, Dynamic Defaults Functions now default pytest_target_runtime_seconds to get_total_looping_time(), evaluated at import; verify this is intended and that LSP toggling after import is not required. def run_behavioral_tests( test_paths: TestFiles, test_framework: str, test_env: dict[str, str], cwd: Path, *, pytest_timeout: int | None = None, pytest_cmd: str = "pytest", verbose: bool = False, pytest_target_runtime_seconds: int = get_total_looping_time(), enable_coverage: bool = False, ) -> tuple[Path, subprocess.CompletedProcess, Path | None, Path | None]: if test_framework == "pytest": test_files: list[str] = []

[github-actions]

PR Code Suggestions ✨

Latest suggestions up to 65a5b6a
Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Make LSP getter import-safe Importing inside these getters can raise import errors at runtime in non-LSP environments or during early import cycles. Wrap the conditional in a try/except to fall back safely when LSP helpers are unavailable. This prevents hard crashes and preserves default behavior. codeflash/code_utils/config_consts.py [21-25] def get_n_candidates() -> int: - from codeflash.lsp.helpers import is_LSP_enabled + try: + from codeflash.lsp.helpers import is_LSP_enabled # local import to avoid hard dependency + return N_CANDIDATES_LSP if is_LSP_enabled() else N_CANDIDATES + except Exception: + # On any import/runtime issue, default to non-LSP values + return N_CANDIDATES - return N_CANDIDATES_LSP if is_LSP_enabled() else N_CANDIDATES - Suggestion importance[1-10]: 7 __ Why: The existing code does a local import that could fail in environments without LSP; adding a try/except improves robustness without changing behavior. Impact is moderate since it guards against runtime import errors but is not a critical bug fix.	Medium
	Guard test count getter Protect the lazy import and LSP check with a try/except to avoid failures when the LSP module is absent or partially initialized. Default to standard values on any exception. codeflash/code_utils/config_consts.py [27-31] def get_n_tests_to_generate() -> int: - from codeflash.lsp.helpers import is_LSP_enabled + try: + from codeflash.lsp.helpers import is_LSP_enabled + return N_TESTS_TO_GENERATE_LSP if is_LSP_enabled() else N_TESTS_TO_GENERATE + except Exception: + return N_TESTS_TO_GENERATE - return N_TESTS_TO_GENERATE_LSP if is_LSP_enabled() else N_TESTS_TO_GENERATE - Suggestion importance[1-10]: 7 __ Why: Similar to the first, wrapping the lazy import prevents hard failures when LSP helpers are unavailable, defaulting safely to standard values. It’s accurate and improves resilience with moderate impact.	Medium
	Harden looping time getter Ensure resilience against missing LSP dependencies by catching exceptions around the import and check. Return the default timing when LSP is unavailable to prevent runtime errors during test runs. codeflash/code_utils/config_consts.py [33-37] def get_total_looping_time() -> float: - from codeflash.lsp.helpers import is_LSP_enabled + try: + from codeflash.lsp.helpers import is_LSP_enabled + return TOTAL_LOOPING_TIME_LSP if is_LSP_enabled() else TOTAL_LOOPING_TIME + except Exception: + return TOTAL_LOOPING_TIME - return TOTAL_LOOPING_TIME_LSP if is_LSP_enabled() else TOTAL_LOOPING_TIME - Suggestion importance[1-10]: 7 __ Why: Catching exceptions around the LSP import/check avoids crashes during timing retrieval, defaulting to the non-LSP constant. Correct and useful for stability, but not a critical fix.	Medium

---

Previous suggestions

Suggestions up to commit 65a5b6a

Category	Suggestion	Impact
General	Use cached LSP flag for tests Reuse a cached LSP-enabled result to prevent repeated dynamic imports and checks, improving performance and avoiding import-time side effects. Keep logic consistent across all getters. codeflash/code_utils/config_consts.py [27-31] def get_n_tests_to_generate() -> int: - from codeflash.lsp.helpers import is_LSP_enabled + return N_TESTS_TO_GENERATE_LSP if _is_lsp_enabled_cached() else N_TESTS_TO_GENERATE - return N_TESTS_TO_GENERATE_LSP if is_LSP_enabled() else N_TESTS_TO_GENERATE - Suggestion importance[1-10]: 6 __ Why: This aligns with the first suggestion’s caching goal and correctly targets the matching lines; modest performance gain and consistency, but dependent on introducing the cached helper safely.	Low
	Cache LSP check for timing Apply the same cached LSP check here to avoid repeated imports and ensure consistent behavior. This prevents unnecessary overhead on frequently called timing paths. codeflash/code_utils/config_consts.py [33-36] def get_total_looping_time() -> float: - from codeflash.lsp.helpers import is_LSP_enabled + return TOTAL_LOOPING_TIME_LSP if _is_lsp_enabled_cached() else TOTAL_LOOPING_TIME - return TOTAL_LOOPING_TIME_LSP if is_LSP_enabled() else TOTAL_LOOPING_TIME - Suggestion importance[1-10]: 6 __ Why: Matches the new hunk and would reduce repeated imports on a hot path; impact is moderate and contingent on the cached helper being added without altering expected dynamic behavior.	Low
	Cache LSP-enabled check Cache the LSP-enabled check to avoid repeated imports and function calls on hot paths. This reduces overhead when these getters are called frequently during optimization. codeflash/code_utils/config_consts.py [21-25] +from functools import cache + +@cache +def _is_lsp_enabled_cached() -> bool: + from codeflash.lsp.helpers import is_LSP_enabled + return is_LSP_enabled() + def get_n_candidates() -> int: - from codeflash.lsp.helpers import is_LSP_enabled + return N_CANDIDATES_LSP if _is_lsp_enabled_cached() else N_CANDIDATES - return N_CANDIDATES_LSP if is_LSP_enabled() else N_CANDIDATES - Suggestion importance[1-10]: 5 __ Why: The existing code matches the new hunk and the proposal can reduce repeated imports/calls, but it introduces global caching that may not reflect runtime changes to LSP state and adds a new helper that isn't shown integrated across the file.	Low

[github-actions]

Persistent review updated to latest commit 65a5b6a

[misrasaurabh1]

we should not decrease this, as this can increase performance noise

[misrasaurabh1]

we should also enable line profiling but reduce the number of candidates for it

[Saga4]

we should also enable line profiling but reduce the number of candidates for it

We never disabled it for VSC, the base profiling and optimization candidate profiling all works same as cli. Yeah we can check on reduction of count though.

[misrasaurabh1]

lets keep it 2

[misrasaurabh1]

lets move the line profiler candidate count to cli

[Saga4]

As discussed with @mohammedahmed18 N_candidates are hardcoded in aiservices currently. so need to make it parameterised with max value.

    optimization_response_items, llm_cost = await optimize_python_code(
        request.user, ctx, data.dependency_code, n=5, python_version=python_version
    )
    ```

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ mohammedahmed18
❌ saga4

---

saga4 seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}