perf(request): optimize hot-path transformations and add runtime benchmark harness by ndycode · Pull Request #17 · ndycode/codex-multi-auth

ndycode · 2026-02-28T23:10:47Z

Summary

Optimize request-path hot functions with behavior parity:
- ilterInput now uses a single pass for filtering/ID stripping and safely skips sparse entries.
- cleanupToolDefinitions now clones and cleans only parameter schema trees instead of deep-cloning whole tool objects.
Add reusable runtime performance harness:
- scripts/benchmark-runtime-path.mjs

pm run bench:runtime-path

pm run bench:runtime-path:quick

Architect-blocker follow-up:
- Added sparse-array regression guard + test coverage for ilterInput.

Performance Evidence

Baseline and after measurements (same local micro-benchmark shape, Node v25.6.1):

Benchmark	Before avg ms	After avg ms	Delta
ilterInput_1200	0.0821	0.0789	-3.9%
cleanupToolDefinitions_140x25	2.7475	2.2972	-16.4%
�ccountHybridLoop_200x200	0.1676	0.1003	-40.2%

Current reusable harness output (
ode scripts/benchmark-runtime-path.mjs --iterations=60):

ilterInput_small: 0.0259 ms
ilterInput_large: 0.1165 ms
cleanupToolDefinitions_medium: 0.3081 ms
cleanupToolDefinitions_large: 2.3519 ms
�ccountHybridSelection_200: 0.0739 ms

Validation

pm run typecheck ✅

pm run build ✅

Targeted regression suites ✅

px vitest run test/request-transformer.test.ts test/property/transformer.property.test.ts

pm test ⚠️ 1 known pre-existing failure on main:

est/documentation.test.ts > uses scoped package only in explicit legacy migration notes

Risk/Compatibility

No public API changes.
Behavior parity preserved for transformed inputs; sparse-array safety now explicitly covered.

note: greptile review for oc-chatgpt-multi-auth. cite files like `lib/foo.ts:123`. confirm regression tests + windows concurrency/token redaction coverage.

Greptile Summary

reduced request-path allocation overhead by optimizing filterInput (single-pass iteration) and cleanupToolDefinitions (selective cloning). added sparse array guard to prevent crashes on malformed inputs with regression test coverage.

filterInput now uses for-loop instead of filter/map chains, guards against sparse/null entries
cleanupToolDefinitions shallow-spreads tool objects, only deep-clones mutated parameters schema
new runtime benchmark harness validates performance improvements
sparse array regression test added (line 384-394 in request-transformer.test.ts)

windows filesystem/token safety: not applicable - no file i/o or credential handling in hot-path changes. benchmark script only writes json output with standard mkdir/writeFile, no cleanup operations requiring retry logic.

vitest coverage: existing tool-utils tests already cover undefined property cleanup (lines 272-309), property-based tests verify filterInput behavior parity.

Confidence Score: 5/5

safe to merge - performance optimization with behavior parity and comprehensive test coverage
all changes are pure function optimizations with existing test coverage. sparse array guard prevents crashes. no concurrency, filesystem, or token safety risks. benchmark harness enables ongoing performance validation.
no files require special attention

Important Files Changed

Filename	Overview
lib/request/request-transformer.ts	optimized `filterInput` from filter/map chains to single-pass for loop, added sparse array guard to prevent crashes on malformed inputs
lib/request/helpers/tool-utils.ts	optimized `cleanupToolDefinitions` to use shallow spread + selective deep clone instead of full deep clone, added undefined property cleanup to maintain behavior parity
test/request-transformer.test.ts	added sparse array regression test covering the filterInput guard, indentation fix for existing test
scripts/benchmark-runtime-path.mjs	new reusable runtime benchmark harness for hot-path functions, measures filterInput, cleanupToolDefinitions, and account selection performance
package.json	added `bench:runtime-path` and `bench:runtime-path:quick` scripts for runtime performance testing

_{Last reviewed commit: 9cd71c4}

Avoid extra array passes in filterInput and clone only tool schema trees during cleanupToolDefinitions, preserving behavior while reducing hot-path work. Co-authored-by: Codex <noreply@openai.com>

Add a reusable benchmark script and npm commands to measure request-path hotspots (filterInput, cleanupToolDefinitions, and account selection) for before/after performance tracking. Co-authored-by: Codex <noreply@openai.com>

coderabbitai · 2026-02-28T23:10:54Z

Warning

Rate limit exceeded

@ndycode has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 11 minutes and 59 seconds before requesting another review.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

📥 Commits

Reviewing files that changed from the base of the PR and between 36cf5d4 and 9cd71c4.

