This flag first moves the `shouldYield()` logic into React itself. We
need this for `postTask` compatibility anyway since this logic is no
longer a concern of the scheduler. This means that there can also be no
global `requestPaint()` that asks for painting earlier. So this is best
rolled out with `enableAlwaysYieldScheduler` (and ideally
`enableYieldingBeforePassive`) instead of `enableRequestPaint`.
Once in React we can change the yield timing heuristics. This uses the
previous 5ms for Idle work to keep everything responsive while doing
background work. However, for Transitions and Retries we have seen that
same thread animations (like loading states animating, or constant
animations like cool Three.js stuff) can take CPU time away from the
Transition that causes moving into new content to slow down. Therefore
we only yield every 25ms.
The purpose of this yield is not to avoid the overhead of yielding,
which is very low, but rather to intentionally block any frequently
occurring other main thread work like animations from starving our work.
If we could we could just tell everyone else to throttle their stuff for
ideal scheduling but that's not quite realistic. In other words, the
purpose of this is to reduce the frame rate of animations to 30 fps and
we achieve this by not yielding. We still do yield to allow the
animations to not just stall. This seems like a good balance.
The 5ms of Idle is because we don't really need to yield less often
since the overhead is low. We keep it low to allow 120 fps animations to
run if necessary and our work may not be the only work within a frame so
we need to yield early enough to leave enough time left.
Similarly we choose 25ms rather than say 35ms to ensure that we push
long enough to guarantee to half the frame rate but low enough that
there's plenty of time left for a rAF to power each animation every
other frame. It's also low enough that if something else interrupts the
work like a new interaction, we can still be responsive to that within
50ms or so. We also need to yield in case there's I/O work that needs to
get bounced through the main thread.
This flag is currently off everywhere since we have so many other
scheduling flags but that means there's some urgency to roll those out
fully so we can test this one. There's also some tests to update since
this doesn't go through the Mock scheduler anymore for yields.
This treats workInProgressRoot work and rootWithPendingPassiveEffects
the same way. Basically as long as there's some work on the root, yield
the current task. Including passive effects. This means that passive
effects are now a continuation instead of a separate callback. This can
mean they're earlier or later than before. Later for Idle in case
there's other non-React work. Earlier for same Default if there's other
Default priority work.
This makes sense since increasing priority of the passive effects beyond
Idle doesn't really make sense for an Idle render.
However, for any given render at same priority it's more important to
complete this work than start something new.
Since we special case continuations to always yield to the browser, this
has the same effect as #31784 without implementing `requestPaint`. At
least assuming nothing else calls `requestPaint`.
<img width="587" alt="Screenshot 2024-12-14 at 5 37 37 PM"
src="https://github.com/user-attachments/assets/8641b172-8842-4191-8bf0-50cbe263a30c"
/>
We introduced the `unstable_useContextWithBailout` API to run compiler
based experiments. This API was designed to be an experiment proxy for
alternative approaches which would be heavier to implement. The
experiment turned out to be inconclusive. Since most of our performance
critical usage is already optimized, we weren't able to find a clear win
with this approach.
Since we don't have further plans for this API, let's clean it up.
Related to #31752.
When hydrating, we have two different ways of handling a Suspense
boundary that the server has already given up on and decided to client
render. If we have already hydrated the parent and then later this
happens, then we'll use the retry lane like any ping. If we discover
that it was already in client-render mode when we discover the Suspense
boundary for the first time, then schedule a default lane to let us
first finish the current render and then upgrade the priority to sync to
try to client render this boundary as soon as possible since we're
holding back content.
We used to use the `DefaultHydrationLane` for this but this is not
really a Hydration. It's actually a client render. If we get any other
updates flowing in from above at the same time we might as well do them
in the same pass instead of two passes. So this should be considered
more like any update.
This also means that visually the client render pass now gets painted as
a render instead of a hydration.
This show the flow of a shell being hydrated at the default priority,
then a Suspense boundary being discovered and hydrated at Idle and then
an inner boundary being discovered as client rendered which gets
upgraded to default.
<img width="1363" alt="Screenshot 2024-12-14 at 12 13 57 AM"
src="https://github.com/user-attachments/assets/a141133e-4856-4f38-a11f-f26bd00b6245"
/>
When scheduling the initial root and when using
`unstable_scheduleHydration` we should use the Hydration Lanes rather
than the raw update lane. This ensures that we're always hydrating using
a Hydration Lane or the Offscreen Lane rather than other lanes getting
some random hydration in it.
This fixes an issue where updating a root while it is still hydrating
causes it to trigger client rendering when it could just hydrate and
then apply the update on top of that.
