Note that bailing out adds false positives for hoisted functions whose
only references are within other functions. For example, this rewrite
would be safe.
```js
// source program
function foo() {
return bar();
}
function bar() {
return 42;
}
// compiler output
let bar;
if (/* deps changed */) {
function foo() {
return bar();
}
bar = function bar() {
return 42;
}
}
```
These false positives are difficult to detect because any maybe-call of
foo before the definition of bar would be invalid.
Instead of bailing out, we should rewrite hoisted function declarations
to the following form.
```js
let bar$0;
if (/* deps changed */) {
// All references within the declaring memo block
// or before the function declaration should use
// the original identifier `bar`
function foo() {
return bar();
}
function bar() {
return 42;
}
bar$0 = bar;
}
// All references after the declaring memo block
// or after the function declaration should use
// the rewritten declaration `bar$0`
```
I noticed that we increase this in the recursive part of the algorithm.
This would mean that we'd count a key more than once if it has Server
Components inside it recursively resolving. This moves it out to where
we enter from toJSON. Which is called once per JSON entry (and therefore
once per key).
This revisits a validation I built a while ago, trying to make it more strict this time to ensure that it's high-signal.
We detect function expressions which are *known* mutable — they definitely can modify a variable defined outside of the function expression itself (modulo control flow). This uses types to look for known Store and Mutate effects only, and disregards mutations of effects. Any such function passed to a location with a Freeze effect is reported as a validation error.
This is behind a flag and disabled by default. If folks agree this makes sense to revisit, i'll test out internally and we can consider enabling by default.
ghstack-source-id: 075a731444
Pull Request resolved: https://github.com/facebook/react/pull/33079
If a function captures a mutable value but never gets called, we don't infer a mutable range for that function. This means that we also don't alias the function with its mutable captures.
This case is tricky, because we don't generally know for sure what is a mutation and what may just be a normal function call. For example:
```js
hook useFoo() {
const x = makeObject();
return () => {
return readObject(x); // could be a mutation!
}
}
```
If we pessimistically assume that all such cases are mutations, we'd have to group lots of memo scopes together unnecessarily. However, if there is definitely a mutation:
```js
hook useFoo(createEntryForKey) {
const cache = new WeakMap();
return (key) => {
let entry = cache.get(key);
if (entry == null) {
entry = createEntryForKey(key);
cache.set(key, entry); // known mutation!
}
return entry;
}
}
```
Then we have to ensure that the function and its mutable captures alias together and end up in the same scope. However, aliasing together isn't enough if the function and operands all have empty mutable ranges (end = start + 1).
This pass finds function expressions and object methods that have an empty mutable range and known-mutable operands which also don't have a mutable range, and ensures that the function and those operands are aliased together *and* that their ranges are updated to end after the function expression. This is sufficient to ensure that a reactive scope is created for the alias set.
NOTE: The alternative is to reject these cases. If we do that we'd also want to similarly disallow cases like passing a mutable function to a hook.
ghstack-source-id: 5d8158246a
Pull Request resolved: https://github.com/facebook/react/pull/33078
Building on mofeiz's recent work to type constructors. Also, types for reanimated values which are useful in the next PR.
ghstack-source-id: 1c81e213a1
Pull Request resolved: https://github.com/facebook/react/pull/33077
This pass didn't previously report the precise difference btw inferred/manual dependencies unless a debug flag was set. But the error message is really good (nice job mofeiz): the only catch is that in theory the inferred dep could be a temporary that can't trivially be reported to the user.
But the messages are really useful for quickly verifying why the compiler couldn't preserve memoization. So here we switch to outputting a detailed message about the discrepancy btw inferred/manual deps so long as the inferred dep root is a named variable. I also slightly adjusted the message to handle the case where there is no diagnostic, which can occur if there were no manual deps but the compiler inferred a dependency.
ghstack-source-id: 534f6f1fec
Pull Request resolved: https://github.com/facebook/react/pull/33095
If a JSX attribute value is a string that contains unicode or other characters that need special escaping, we wrap the attribute value in an expression container. However, our unicode to detect this only handled the basic unicode character plane, not the "astral" plane which includes emojis.