📒 Files selected for processing (5)

lib/request/helpers/tool-utils.ts
lib/request/request-transformer.ts
package.json
scripts/benchmark-runtime-path.mjs
test/request-transformer.test.ts

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment
Commit unit tests in branch audit/perf-runtime-path-2026-03-01

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Prevent sparse-array holes from causing runtime errors in the single-pass filterInput loop and add regression coverage for sparse entries. Co-authored-by: Codex <noreply@openai.com>

ndycode · 2026-02-28T23:33:49Z

@coderabbitai review

coderabbitai · 2026-02-28T23:33:56Z

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

Source-PR: #17 Source-Branch: audit/perf-runtime-path-2026-03-01 Source-Head: 9cd71c4 Co-authored-by: Codex <noreply@openai.com>

ndycode · 2026-03-01T10:02:09Z

Fixed items

Superseded by PR supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch #27 (eat/unified-supersede-all-prs), which consolidates PR perf(request): optimize hot-path transformations and add runtime benchmark harness #17 ($branch) into the single canonical review branch.

Deferred items

None.

Blocked items

None.

Validation evidence

pm run clean:repo:check - pass

pm run audit:ci - pass

pm run typecheck - pass

pm run lint - pass

pm run build - pass

pm test - pass (92 files, 2113 tests)

PR supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch #27 checks - pass (CodeRabbit, Graphite / AI Reviews, Greptile Review)

supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch

Source-PR: #17 Source-Branch: audit/perf-runtime-path-2026-03-01 Source-Head: 9cd71c4 Co-authored-by: Codex <noreply@openai.com>

supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch

Completes partial fixes flagged in the third review, bumps to v2.2.1. Real bugs (prior fixes were incomplete): - request/fetch-helpers: isUnsupportedCodexModelForChatGpt (the handleErrorResponse path) now also matches NORMALIZED_UNSUPPORTED_MODEL_PATTERN, so a 400 'model not currently available for this chatgpt account' gets the entitlement rewrite, not generic error guidance (#13) - forecast: quota-exhausted accounts are classified 'delayed' (not 'unavailable'), so they slipped the recommendation filter; added an explicit exhausted flag and excluded it, returning null when the whole pool is exhausted (#10) - storage/record-utils: clampIndex guards NaN -> 0 (Math.trunc(NaN) propagated) (#16) - local-client-tokens: debounce lastUsedAt persistence (60s threshold) so the bearer-verify hot path stops writing to disk every request; +chmod 0o700 re-assert on the token-store dir (#12, #11) - mcodex: relay SIGTERM/SIGINT to the spawned child so it isn't orphaned (#15) Test quality: - local-client-tokens: parameterized rename-retry test over ENOTEMPTY/EAGAIN/EACCES (#18) - storage-flagged: clear the H4 deadlock-guard timer on the happy path (#17) - storage: removeWithRetry for suite cleanup (#19) Release: - bump package.json + .codex-plugin/plugin.json to 2.2.1 - add docs/releases/v2.2.1.md; point docs portal + README at it Known follow-ups (documented, deferred — need design, not rushed into a patch): config env-path save is a single-process CAS not a true cross-process lock (#8/#9); verifyLocalClientBearerToken read stays serialized but a fuller lease is future work; runtime proxy routingMutex='enabled' still has a select/commit cursor race (#14) requiring an async refactor of chooseAccount across its call sites. Full suite: 4337 passed, 3 skipped, 0 failed; typecheck + lint clean.

Real bug introduced by the round-3 config lock (#18): - the cross-process config lock fixed expiresAt once and never renewed, and released by unconditionally unlinking the lockfile. A save running longer than CONFIG_LOCK_TTL_MS could be deemed stale and stolen by another process, after which the original holder would delete the NEW owner's lock and reopen concurrent saves. Added a per-acquisition owner token (randomUUID) to the lock payload and a releaseConfigLockIfOwner() that compares-before-unlink, so a holder never deletes a lock it no longer owns. Regressions: stale-foreign-lock takeover (cleans only its own lock) + live-foreign-lock respected (times out, foreign lock untouched, no partial apply). Nits: - mcodex-launcher test: dropped the explicit vitest globals import (#17). - runtime-rotation-proxy: replaced the stale 'KNOWN GAP (L4)' comment on chooseAccount with accurate docs — the race is closed via the reentrant withRoutingMutex on the hot path (#19). Full suite: 4343 passed, 3 skipped, 0 failed; typecheck + lint clean.