It also fixes a potential performance issue where
`unstable_scheduleHydration` gets batched with an update that then ends
up forcing an update of a boundary that requires it to rewind to do the
hydration lane anyway. Might as well just start with the hydration
without the update applied first.
I added a kill switch (`enableHydrationLaneScheduling`) just in case but
seems very safe given that using `unstable_scheduleHydration` at all is
very rare and updating the root before the shell hydrates is extremely
rare (and used to trigger a recoverable error).
A long standing issue for React has been that if you reorder stateful
nodes, they may lose their state and reload. The thing moving loses its
state. There's no way to solve this in general where two stateful nodes
swap.
The [`moveBefore()`
proposal](https://chromestatus.com/feature/5135990159835136?gate=5177450351558656)
has now moved to
[intent-to-ship](https://groups.google.com/a/chromium.org/g/blink-dev/c/YE_xLH6MkRs/m/_7CD0NYMAAAJ).
This function is kind of like `insertBefore` but preserves state.
There's [a demo here](https://state-preserving-atomic-move.glitch.me/).
Ideally we'd port this demo to a fixture so we can try it.
Currently this flag is always off - even in experimental. That's because
this is still behind a Chrome flag so it's a little early to turn it on
even in experimental. So you need a custom build. It's on in RN but only
because it doesn't apply there which makes it easier to tell that it's
safe to ship once it's on everywhere else.
The other reason it's still off is because there's currently a semantic
breaking change. `moveBefore()` errors if both nodes are disconnected.
That happens if we're inside a completely disconnected React root.
That's not usually how you should use React because it means effects
can't read layout etc. However, it is currently supported. To handle
this we'd have to try/catch the `moveBefore` to handle this case but we
hope this semantic will change before it ships. Before we turn this on
in experimental we either have to wait for the implementation to not
error in the disconnected-disconnected case in Chrome or we'd have to
add try/catch.
In preparation for the next RC, I set this feature flag to true
everywhere. I did not delete the feature flag yet, in case there are yet
more bugs to be discovered.
I also didn't remove the dynamic feature flag from the Meta builds; I'll
let the Meta folks handle that.
Reverts facebook/react#31403 to reenable lazy context propagation
The disabling was to produce a build that could help track down whether
this flag is causing a possibly related bug in transitions but we don't
intend to disable it just fix forward once we figure out what the
problem is
disables lazy context propagation in oss to help determine if it is
causing bugs in startTransition. Will reenable after cutting a canary
release with this flag disabled
We're seeing issues with this feature internally including bugs with
sibling prerendering and errors that are difficult for developers to
action on. We'll turn off the feature for the time being until we can
improve the stability and ergonomics.
This PR does two things:
- Turn off `enableInfiniteLoopDetection` everywhere while leaving it as
a variant on www so we can do further experimentation.
- Revert https://github.com/facebook/react/pull/31061 which was a
temporary change for debugging. This brings the feature back to
baseline.
This flag will be used to gate a new timeline profiler that's integrate
with the Performance Tab and the new performance.measure extensions in
Chrome.
It replaces the existing DevTools feature so this disables
enableSchedulingProfiler when it is enabled since they can interplay in
weird ways potentially.
This means that experimental React now disable scheduling profiler and
enables this new approach.
Insertion effects do not unmount when a subtree is removed while
offscreen.
Current behavior for an insertion effect is if the component goes
- *visible -> removed:* calls insertion effect cleanup
- *visible -> offscreen -> removed:* insertion effect cleanup is never
called
This makes it so we always call insertion effect cleanup when removing
the component.
Likely also fixes https://github.com/facebook/react/issues/26670
---------
Co-authored-by: Rick Hanlon <rickhanlonii@fb.com>
To recap. This only affects DEV and RSC. It patches console on the
server in DEV (similar to how React DevTools already does and what we
did for the double logging). Then replays those logs with a `[Server]`
badge on the client so you don't need a server terminal open.
This has been on for over 6 months now in our experimental channel and
we've had a lot of coverage in Next.js due to various experimental flags
like taint and ppr.
It's non-invasive in that even if something throws we just serialize
that as an unknown value.
The main feedback we've gotten was:
- The serialization depth wasn't deep enough which I addressed in #30294
and haven't really had any issues since. This could still be an issue or
the inverse that you serialize too many logs that are also too deep.
This is not so much an issue with intentional logging and things like
accidental errors don't typically have unbounded arguments (e.g. React
errors are always string arguments). The ideal would be some way to
retain objects and then load them on-demand but that needs more
plumbing. Which can be later.
