Skip to content

[compiler] Patch array and argument spread mutability #32521

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 13, 2025
Merged

[compiler] Patch array and argument spread mutability #32521

merged 1 commit into from
Mar 13, 2025

Conversation

mofeiZ
Copy link
Contributor

@mofeiZ mofeiZ commented Mar 4, 2025

Array and argument spreads may mutate stateful iterables. Spread sites should have ConditionallyMutate effects (e.g. mutate if the ValueKind is mutable, otherwise read).

See

Note that

  • Object and JSX Attribute spreads do not evaluate iterables (srcs mozilla, ecma)
  • An ideal mutability inference system could model known collections (i.e. Arrays or Sets) as a "mutated collection of non-mutable objects" (see todo-granular-iterator-semantics), but this is not what we do today. As such, an array / argument spread will always extend the range of built-in arrays, sets, etc
  • Due to HIR limitations, call expressions with argument spreads may cause unnecessary bailouts and/or scope merging when we know the call itself has freeze, capture, or read semantics (e.g. useHook(...mutableValue))
    We can deal with this by rewriting these call instructions to (1) create an intermediate array to consume the iterator and (2) capture and spread the array at the callsite

Stack created with Sapling. Best reviewed with ReviewStack.

Comment on lines +8 to +9
const items = makeArray(0, 1, 2, null, 4, false, 6);
return useIdentity(...items.values());
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is the hook call bailout case. This new bailout is good (better correctness, as we were previously memoizing the items.values() iterable)

Ideally we can rewrite this to 1) create an intermediate array to consume the iterable and (2) capture and spread the array at the callsite.

const items = makeArray(0, 1, 2, null, 4, false, 6);
const tmp = [...items.values()]; // this spread has conditionallyMutate effects
return useIdentity([...tmp]); // but this spread has freeze effects

Copy link
Member

@josephsavona josephsavona left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is great! See comments for a couple concrete things to take a look at before landing, up to you if you want to address them here or a follow-up.

Comment on lines 878 to 885
state.referenceAndRecordEffects(
freezeActions,
element.place,
Effect.ConditionallyMutate,
ValueReason.Other,
);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder if we could address most of the negative consequences of this change by just special-casing array here. Basically isBuiltInArrayType(element) ? Effect.Capture : Effect.ConditionallyMutate as the effect. Because spreading doesn't mutate arrays, it only mutates if you have an iterator.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point, done!

@@ -1330,7 +1344,8 @@ function inferBlock(
state.referenceAndRecordEffects(
freezeActions,
place,
Effect.Read,
// see call-spread-argument-mutable-iterator test fixture
arg.kind === 'Spread' ? Effect.ConditionallyMutate : Effect.Read,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

hmm this doesn't seem like it should be necessary - we're inside an if (areArgumentsImmutableAndNonMutating()) check

Array and argument spreads may mutate stateful iterables. Spread sites should have `ConditionallyMutate` effects (e.g. mutate if the ValueKind is mutable, otherwise read).

See
- [ecma spec (13.2.4.1 Runtime Semantics: ArrayAccumulation. SpreadElement : ... AssignmentExpression)](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-arrayaccumulation).
- [ecma spec 13.3.8.1 Runtime Semantics: ArgumentListEvaluation](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-argumentlistevaluation)

Note that
- Object and JSX Attribute spreads do not evaluate iterables (srcs [mozilla](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Spread_syntax#description), [ecma](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-propertydefinitionevaluation))
- An ideal mutability inference system could model known collections (i.e. Arrays or Sets) as a "mutated collection of non-mutable objects" (see `todo-granular-iterator-semantics`), but this is not what we do today. As such, an array / argument spread will always extend the range of built-in arrays, sets, etc
- Due to HIR limitations, call expressions with argument spreads may cause unnecessary bailouts and/or scope merging when we know the call itself has `freeze`, `capture`, or `read` semantics (e.g. `useHook(...mutableValue)`)
  We can deal with this by rewriting these call instructions to (1) create an intermediate array to consume the iterator and (2) capture and spread the array at the callsite
@mofeiZ mofeiZ merged commit ed1264f into main Mar 13, 2025
32 of 35 checks passed
mofeiZ added a commit that referenced this pull request Mar 13, 2025
(see title)
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32522).
* #32596
* #32595
* #32594
* #32593
* __->__ #32522
* #32521
mofeiZ added a commit that referenced this pull request Mar 13, 2025
…ties (#32593)

Expand type inference to infer mixedReadOnly types for numeric and
computed property accesses.
```js
function Component({idx})
  const data = useFragment(...)
  // we want to type `posts` correctly as Array
  const posts = data.viewers[idx].posts.slice(0, 5);
  // ...
}
```
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32593).
* #32596
* #32595
* #32594
* __->__ #32593
* #32522
* #32521
mofeiZ added a commit that referenced this pull request Mar 13, 2025
- Add `at`, `indexOf`, and `includes`
- Optimize MixedReadOnly which is currently only used by hook return
values. Hook return values are typed as Frozen, this change propagates
that to return values of aliasing function calls (such as `at`). One
potential issue is that developers may pass
`enableAssumeHooksFollowRulesOfReact:false` and set
`transitiveMixedData`, expecting their transitive mixed data to be
mutable. This is a bit of an edge case and already doesn't have clear
semantics.
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32594).
* #32596
* #32595
* __->__ #32594
* #32593
* #32522
* #32521
github-actions bot pushed a commit that referenced this pull request Mar 13, 2025
(see title)
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32522).
* #32596
* #32595
* #32594
* #32593
* __->__ #32522
* #32521

DiffTrain build for [38a7600](38a7600)
mofeiZ added a commit that referenced this pull request Mar 13, 2025
Move all gating tests to `gating/`
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32595).
* #32596
* __->__ #32595
* #32594
* #32593
* #32522
* #32521
github-actions bot pushed a commit that referenced this pull request Mar 13, 2025
Array and argument spreads may mutate stateful iterables. Spread sites
should have `ConditionallyMutate` effects (e.g. mutate if the ValueKind
is mutable, otherwise read).

