* Support ForwardRef type of work in TestRenderer and ShallowRenderer.
* Release script now updates inter-package dependencies too (e.g. react-test-renderer depends on react-is).
We'll use this in www to test whether the polyfill is better at
scheduling high-pri async work than the native one. My preliminary tests
suggest "yes" but it's hard to say for certain, given how difficult it
is to consistently reproduce the starvation issues we've been seeing.
* Use module pattern so context stack is isolated per renderer
* Unify context implementations
Implements the new context API on top of the existing ReactStack that we
already use for host context and legacy context. Now there is a single
array that we push and pop from.
This makes the interrupt path slightly slower, since when we reset the
unit of work pointer, we have to iterate over the stack (like before)
*and* switch on the type of work (not like before). On the other hand,
this unifies all of the unwinding behavior in the UnwindWork module.
* Add DEV only warning if stack is not reset properly
We have other tests that would have caught this if resuming were enabled
in all cases, but since it's currently only enabled for error
boundaries, the test I've added to prevent a regression is a
bit contrived.
* Context providers and consumers should bail-out on already finished work
Fixes bug where a consumer would re-render even if its props and context
had not changed.
* Encode output as JSON string
* Add weights to random action generator
* Add context to triangle fuzz tester
* Move bailouts to as early as possible
* Bailout if neither context value nor children haven't changed (sCU)
* Change prop type invariant to a DEV-only warning
Using `new Map(iterable)` isn't supported in IE11, so it ends up trying to iterate through an empty map and these attributes don't get defined in properties. Since this is only run once on startup inlining the attributeName array is probably fine.
Changes `createResource` to return an object with `read` and `preload`
methods. Future methods may include `set`, `subscribe`, `invalidate`,
and so on.
* Don't expose ReactGlobalSharedState on React Native renderer
We should just go through the "react" package if need access to this one.
Removed the dependencies in React Native.
* No longer used by InspectorUtils
`oldProps` was null. This went uncaught by the unit tests because
ReactNoop did not use `oldProps` in either `prepareUpdate` or
`completeUpdate`. I added some invariants so we don't regress in
the future.
* Revert "Replace danger token with a refreshed facebook-open-source-bot token (#12295)"
This reverts commit 2d511479c4.
* Revert "Temporarily disable Danger in CI (#12291)"
This reverts commit 925fc93389.
* Add stack unwinding phase for handling errors
A rewrite of error handling, with semantics that more closely match
stack unwinding.
Errors that are thrown during the render phase unwind to the nearest
error boundary, like before. But rather than synchronously unmount the
children before retrying, we restart the failed subtree within the same
render phase. The failed children are still unmounted (as if all their
keys changed) but without an extra commit.
Commit phase errors are different. They work by scheduling an error on
the update queue of the error boundary. When we enter the render phase,
the error is popped off the queue. The rest of the algorithm is
the same.
This approach is designed to work for throwing non-errors, too, though
that feature is not implemented yet.
* Add experimental getDerivedStateFromCatch lifecycle
Fires during the render phase, so you can recover from an error within the same
pass. This aligns error boundaries more closely with try-catch semantics.
Let's keep this behind a feature flag until a future release. For now, the
recommendation is to keep using componentDidCatch. Eventually, the advice will
be to use getDerivedStateFromCatch for handling errors and componentDidCatch
only for logging.
* Reconcile twice to remount failed children, instead of using a boolean
* Handle effect immediately after its thrown
This way we don't have to store the thrown values on the effect list.
* ReactFiberIncompleteWork -> ReactFiberUnwindWork
* Remove startTime
* Remove TypeOfException
We don't need it yet. We'll reconsider once we add another exception type.
* Move replay to outer catch block
This moves it out of the hot path.
* Add test exercising public API to test BeforeInputEventPlugin + FallbackCompositionState
- I've adopted a similar approach to the existing test for BeforeInputEventPlugin
- I've simulated events and then assert the event handler for onBeforeInput is fired or not fired based on the test conditions
- The scenarios are tested against IE11, Webkite and Presto environment simulations
- I've encorporated what I understand to be the functionality in the FallbackCompositionState test
* Prettier
* Linting fixes
* Remove test for contenteditable in Presto - the contenteditable type is not supported in Presto powered browsers (Opera).
* Remove mention of Presto as this explicit condition is no longer handled in BeforeInputEventPlugin.
We still need to exercise usage of FallbackCompositionState though so let's keep a test where the env does not support Composition and Text events.
* Add tests for envs with only CompositionEvent support
* Remove internal tests no longer needed
* Shorten test case names to satisfy lint rules
* Add tests for onCompositionStart and onCompositionUpdte events
The BeforeInputEventPlugin is responsible for emitting these events so we need to add tests for this. This also ensure we exercise the code path that, L207, that was not previously exercised with the public tests.
If we have multiple RN renderers running simultaneously, we should be able to send a single event to all of them and only if it recognizes the event will it do anything with it. Crucially, this avoids the 'Unsupported top level event type "%s" dispatched' invariant in those cases.
* [experimental] simple-cache-provider
Pushing an early version of this for testing and demonstration purposes.
* Change invariant to DEV-only warning
* Use function overloading for createResource type
Expresses that primitive keys do not require a hash function, but
non-primitive keys do.
* More tests
* Use export *
* Make Record type a disjoint union
* Pass miss argument separate from key to avoid a closure
* Additional release script options for publishing canary versions
- `branch` specifies a branch other than master
- `local` skips pulling from the remote branch and checking CircleCI
- `tag` specifies an npm dist tag other than `latest` or `next`
We may add a higher-level `canary` option in the future.
* Address Brian's feedback:
- Updated description of `local` option
- Throws if the `latest` tag is specified for a prerelease version