- The other was that double logging on the server is annoying if the
same terminal does both the RSC render and SSR render which was
addressed in #30207. It is now off by default in node/edge-builds of the
client, on by default in browser builds. With the `replayConsole` option
to either opt-in or out.
We've reached a good spot now I think.
These are better with `enableOwnerStacks` but that's a separate track
and not needed.
The only thing to document here, other than maybe that we're doing it,
is the `replayConsole` option but that's part of the RSC renderers that
themselves are not documented so nowhere to document it.
enableHalt turns on a mode for flight prerenders where aborts are
treated like infinitely stalled outcomes while still completing the
prerender. For regular tasks we simply serialize the slot as a promise
that never settles. For ReadableStream, Blob, and Async Iterators we
just never advance the serialization so they remain unfinished when
consumed on the client.
When enableHalt is turned on aborts of prerenders will halt rather than
error. The abort reason is forwarded to the upstream produces of the
aforementioned async iterators, blobs, and ReadableStreams. In the
future if we expose a signal that you can consume from within a render
to cancel additional work the abort reason will also be forwarded there
Persistent renderers used the `Update` effect flag to check if a subtree
needs to be cloned. In some cases, that causes extra renders, such as
when a layout effect is triggered which only has an effect on the JS
side, but doesn't update the host components.
It's been a bit tricky to find the right places where this needs to be
set and I'm not 100% sure I got all the cases even though the tests
passed.
There is currently a mismatch in how the persistent mode JS API and the
Fabric native code interpret `completeRoot`.
This is a short-lived experiment to see the effect of moving the Fabric
`completeRoot` call from `finalizeContainerChildren` to
`replaceContainerChildren` which in some cases does not get called.
**This API is not intended to ship. This is a temporary unstable hook
for internal performance profiling.**
This PR exposes `unstable_useContextWithBailout`, which takes a compare
function in addition to Context. The comparison function is run to
determine if Context propagation and render should bail out earlier.
`unstable_useContextWithBailout` returns the full Context value, same as
`useContext`.
We can profile this API against `useContext` to better measure the cost
of Context value updates and gather more data around propagation and
render performance.
The bailout logic and test cases are based on
https://github.com/facebook/react/pull/20646
Additionally, this implementation allows multiple values to be compared
in one hook by returning a tuple to avoid requiring additional Context
consumer hooks.
Following https://github.com/facebook/react/pull/30436
Concurrent by default strategy has been unshipped. Here we clean up the
`allowConcurrentByDefault` path and related logic/tests.
For now, this keeps the `concurrentUpdatesByDefaultOverride` argument in
`createContainer` and `createHydrationContainer` and ignores the value
to prevent more breaking changes to `react-reconciler` in the RC stage.
Concurrent by default has been unshipped! Let's clean it up.
Here we remove `forceConcurrentByDefaultForTesting`, which allows us to
run tests against both concurrent strategies. In the next PR, we'll
remove the actual concurrent by default code path.
While the goal is to remove legacy context completely, I think we can
already land the removal of legacy context for function components. I
didn't even know this feature existed until reading the code recently.
The win is just a couple of property lookups on function renders, but it
trims down the API already as the full removal will likely still take a
bit more time.
www: Starting with enabled test renderer and a feature flag for
production rollout.
RN: Not enabled, will follow up on this.
Object literals should be faster at least on React Native with Hermes as
the JS engine.
It might also be interesting to confirm the old comments in this file
from years ago are even still valid. Creating an object from a literal
should be a simpler operation.
It's a bit unfortunate that this introduces a bunch of copied code, but
since we rearely update the fields on fibers, this seems like an okay
tradeoff for a hot code path. An alternative would be some sort of macro
system, but that doesn't seem worth the extra complexity.
## Summary
We currently do deep diffing for object props, and also use custom
differs, if they are defined, for props with custom attribute config.
The idea is to simply do a `===` comparison instead of all that work. We
will do less computation on the JS side, but send more data to native.
The hypothesis is that this change should be neutral in terms of
performance. If that's the case, we'll be able to get rid of custom
differs, and be one step closer to deleting view configs.
This PR adds the `enableShallowPropDiffing` feature flag to support this
experiment.
## How did you test this change?
With `enableShallowPropDiffing` hardcoded to `true`:
```
yarn test packages/react-native-renderer
```
This fails on the following test cases:
- should use the diff attribute
- should do deep diffs of Objects by default
- should skip deeply-nested changed functions
Which makes sense with this change. These test cases should be deleted
if the experiment is shipped.
