feat(minifier): drop write-only property assignments to unused local bindings by default#24112
Merged
Merged
Conversation
Merging this PR will not alter performance
Comparing Footnotes
|
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 3, 2026 09:20
035a3cb to
ae445d3
Compare
Dunqing
changed the base branch from
graphite-base/24112
to
minifier-guard-inversion
July 3, 2026 09:20
Member
Author
How to use the Graphite Merge QueueAdd either label to this PR to merge it via the merge queue:
You must have a Graphite account in order to use the merge queue. Sign up using this link. An organization admin has enabled the Graphite Merge Queue in this repository. Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue. This stack of pull requests is managed by Graphite. Learn more about stacking. |
Contributor
Monitor OxcCommit:
|
Dunqing
force-pushed
the
minifier-guard-inversion
branch
from
July 3, 2026 09:56
abad106 to
caca822
Compare
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 3, 2026 09:56
ae445d3 to
9621c74
Compare
Dunqing
force-pushed
the
minifier-guard-inversion
branch
from
July 6, 2026 07:36
caca822 to
a77d6c4
Compare
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 6, 2026 07:36
9621c74 to
45a99cb
Compare
graphite-app
Bot
force-pushed
the
minifier-guard-inversion
branch
2 times, most recently
from
July 6, 2026 08:31
9e679b3 to
f39d74e
Compare
graphite-app
Bot
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 6, 2026 08:31
45a99cb to
d7c31b4
Compare
Dunqing
changed the base branch from
minifier-guard-inversion
to
graphite-base/24112
July 6, 2026 08:49
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 6, 2026 08:49
d7c31b4 to
d59bcc1
Compare
Dunqing
force-pushed
the
graphite-base/24112
branch
from
July 6, 2026 08:49
f39d74e to
2d9b0b3
Compare
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
3 times, most recently
from
July 7, 2026 03:06
3e1637a to
e5478f8
Compare
Dunqing
force-pushed
the
minifier-write-only-prop-dce
branch
3 times, most recently
from
July 7, 2026 14:58
b1c8859 to
144288b
Compare
Dunqing
marked this pull request as ready for review
July 8, 2026 02:25
Contributor
Merge activity
|
…bindings by default (#24112) ## What Drop `o.x = value` when `o` is a local variable that is never read — by default. Until now this only happened with the `treeshake.property_write_side_effects: false` opt-in. terser does this by default; esbuild does not do it at all. This matters because one leftover write keeps a lot of code alive. antd bundles 820 icon modules that each end with `o.displayName = 'UserOutlined'`; that single write kept each module's whole require-chain alive. **antd.js: 2.22 MB -> 2.13 MB minified** (458 kB -> 455 kB gzip). oxc now beats esbuild on antd by ~180 kB. ## How The drop itself uses machinery that already exists for the opt-in path (`is_member_assign_to_unused_binding`: the object must be a fresh value — function/class/object/array literal — not exported, never written as a whole, and every read of it must be the target of a property write). What is new is making it safe without the opt-in: - A program-wide fact table (`PersistentSymbolFacts`, bit `SymbolFact::MEMBER_WRITE_HAZARD`) is filled once by `Normalize`, before the compress loop. The container is generic on purpose: monotone symbol facts, seeded pre-loop, never cleared — a stale set bit only forgoes an optimization. A follow-up migrates `proto_write_symbols` into it as `PROTO_WRITTEN`, fixing that field's execution-order TODO. A symbol goes in when any of its property writes could be observed: read-modify ops (`o.x += 1` reads the property), chained writes (`a.b.c = 1` reads `a.b`, so dropping `a.b = {}` would throw), and `__proto__` or computed keys (may install setters). A hazarded symbol never gets its writes dropped. - The set also grows mid-loop when a pass rewrites `o.x = o.x + 1` into `o.x += 1`. - Only plain `=` with a single-level, non-`__proto__` key qualifies. `o[k()] = v` bails. - A side-effectful right side is kept: `o.x = impure()` becomes `impure()`. Once the write is gone, the variable is unused and the existing passes delete the declaration and everything only it referenced. ## Behavior change The only assumption this makes (documented in `docs/ASSUMPTIONS.md`): setters installed on `Object.prototype` / `Function.prototype` / `Array.prototype` are not observed by these writes — the same assumption the existing opt-in flag makes, and terser's default. A 98-case audit (node ground truth vs oxc vs terser, strict and sloppy) hardened everything else. Writes that throw or have exotic effects are now kept even though the binding is unused, via a kind-aware key denylist: `caller`/`arguments`/`name`/`length` on function- and class-valued bindings, `prototype` on class-valued (class prototypes are non-writable; plain function `prototype` writes still drop — they are writable and a real size win), `length` on array-valued (the array length setter coerces its RHS and can throw RangeError), private-field writes (brand-check TypeError), and classes with static blocks, decorators, or static accessors are not treated as fresh. Post-audit the matrix shows zero unsound cases beyond the documented prototype-setter class — stricter than terser, which for example still drops `a.length = -1` (losing the RangeError) where oxc now keeps it. No size cost: the minsize corpus is byte-identical, antd keeps the full 90 kB win. ## Example ```js // input (function() { var r = require('react'); var i = require('./asn/UserOutlined'); var o = function(e, t) { return r.createElement(i.a, e, t); }; o.displayName = 'UserOutlined'; })(); ``` ```js // before (everything kept) (function(){var e=require("react"),t=require("./asn/UserOutlined"),n=function(n,r){return e.createElement(t.a,n,r)};n.displayName=`UserOutlined`})(); // after (only the side effects remain) (function(){require("react"),require("./asn/UserOutlined")})(); ``` Found by re-minifying oxc's minsize output with `terser -c`; this was the single largest gap between oxc and terser. Full conformance suite passes with zero snapshot changes; minsize stays a fixed point (antd takes one extra compress iteration). Implemented with Claude Code, reviewed and verified by me.