See
- [ecma spec (13.2.4.1 Runtime Semantics: ArrayAccumulation.
SpreadElement : ...
AssignmentExpression)](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-arrayaccumulation).
- [ecma spec 13.3.8.1 Runtime Semantics:
ArgumentListEvaluation](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-argumentlistevaluation)

Note that
- Object and JSX Attribute spreads do not evaluate iterables (srcs
[mozilla](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Spread_syntax#description),
[ecma](https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#sec-runtime-semantics-propertydefinitionevaluation))
- An ideal mutability inference system could model known collections
(i.e. Arrays or Sets) as a "mutated collection of non-mutable objects"
(see `todo-granular-iterator-semantics`), but this is not what we do
today. As such, an array / argument spread will always extend the range
of built-in arrays, sets, etc
- Due to HIR limitations, call expressions with argument spreads may
cause unnecessary bailouts and/or scope merging when we know the call
itself has `freeze`, `capture`, or `read` semantics (e.g.
`useHook(...mutableValue)`)
We can deal with this by rewriting these call instructions to (1) create
an intermediate array to consume the iterator and (2) capture and spread
the array at the callsite
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32521).
* #32596
* #32595
* #32594
* #32593
* #32522
* __->__ #32521

DiffTrain build for [ed1264f](ed1264f)
github-actions bot pushed a commit that referenced this pull request Mar 13, 2025
…ties (#32593)

Expand type inference to infer mixedReadOnly types for numeric and
computed property accesses.
```js
function Component({idx})
  const data = useFragment(...)
  // we want to type `posts` correctly as Array
  const posts = data.viewers[idx].posts.slice(0, 5);
  // ...
}
```
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32593).
* #32596
* #32595
* #32594
* __->__ #32593
* #32522
* #32521

DiffTrain build for [eb53139](eb53139)
github-actions bot pushed a commit that referenced this pull request Mar 13, 2025
- Add `at`, `indexOf`, and `includes`
- Optimize MixedReadOnly which is currently only used by hook return
values. Hook return values are typed as Frozen, this change propagates
that to return values of aliasing function calls (such as `at`). One
potential issue is that developers may pass
`enableAssumeHooksFollowRulesOfReact:false` and set
`transitiveMixedData`, expecting their transitive mixed data to be
mutable. This is a bit of an edge case and already doesn't have clear
semantics.
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32594).
* #32596
* #32595
* __->__ #32594
* #32593
* #32522
* #32521

DiffTrain build for [89a46a5](89a46a5)
mofeiZ added a commit that referenced this pull request Mar 13, 2025
…iers (#32596)

Reduce false positive bailouts by using the same
`isReferencedIdentifier` logic that the compiler also uses for
determining context variables and a function's own hoisted declarations.

Details:
Previously, we counted every babel identifier as a reference. This is
problematic because babel counts most string symbols as an identifier.

```js
print(x);  // x is an identifier as expected
obj.x      // x is.. also an identifier here
{x: 2}     // x is also an identifier here
```

This PR adds a check for `isReferencedIdentifier`. Note that only
non-lval
references pass this check. This should be fine as we don't need to
hoist function declarations before writes to the same lvalue (which
should error in strict mode anyways)
```js
print(x);  // isReferencedIdentifier(x) -> true
obj.x      // isReferencedIdentifier(x) -> false
{x: 2}     // isReferencedIdentifier(x) -> false
x = 2      // isReferencedIdentifier(x) -> false
```
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32596).
* __->__ #32596
* #32595
* #32594
* #32593
* #32522
* #32521
github-actions bot pushed a commit that referenced this pull request Mar 13, 2025
…iers (#32596)

Reduce false positive bailouts by using the same
`isReferencedIdentifier` logic that the compiler also uses for
determining context variables and a function's own hoisted declarations.

Details:
Previously, we counted every babel identifier as a reference. This is
problematic because babel counts most string symbols as an identifier.

```js
print(x);  // x is an identifier as expected
obj.x      // x is.. also an identifier here
{x: 2}     // x is also an identifier here
```

This PR adds a check for `isReferencedIdentifier`. Note that only
non-lval
references pass this check. This should be fine as we don't need to
hoist function declarations before writes to the same lvalue (which
should error in strict mode anyways)
```js
print(x);  // isReferencedIdentifier(x) -> true
obj.x      // isReferencedIdentifier(x) -> false
{x: 2}     // isReferencedIdentifier(x) -> false
x = 2      // isReferencedIdentifier(x) -> false
```
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/32596).
* __->__ #32596
* #32595
* #32594
* #32593
* #32522
* #32521

DiffTrain build for [f457d0b](f457d0b)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CLA Signed React Core Team Opened by a member of the React Core Team
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants