fix(perf): cache lengths/locals in more hot loops#225
Merged
alexander-akait merged 7 commits intomainfrom Apr 20, 2026
Merged
Conversation
Extends the earlier perf pass (#224) to the remaining hot spots: - HookCodeFactory.contentWithInterceptors: short-circuit the no-interceptor case up front (skip the closure-wrapping and the Object.assign on options); hoist the `interceptors` array and its length to locals so the inner onError/onResult/onDone closures don't re-read them on every iteration. - HookCodeFactory.callTap: cache `this.options.taps` and `this.options.interceptors` once at the top. - HookCodeFactory.callTapsLooping: inline Array#every to drop the per-call callback allocation; cache interceptors length. - HookCodeFactory.needContext: cache taps length. - Hook.intercept: cache `taps`, its length, and `interceptor.register` so the re-registration loop stops doing three property lookups per iteration. - MultiHook.{tap,tapAsync,tapPromise,isUsed,intercept}: cache `hooks.length` so the loop condition doesn't re-read it each pass. Benchmarks (npm run benchmark, --no-opt --predictable): hook-map for(new key) x 10, no interceptors 174k -> 226k ops/s (+30%) hook-map for(new key) x 10, 3 interceptors 157k -> 190k ops/s (+21%) multi-hook intercept across 3 hooks 162k -> 178k ops/s (+10%) multi-hook tap x 10 across 3 hooks 103k -> 109k ops/s (+6%) interceptors-sync call interceptor 6.46k -> 6.93k ops/s (+7%) interceptors-sync register + 10 taps 255k -> 267k ops/s (+5%) All 126 jest tests pass. https://claude.ai/code/session_01863WjdJD1wESVQcdrqgoJL
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #225 +/- ##
==========================================
+ Coverage 98.96% 99.09% +0.12%
==========================================
Files 15 15
Lines 776 775 -1
Branches 130 132 +2
==========================================
Hits 768 768
+ Misses 8 7 -1
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
This reverts commit 0f7d7e4.
Three changes in HookCodeFactory that actually reduce allocations on the compile path (visible on CodSpeed's instruction-count runs, where the previous length-caching tweaks were not): - init(): stop calling `options.args.slice()`. `_args` is only read (length, join, indexing) by this class and its subclasses - never mutated - so sharing the caller's array is safe and saves a copy on every compile. - args(): the `before` / `after` branches used to allocate `[before, ...allArgs]` and/or `[...allArgs, after]`, then re-join them. The no-customization result is already cached as `_joinedArgs`; the customized variants can be built by concatenating that cached string, avoiding both the array spread and the second join. This runs for every tap (async gets `after: "_callback"`, and interceptors with context get `before: "_context"`), so it accumulates. - create(): read `options.type` from the parameter instead of re-reading `this.options.type` (set one line earlier by init()). Generated code output is byte-identical - all 88 HookCodeFactory snapshots still pass - so no hooks library consumers see any change. https://claude.ai/code/session_01863WjdJD1wESVQcdrqgoJL
Three independent allocation reductions:
- HookCodeFactory.setup(): swap `Array.from({length})` for
`new Array(length)`. The former walks the array-like protocol and
writes `undefined` into every slot before the loop immediately
overwrites it; the latter just reserves the backing store.
- HookCodeFactory.contentWithInterceptors(): assign directly to
`options.onError`/`onResult`/`onDone` instead of
`Object.assign(options, {…})`. Avoids allocating the RHS literal
(which is thrown away after the merge) on every compile that has
interceptors.
- Hook._insert(): fast path for single-name `before`. When
`item.before` is `"name"` (or `["name"]`) we now hold the name in
a string and do `tap.name === beforeName` instead of building a
Set just to call `.has()` once per walk step. The existing Set
path is kept for multi-name before.
Behavior unchanged: all 126 jest tests and 88 snapshots still pass,
generated code is byte-identical.
https://claude.ai/code/session_01863WjdJD1wESVQcdrqgoJL
The previous commit added a single-string `before` fast path to
_insert but routed the stage-only path (no `before`, hit by
`tap({stage})` insertions like the `with stages` bench) through the
same loop, which then had to do `beforeName !== undefined` and
`before !== undefined` checks on every iteration. That showed as a
regression on CodSpeed instruction counts.
Fork the walk into two variants:
- No `before`: pure stage walk with zero extra branches - matches
pre-fast-path instruction count on that path.
- Has `before`: keeps the beforeName / Set logic.
Also drop the `before: ["x"]` -> beforeName coercion. Single-element
`before` arrays are rare; removing the special case simplifies the
setup block and the generated hidden class.
All 126 jest tests and 88 snapshots still pass, lint clean.
https://claude.ai/code/session_01863WjdJD1wESVQcdrqgoJL
…ead" This reverts commit 5b2bb30.
This reverts commit 4355011.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
more perf improvements