graphite-app
Bot
force-pushed
the
minifier-write-only-prop-dce
branch
from
July 8, 2026 02:28
144288b to
34ff7b4
Compare
graphite-app Bot
pushed a commit
that referenced
this pull request
Jul 8, 2026
…24292) ## Issue The `Allocations` CI job can fail on PRs with a 1-allocation `Sys allocs` difference that local `cargo allocs` cannot reproduce. #22621 fixed one instance of this in the transformer stage. #24112 and #24017 hit the same failure again in the minifier stage: `kitchen-sink.tsx` measures 2309 Sys allocs on macOS/arm64 but 2310 on Linux/x64, so whichever platform regenerates the snapshot last breaks CI for the other. ## Why The tracker counts every call to the system allocator. That includes the chunks bump arenas request. Whether an arena needs one more chunk depends on the total number of bytes in the arena, and byte totals are not the same on every target. For example, hashbrown tables use 16-byte control groups on x86_64 but 8-byte groups on aarch64. When an arena's content size sits close to a chunk boundary, one target crosses it and the other does not. In the observed failure, the minifier's mangler-phase semantic build creates a fresh `Scoping`. Its arena holds symbol names, resolved-reference vectors, and the per-scope `bindings` hash tables, which are slightly larger on x86_64. The traced Linux-only allocation happened under `Scoping::add_resolved_reference -> ArenaVec::push -> Arena::new_chunk`, when the push overflowed the current chunk and requested a new 720880-byte chunk. On aarch64 the same content fit in the existing chunk, so the count stayed one lower. Every other allocation class in the measurement grows on element counts, which are identical across platforms. Logging every allocation size during the minifier stage on macOS/arm64, Linux/arm64, and Linux/x64 showed the three sequences match one-to-one except for arena chunk requests. Any PR that shifts arena content near a chunk boundary can trip this again in any stage, so fixing one stage at a time does not close the class. ## Fix Count chunk allocations in `oxc_allocator` under the tracker-only `track_allocations` feature, and subtract them from the reported Sys allocs in the tracker. The counter is global rather than per-arena because short-lived arenas (like the one backing `Scoping`) are created and dropped inside the measured operation. Arena usage is still measured by the `Arena allocs` / `Arena reallocs` columns; only the platform-sensitive chunk-request count is excluded from `Sys allocs`. Parser and transformer rows are unchanged (the main arena is pre-warmed, and #22621 already removed transformer scoping growth). Semantic, minifier, and formatter rows drop by their per-stage chunk counts. ## Validation `cargo allocs` produces byte-identical snapshots on macOS/arm64, Linux/arm64 (Docker), and Linux/x64 (Docker), and is idempotent across reruns. `cargo test -p oxc_allocator` passes; clippy is clean with `track_allocations` both on and off, and the counter compiles out entirely without the feature. AI usage: AI assistance was used to investigate, implement, and validate this change.