ndycode and others added 2 commits March 1, 2026 07:09

perf(request): reduce request-path allocation overhead

b624ae2

Avoid extra array passes in filterInput and clone only tool schema trees during cleanupToolDefinitions, preserving behavior while reducing hot-path work. Co-authored-by: Codex <noreply@openai.com>

chore(bench): add runtime request-path benchmark harness

d1c3596

Add a reusable benchmark script and npm commands to measure request-path hotspots (filterInput, cleanupToolDefinitions, and account selection) for before/after performance tracking. Co-authored-by: Codex <noreply@openai.com>

fix(request): guard sparse inputs in filterInput

9cd71c4

Prevent sparse-array holes from causing runtime errors in the single-pass filterInput loop and add regression coverage for sparse entries. Co-authored-by: Codex <noreply@openai.com>

ndycode marked this pull request as ready for review February 28, 2026 23:26

ndycode added a commit that referenced this pull request Mar 1, 2026

supersede(pr17): import request hot-path performance updates

3f3422e

Source-PR: #17 Source-Branch: audit/perf-runtime-path-2026-03-01 Source-Head: 9cd71c4 Co-authored-by: Codex <noreply@openai.com>

ndycode closed this Mar 1, 2026

ndycode mentioned this pull request Mar 1, 2026

supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch #27

Merged

ndycode added a commit that referenced this pull request Mar 3, 2026

Merge pull request #27 from ndycode/feat/unified-supersede-all-prs

3ebb7bf

supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch

ndycode deleted the audit/perf-runtime-path-2026-03-01 branch March 3, 2026 12:12

ndycode added a commit that referenced this pull request Apr 6, 2026

supersede(pr17): import request hot-path performance updates

bc4c92a

Source-PR: #17 Source-Branch: audit/perf-runtime-path-2026-03-01 Source-Head: 9cd71c4 Co-authored-by: Codex <noreply@openai.com>

ndycode added a commit that referenced this pull request Apr 6, 2026

Merge pull request #27 from ndycode/feat/unified-supersede-all-prs

f426c7b

supersede: unify PRs #15 #16 #17 #18 #19 #20 #24 #25 into one branch

ndycode mentioned this pull request Jun 2, 2026

fix: v2.2.0 deep-audit remediation (4 HIGH, 6 MEDIUM) #505

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

perf(request): optimize hot-path transformations and add runtime benchmark harness#17

perf(request): optimize hot-path transformations and add runtime benchmark harness#17
ndycode wants to merge 3 commits into
mainfrom
audit/perf-runtime-path-2026-03-01

ndycode commented Feb 28, 2026 •

edited by greptile-apps Bot

Loading

Uh oh!

coderabbitai Bot commented Feb 28, 2026 •

edited

Loading

Rate limit exceeded

Uh oh!

ndycode commented Feb 28, 2026

Uh oh!

coderabbitai Bot commented Feb 28, 2026

Uh oh!

ndycode commented Mar 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

ndycode commented Feb 28, 2026 • edited by greptile-apps Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

pm run bench:runtime-path

Performance Evidence

Validation

pm run typecheck ✅

Targeted regression suites ✅

px vitest run test/request-transformer.test.ts test/property/transformer.property.test.ts

Risk/Compatibility

note: greptile review for oc-chatgpt-multi-auth. cite files like lib/foo.ts:123. confirm regression tests + windows concurrency/token redaction coverage.

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Uh oh!

coderabbitai Bot commented Feb 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Rate limit exceeded

Uh oh!

ndycode commented Feb 28, 2026

Uh oh!

coderabbitai Bot commented Feb 28, 2026

Uh oh!

ndycode commented Mar 1, 2026

Fixed items

Deferred items

Blocked items

Validation evidence

pm run clean:repo:check - pass

pm run audit:ci - pass

pm run typecheck - pass

pm run lint - pass

pm run build - pass

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

ndycode commented Feb 28, 2026 •

edited by greptile-apps Bot

Loading

note: greptile review for oc-chatgpt-multi-auth. cite files like `lib/foo.ts:123`. confirm regression tests + windows concurrency/token redaction coverage.

coderabbitai Bot commented Feb 28, 2026 •

edited

Loading