This PR updates the regex to detect such extended characters and also use an expression container.
ghstack-source-id: 6d9c8e4dd2
Pull Request resolved: https://github.com/facebook/react/pull/33096
Our runtime results are still bogus, but so far this tweak let's the llm
use the tools a little more intelligently.
example:
---
```js
import React from 'react';
function App() {
const [count, setCount] = React.useState(0);
const [data, setData] = React.useState([]);
React.useEffect(() => {
const processData = () => {
console.log("Processing data...");
const newData = [];
for (let i = 0; i < 1000000000; i++) {
newData.push({ id: i, value: Math.random() * count });
}
return newData;
};
const processedData = processData();
setData(processedData);
}, [count]);
return (
<div>
<p>Count: {count}</p>
<p>Data Length: {data.length}</p>
<ul>
{data.slice(0, 5).map((item) => (
<li key={item.id}>Item {item.id}: {item.value.toFixed(2)}</li>
))}
</ul>
</div>
);
};
```
why is this component slow? use the react tools to figure this out.
> I'll analyze why this component is slow using the React tools. Let's
first evaluate the performance of the original code.
>
[review-react-runtime ...]
>
> The performance measurement tool timed out, which is a strong
indication of just how severe the performance issue is in this
component. Let's now try to compile the code to see if React Compiler
can identify any issues:
>
[compile...]
>
> Now I understand the issue. Let's analyze why this component > is
extremely slow:
> Main Performance Issues:
>
> Massive Data Processing: The component is trying to create an array
with a billion (1,000,000,000) objects in the processData function. This
is an enormous amount of data that will:
>
> Consume massive amounts of memory
> Take an extremely long time to process
> Likely crash the browser
>
>
> Unnecessary Recalculation: The effect depends on [count], meaning
every time the count changes, it will reprocess this enormous dataset.
> Inefficient Data Handling: Even though only 5 items are displayed, the
component processes a billion items.
>
> Let's fix these issues:
>
[compile...]
>
> Now let's check if our modified code performs better:
>
[review-react-runtime...]
>
> Excellent! The performance has dramatically improved. Now let's check
the React documentation to make sure we're following the best practices:
>
[query-react-dev-docs...]
(hit the maximum length conversation limit because i'm on a free plan)
Since we use esbuild we need to correctly move dependencies that are
required at runtime into `dependencies` and other packages that are only
used in development in to `devDependencies`. This ensures the correct
packages are included in the build.
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/33101).
* #33085
* #33084
* #33083
* #33082
* __->__ #33101
Because we now decided whether to outline in the flushing phase, when
we're writing the preamble we don't yet know if we will make that
decision so we don't know if it's safe to omit the external runtime.
However, if you are providing an external runtime it's probably a pretty
safe bet you're streaming something dynamically that's likely to need it
so we can always include it.
The main thing is that this makes it hard to test it because it affects
our tests in ways it wouldn't otherwise so we have to add a bunch of
conditions.
Stacked on #33076.
This fixes a bug where we used the "complete" status but the
DOMContentLoaded event. This checks for not "loading" instead.
We also add a new status where the boundary has been marked as complete
by the server but has not yet flushed either due to being throttled,
suspended on CSS or animating.
Stacked on #33073.
React semantics is that Suspense boundaries reveal with a throttle
(300ms). That helps avoid flashing reveals when a stream reveals many
individual steps back to back. It can also improve overall performance
by batching the layout and paint work that has to happen at each step.
Unfortunately we never implemented this for SSR streaming - only for
client navigations. This is highly noticeable on very dynamic sites with
lots of Suspense boundaries. It can look good with a client nav but feel
glitchy when you reload the page or initial load.
This fixes the Fizz runtime to be throttled and reveals batched into a
single paint at a time. We do this by first tracking the last paint
after the complete (this will be the first paint if `rel="expect"` is
respected). Then in the `completeBoundary` operation we queue the
operation and then flush it all into a throttled batch.
Another motivation is that View Transitions need to operate as a batch
and individual steps get queued in a sequence so it's extra important to
include as much content as possible in each animated step. This will be
done in a follow up for SSR View Transitions.
Stacked on #33066 and #33068.
Currently we're passing `errorDigest` to `completeBoundary` if there is
a client side error (only CSS loading atm). This only exists because of
`completeBoundaryWithStyles`. Normally if there's a server-side error
we'd emit the `clientRenderBoundary` instruction instead. This adds
unnecessary code to the common case where all styles are in the head.
This is about to get worse with batching because client render shouldn't
be throttled but complete should be.
The first commit moves the client render logic inline into
`completeBoundaryWithStyles` so we only pay for it when styles are used.
However, the approach I went with in the second commit is to reuse the
`$RX` instruction instead (`clientRenderBoundary`). That way if you have
both it ends up being amortized. However, it does mean we have to emit
the `$RX` (along with the `$RC` helper if any
`completeBoundaryWithStyles` instruction is needed.
Stacked on #33065.
The runtime is about to be a lot more complicated so we need to start
sharing some more code.
The problem with sharing code is that we want the inline runtime to as
much as possible be isolated in its scope using only a few global
variables to refer across runtimes.
A problem with Closure Compiler is that it refuses to inline functions
if they have closures inside of them. Which makes sense because of how
VMs work it can cause memory leaks. However, in our cases this doesn't
matter and code size matters more. So we can't use many clever tricks.
So this just favors writing the source in the inline form. Then we add
an extra compiler pass to turn those global variables into local
variables in the external runtime.
We weren't treating terminal operands as eligible for memoization in PruneNonEscapingScopes, which meant that they could end up un-memoized. Terminal operands can also be compound ReactiveValues like SequenceExpressions, so part of the fix is to make sure we don't just recurse into compound values but record the full aliasing information we would for top-level instructions.
Still WIP, this needs to handle terminals other than for..of.
ghstack-source-id: 09a2923051
Pull Request resolved: https://github.com/facebook/react/pull/33062
## Summary
Fix babel presets, and add a bit more context to the tool so that it is
more reliable
## How did you test this change?
Manually tested the mcp integrated with claude desktop
We normally expect the segment to exist whatever the client does while
streaming. However, when hydration errors at the root of the shell for a
whole document render, then we clear nodes from body which can include
our segments. We don't need them anymore because we switched to client
rendering.
It triggers an error accessing parent node which can safely be ignored.
This just helps avoid confusion in this scenario.
This also covers up the error in #33067. Which doesn't actually cause
any visible problems other than error logging. However, ideally we
wouldn't emit completeBoundary instructions if the boundary is inside a
cancelled fallback.
```js
function Component() {
useEffect(() => {
let hasCleanedUp = false;
document.addEventListener(..., () => hasCleanedUp ? foo() : bar());
// effect return values shouldn't be typed as frozen
return () => {
hasCleanedUp = true;
}
};
}
```
### Problem
`PruneHoistedContexts` currently strips hoisted declarations and
rewrites the first `StoreContext` reassignment to a declaration. For
example, in the following example, instruction 0 is removed while a
synthetic `DeclareContext let` is inserted before instruction 1.
```js
// source
const cb = () => x; // reference that causes x to be hoisted
let x = 4;
x = 5;
// React Compiler IR
[0] DeclareContext HoistedLet 'x'
...
[1] StoreContext reassign 'x' = 4
[2] StoreContext reassign 'x' = 5
```
Currently, we don't account for `DeclareContext let`. As a result, we're
rewriting to insert duplicate declarations.
```js
// source
const cb = () => x; // reference that causes x to be hoisted
let x;
x = 5;
// React Compiler IR
[0] DeclareContext HoistedLet 'x'
...
[1] DeclareContext Let 'x'
[2] StoreContext reassign 'x' = 5
```
### Solution
Instead of always lowering context variables to a DeclareContext
followed by a StoreContext reassign, we can keep `kind: 'Const' | 'Let'
| 'Reassign' | etc` on StoreContext.
Pros:
- retain more information in HIR, so we can codegen easily `const` and
`let` context variable declarations back
- pruning hoisted `DeclareContext` instructions is simple.
Cons:
- passes are more verbose as we need to check for both `DeclareContext`
and `StoreContext` declarations
~(note: also see alternative implementation in
https://github.com/facebook/react/pull/32745)~
### Testing
Context variables are tricky. I synced and diffed changes in a large
meta codebase and feel pretty confident about landing this. About 0.01%
of compiled files changed. Among these changes, ~25% were [direct
bugfixes](https://www.internalfb.com/phabricator/paste/view/P1800029094).
The [other
changes](https://www.internalfb.com/phabricator/paste/view/P1800028575)
were primarily due to changed (corrected) mutable ranges from
https://github.com/facebook/react/pull/33047. I tried to represent most
interesting changes in new test fixtures
`
Fixes an edge case in React Compiler's effects inference model.
Returned values should only be typed as 'frozen' if they are (1) local
and (2) not a function expression which may capture and mutate this
function's outer context. See test fixtures for details
---
[//]: # (BEGIN SAPLING FOOTER)
Stack created with [Sapling](https://sapling-scm.com). Best reviewed
with [ReviewStack](https://reviewstack.dev/facebook/react/pull/33047).
* #32765
* #32747
* __->__ #33047
## Summary
Add a way for the agent to get some data on the performance of react
code
## How did you test this change?
Tested function independently and directly with claude desktop app
---------
Co-authored-by: Sebastian "Sebbie" Silbermann <sebastian.silbermann@vercel.com>
Same principle as #33029 but for Flight.
We pretty aggressively create separate rows for things in Flight (every
Server Component that's an async function create a microtask). However,
sync Server Components and just plain Host Components are not. Plus we
should ideally ideally inline more of the async ones in the same way
Fizz does.
This means that we can create rows that end up very large. Especially if
all the data is already available. We can't show the parent content
until the whole thing loads on the client.
We don't really know where Suspense boundaries are for Flight but any
Element is potentially a point that can be split.
This heuristic counts roughly how much we've serialized to block the
current chunk and once a limit is exceeded, we start deferring all
Elements. That way they get outlined into future chunks that are later
in the stream. Since they get replaced by Lazy references the parent can
potentially get unblocked.
This can help if you're trying to stream a very large document with a
client nav for example.
Adds Fragment Ref support to RN through the Fabric config, starting with
`observeUsing`/`unobserveUsing`. This is mostly a copy from the
implementation on DOM, and some of it can likely be shared in the future
but keeping it separate for now and we can refactor as we add more
features.
Added a basic test with Fabric, but testing specific methods requires so
much mocking that it doesn't seem valuable here.
I built Fabric and ran on the Catalyst app internally to test with
intersection observers end to end.
Stacked on #32736.
That way you can find the owner stack of each component that rerendered
for context.
In addition to the JSX callsite tasks that we already track, I also
added tracking of the first `setState` call before rendering.
We then run the "Update" entries in that task. That way you can find the
callsite of the first setState and therefore the "cause" of a render
starting by selecting the "Update" track.
Unfortunately this is blocked on bugs in Chrome that makes it so that
these stacks are not reliable in the Performance tab. It basically just
doesn't work.
This is a new extension that Chrome added to the existing
`console.timeStamp` similar to the extensions added to
`performance.measure`. This one should be significantly faster because
it doesn't have the extra object indirection, it doesn't return a
`PerformanceMeasure` entry and doesn't register itself with the global
system of entries.
I also use `performance.measure` in DEV for errors since we can attach
the error to the `properties` extension which doesn't exist for
`console.timeStamp`.
A downside of using this API is that there's no programmatic API for the
site itself to collect its own logs from React. Which the previous
allowed us to use the standard `performance.getEntries()` for. The
recommendation instead will be for the site to patch `console.timeStamp`
if it wants to collect measurements from React just like you're
recommended to patch `console.error` or `fetch` or whatever to collect
other instrumentation metrics.
This extension works in Chrome canary but it doesn't yet work fully in
Chrome stable. We might want to wait until it has propagated to Chrome
to stable. It should be in Chrome 136.
Follow up to #33027.
This enhances the heuristic so that we accumulate the size of the
currently written boundaries. Starting from the size of the root (minus
preamble) for the shell.
This ensures that if you have many small boundaries they don't all
continue to get inlined. For example, you can wrap each paragraph in a
document in a Suspense boundary to regain document streaming
capabilities if that's what you want.
However, one consideration is if it's worth producing a fallback at all.
Maybe if it's like `null` it's free but if it's like a whole alternative
page, then it's not. It's possible to have completely useless Suspense
boundaries such as when you nest several directly inside each other. So
this uses a limit of at least 500 bytes of the content itself for it to
be worth outlining at all. It also can't be too small because then for
example a long list of paragraphs can never be outlined.
In the fixture I straddle this limit so some paragraphs are too small to
be considered. An unfortunate effect of that is that you can end up with
some of them not being outlined which means that they appear out of
order. SuspenseList is supposed to address that but it's unfortunate.
The limit is still fairly high though so it's unlikely that by default
you'd start outlining anything within the viewport at all. I had to
reduce the `progressiveChunkSize` by an order of magnitude in my fixture
to try it out properly.
`requestFormReset` incorrectly tries to get the current dispatch queue
from the Fiber. However, the Fiber might be the workInProgress which is
an inconsistent state.
This hack just tries the other Fiber if it detects one of the known
inconsistent states but there can be more.
Really we should stash the dispatch queue somewhere stateful which is
effectively what `setState` does by binding it to the closure.
## Summary
We don't need the isArray check for this experiment, as
`fastAddProperties` already does the same. Also renaming
slowAddProperties to make it clearer we can fully remove this codepath
once fastAddProperties is fully rolled out.
## How did you test this change?
```
yarn test packages/react-native-renderer -r=xplat --variant=true
```
When we end up creating an incomplete state in the shell we end up not
flushing anything. As a hack, in this case we need to reset the
ResumableState because some of the ResumableState is still relevant
(e.g. any preloads that went into headers) but some of the
ResumableState needs to be reset since they assume that what we produced
actually flushed.
We didn't reset the instructions state but we haven't actually flushed
any of the instructions so it needs to reset.
Since the very beginning we have had the `progressiveChunkSize` option
but we never actually took advantage of it because we didn't count the
bytes that we emitted. This starts counting the bytes by taking a pass
over the added chunks each time a segment completes.
That allows us to outline a Suspense boundary to stream in late even if
it is already loaded by the time that back-pressure flow and in a
`prerender`. Meaning it gets inserted with script.
The effect can be seen in the fixture where if you have large HTML
content that can block initial paint (thanks to
[`rel="expect"`](https://github.com/facebook/react/pull/33016) but also
nested Suspense boundaries). Before this fix, the paint would be blocked
until the large content loaded. This lets us paint the fallback first in
the case that the raw bytes of the content takes a while to download.
You can set it to `Infinity` to opt-out. E.g. if you want to ensure
there's never any scripts. It's always set to `Infinity` in
`renderToHTML` and the legacy `renderToString`.
One downside is that if we might choose to outline a boundary, we need
to let its fallback complete.
We don't currently discount the size of the fallback but really just
consider them additive even though in theory the fallback itself could
also add significant size or even more than the content. It should maybe
really be considered the delta but that would require us to track the
size of the fallback separately which is tricky.
One problem with the current heuristic is that we just consider the size
of the boundary content itself down to the next boundary. If you have a
lot of small boundaries adding up, it'll never kick in. I intend to
address that in a follow up.
When effect dependencies cannot be inferred due to memoization-related
bailouts or unexpected mutable ranges (which currently often have to do
with writes to refs), fall back to traversing the effect lambda itself.
This fallback uses the same logic as PropagateScopeDependencies:
1. Collect a sidemap of loads and property loads
2. Find hoistable accesses from the control flow graph. Note that here,
we currently take into account the mutable ranges of instructions (see
`mutate-after-useeffect-granular-access` fixture)
3. Collect the set of property paths accessed by the effect
4. Merge to get the set of minimal dependencies
Inferred effect dependencies and inlined jsx (both experimental
features) rely on `InferReactivePlaces` to determine their dependencies.
Since adding type inference for phi nodes
(https://github.com/facebook/react/pull/30796), we have been incorrectly
inferring stable-typed value blocks (e.g. `props.cond ? setState1 :
setState2`) as non-reactive. This fix patches InferReactivePlaces
instead of adding a new pass since we want non-reactivity propagated
correctly
See https://github.com/rollup/plugins/issues/1425
Currently, `@babel/helper-string-parser/lib/index.js` is either emitted
as a wrapped esmodule or inline depending on the ordering of async
functions in `rollup/commonjs`. Specifically,
`@babel/types/lib/definitions/core.js` is cyclic (i.e. transitively
depends upon itself), but sometimes
`@babel/helper-string-parser/lib/index.js` is emitted before this is
realized.
A relatively straightforward patch is to wrap all modules (see
https://github.com/rollup/plugins/issues/1425#issuecomment-1465626736).
This only regresses `eslint-plugin-react-hooks` bundle size by ~1.8% and
is safer (see
https://github.com/rollup/plugins/blob/master/packages/commonjs/README.md#strictrequires)
> The default value of true will wrap all CommonJS files in functions
which are executed when they are required for the first time, preserving
NodeJS semantics. This is the safest setting and should be used if the
generated code does not work correctly with "auto". Note that
strictRequires: true can have a small impact on the size and performance
of generated code, but less so if the code is minified.
(note that we're on an earlier version of `@rollup/commonjs` which does
not default to `strictRequires: true`)
Mofei Zhang
mofeiZ
Sebastian Markbåge
Stephen Zhou
Sebastian "Sebbie" Silbermann
Joe Savona
Jan Kassens
lauren
Jorge Cabiedes
Jack Pope
Pieter De Baets