Boshen
added a commit
that referenced
this pull request
Jul 14, 2026
### 🚀 Features - 616bfa2 minifier: Remove unreachable code after terminating statements (#24441) (Dunqing) - ddab89a data_structures: Add `likely` and `unlikely` functions (#24368) (overlookmotel) - a3a39f9 react_compiler: Implement enableEmitHookGuards codegen (#24329) (Boshen) - b79eef7 minifier: Apply De Morgan's law to negated comparison chains in jump guards and loop tests (#24279) (Dunqing) - 34ff7b4 minifier: Drop write-only property assignments to unused local bindings by default (#24112) (Dunqing) - 1b829d8 semantic: Record const enums in EnumData (#24268) (Dunqing) - ba0944c semantic: Add `Scoping::set_symbol_span` (#24221) (camc314) ### 🐛 Bug Fixes - 7d33363 minifier: Preserve guaranteed throws from class heritage evaluation (#24349) (Dunqing) - 058a62f semantic: Track ambient contexts in `SemanticBuilder` (#24327) (camc314) - 721eb0b transformer/decorator: Scope accessor class binding (#24330) (camc314) - 1ebdce3 semantic: Allow reserved keywords in ambient declaration types (#24325) (camc314) - 460176a track-memory-allocations: Exclude arena chunks from Sys allocs (#24292) (Dunqing) - af4922b transformer: Clear lowered namespace redeclarations (#24300) (camc314) - ffd2765 semantic: Mark declared computed `MethodDefinition`s as type references (#24296) (camc314) - f17514b isolated-declarations: Emit const readonly fields as types (#24288) (camc314) - 40f769d minifier: Make `__proto__` write tracking execution-order independent (#24280) (Dunqing) - 6371fed transformer: Remove stale enum member bindings (#24272) (camc314) - f05dfab transformer: Correct symbol flags for lowered namespaces (#24271) (Dunqing) - 84eeb55 transformer: Correct symbol flags for lowered enums (#24269) (Dunqing) - c3057da transformer: Preserve generated class binding spans (#24220) (camc314) - 8260096 transformer: Correct span for lowered namespace symbol (#24222) (camc314) - 42d00d3 semantic: Mark declared class heritage as type references (#24237) (camc314) - 588d997 semantic: Mark TS `PropertyDefinition`s computed fields as type references (#24233) (camc314) - 9b95632 semantic: Mark computed method keys in `TSMethodSignature`s as type references (#24232) (camc314) ### ⚡ Performance - 5b26643 transformer_plugins: Dispatch global defines by trailing name (#23666) (Boshen) - dce0f29 react_compiler: Replace all compiled functions in a single AST walk (#24403) (Boshen) - f85f0d8 ast: Delegate inherited enum variants in clone_in and estree derives (#23555) (Boshen) - 3ff0234 allocator: Remove `unwrap` from `ReplaceWith` (#24365) (overlookmotel) - ab22e80 transformer: Fix Rust 1.97 performance regression (#24354) (camc314) - b47585c parser: Use `ReplaceWith` instead of `TakeIn` (#24018) (overlookmotel) - b227a06 minifier: Use `ReplaceWith` instead of `TakeIn` (#24017) (overlookmotel) Co-authored-by: Boshen <[email protected]>
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
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.

What
Drop
o.x = valuewhenois a local variable that is never read — by default. Until now this only happened with thetreeshake.property_write_side_effects: falseopt-in. terser does this by default; esbuild does not do it at all.This matters because one leftover write keeps a lot of code alive. antd bundles 820 icon modules that each end with
o.displayName = 'UserOutlined'; that single write kept each module's whole require-chain alive. antd.js: 2.22 MB -> 2.13 MB minified (458 kB -> 455 kB gzip). oxc now beats esbuild on antd by ~180 kB.How
The drop itself uses machinery that already exists for the opt-in path (
is_member_assign_to_unused_binding: the object must be a fresh value — function/class/object/array literal — not exported, never written as a whole, and every read of it must be the target of a property write).What is new is making it safe without the opt-in:
PersistentSymbolFacts, bitSymbolFact::MEMBER_WRITE_HAZARD) is filled once byNormalize, before the compress loop. The container is generic on purpose: monotone symbol facts, seeded pre-loop, never cleared — a stale set bit only forgoes an optimization. A follow-up migratesproto_write_symbolsinto it asPROTO_WRITTEN, fixing that field's execution-order TODO. A symbol goes in when any of its property writes could be observed: read-modify ops (o.x += 1reads the property), chained writes (a.b.c = 1readsa.b, so droppinga.b = {}would throw), and__proto__or computed keys (may install setters). A hazarded symbol never gets its writes dropped.o.x = o.x + 1intoo.x += 1.=with a single-level, non-__proto__key qualifies.o[k()] = vbails.o.x = impure()becomesimpure().Once the write is gone, the variable is unused and the existing passes delete the declaration and everything only it referenced.
Behavior change
The only assumption this makes (documented in
docs/ASSUMPTIONS.md): setters installed onObject.prototype/Function.prototype/Array.prototypeare not observed by these writes — the same assumption the existing opt-in flag makes, and terser's default.A 98-case audit (node ground truth vs oxc vs terser, strict and sloppy) hardened everything else. Writes that throw or have exotic effects are now kept even though the binding is unused, via a kind-aware key denylist:
caller/arguments/name/lengthon function- and class-valued bindings,prototypeon class-valued (class prototypes are non-writable; plain functionprototypewrites still drop — they are writable and a real size win),lengthon array-valued (the array length setter coerces its RHS and can throw RangeError), private-field writes (brand-check TypeError), and classes with static blocks, decorators, or static accessors are not treated as fresh. Post-audit the matrix shows zero unsound cases beyond the documented prototype-setter class — stricter than terser, which for example still dropsa.length = -1(losing the RangeError) where oxc now keeps it. No size cost: the minsize corpus is byte-identical, antd keeps the full 90 kB win.Example
Found by re-minifying oxc's minsize output with
terser -c; this was the single largest gap between oxc and terser. Full conformance suite passes with zero snapshot changes; minsize stays a fixed point (antd takes one extra compress iteration).Implemented with Claude Code, reviewed and verified by me.