We warn about getInitialState() usage in class components because users may have accidentally used it instead of state. If they have also specified state though then we should skip this warning.
Context: https://twitter.com/soprano/status/810003963286163456
* (WIP) Nesting warnings
* Remove indirection
* Add a note about namespace
* Fix Flow and make host context required
This makes it easier to avoid accidental nulls.
I also added a test for the production case to avoid regressing because of __DEV__ branches.
We should be able to support passing a function to replaceState, which
receives the accumulation of all the previously applied state updates.
Which means we shouldn't drop those updates from the queue. Technically,
we could drop them only when an object is passed to replaceState, but
that seems like more trouble than it's worth.
Go back to using a flag, instead. I removed it before because I thought
we might want to get rid of the top-level UpdateQueue type and put
the fields directly on the fiber, but since we're keeping UpdateQueue
we can put hasForceUpdate on there.
The update is scheduled as if the current processing update has already
been processed; if it has the same or higher priority, it will be
flushed in the same batch.
We also print a warning.
The queue maintains a pointer to the last progressed update in the list.
Updates that come after that pointer are pending. The pointer is set to
the end of the list during reconciliation.
Pending updates are sorted by priority then insertion. Progressed
updates are sorted by the order in which they were applied during
reconciliation, which may not be by priority: if a component bails out
before the updates are committed, in the next render, the progressed
updates are applied in the same order that they were previously, even if
a higher priority update comes in.
Once a progressed update is flushed/committed, it's removed from
the queue.
We need to be able to access both, and since the list uses forward
pointers, it makes more sense to point to the one that comes first.
Otherwise to get the last progressed update you have to start at the
beginning of the list.
When resetting the priority in the complete phase, check the priority of
the update queue so that updates aren't dropped.
Updates inside render, child cWRP, etc are no longer dropped.
The next step is sort the queue by priority and only flush updates that
match the current priority level.
Restructures the update queue to maintain a pointer to the first
pending update, which solves a few problems:
- Updates that occur during the begin phase (e.g. in cWRP of a child)
aren't dropped, like they are currently. This isn't working yet because
the work priority is reset during completion. The following item will
fix it.
- Sets us up to be able to add separate priorities to each update in
the queue. I'll add this in a subsequent commit.
* Implement component stack for some warnings in Fiber
* Keep Fiber debug source up to date
When an element changes, we should copy the source and owner again.
Otherwise they can get stale since we're not reading them from the element.
* Remove outdated TODOs from tests
* Explicitly specify Fiber types to include in the stack
Fixes an accidental omission when both source and owner are null but displayName exists.
* Fix mised Stack+Fiber test to not expect extra warnings
When we're in Fiber mode we don't actually expect that warning being printed.
* Warn on passing different props to super()
* Implement duplicate key warnings
We keep known keys in a set in development. There is an annoying special case where we know we'll check it again because we break out of the loop early.
One test in the tree hook regresses to the failing state because it checks that the tree hook works without a Set available, but we started using Set in this code. It is not essential and we can clean this up later when we decide how to deal with polyfills.
* Move ReactTypeOfWork to src/shared
It needs to be available both to Fiber and Isomorphic because the tree hook lives in Isomorphic but pretty-prints Fiber stack.
* Add dev-only ReactDebugCurrentFiber for warnings
The goal is to use ReactCurrentOwner less and rely on ReactDebugCurrentFiber for warning owner name and stack.
* Make Stack invariant messages more consistent
Fiber used a helper so two tests had the same phrasing.
Stack also used a helper for most invariants but hardcoded a different phrase in one place.
I changed that invariant message to use a helper which made it consistent with what it prints in Fiber.
* Make CSSPropertyOperations use getCurrentFiberOwnerName()
This gets mount-time CSS warnings to be printed.
However update-time warnings are currently ignored because current fiber is not yet available during the commit phase.
We also regress on HostOperation hook tests but this doesn't matter because it's only used by ReactPerf and it doesn't work with Fiber yet anyway. We'll have to think more about it later.
* Set ReactDebugCurrentFiber during the commit phase
This makes it available during updates, fixing the last failing test in CSSPropertyOperations.
* Add DOM warnings by calling hooks directly
It is not clear if the old hook system is worth it in its generic incarnation. For now I am just hooking it up to the DOMFiber renderer directly.
* Add client-side counterparts for some warning tests
This helps us track which warnings are really failing in Fiber, and which ones depend on SSR.
This matches what we do in Fiber -- and doing it this way is the only way we can prepare new views in the background before unmounting old ones.
In particular, this breaks this pattern:
```js
class Child1 extends React.Component {
render() { ... }
componentWillMount() {
this.props.registerChild(this);
}
componentWillUnmount() {
this.props.unregisterChild();
}
}
class Child2 extends React.Component {
render() { ... }
componentWillMount() {
this.props.registerChild(this);
}
componentWillUnmount() {
this.props.unregisterChild();
}
}
class Parent extends React.Component {
render() {
return (
showChild1 ?
<Child1
registerChild={(child) => this.registered = child}
unregisterChild={() => this.registered = null}
/> :
<Child2
registerChild={(child) => this.registered = child}
unregisterChild={() => this.registered = null}
/>
);
}
}
```
Previously, `this.registered` would always be set -- now, after a rerender, `this.registered` gets stuck at null because the old child's componentWillUnmount runs *after* the new child's componentWillMount.
A correct fix here is to use componentDidMount rather than componentWillMount. (In general, componentWillMount should not have side effects.) If Parent stored a list or set of registered children instead, there would also be no issue.
This matches the behavior in Fiber. Normally we would change Fiber
to match Stack to minimize breaking changes for the initial release.
However, in this case it would require too large a compromise to change
Fiber to act like Stack.
* Remove array indirection in host context
* Keep a single context stack with a null sentinel
This lets us keep subtrees separated without maintaining independent context arrays for subtrees.
* There is always exactly one null by the time we pop a portal
I was trying to be smart but didn't need to.
* Cache current context
* Enable additional (failing) portal tests
* Fix portal unmounting
When unmount a portal, we need to unmount *its* children from itself.
This is similar to what we would do for a root if we allowed deleting roots.
* Skip portals when looking for host siblings
A portal is not part of that host tree despite being a child.
* Fix comment typo
* Add a failing test for portal child reconciliation
It is failing because portal bails out of update, seeing null in pendingProps.
It is null because we set pendingProps to nextPortal.children, which is null in this test.
* Fix the bug when switching to a null portal child
If pendingProps is null, we do a bailout in beginWork.
This prevents unmounting of the existing child when the new child is null.
We fix this by changing portal fiber's pendingProps to be the portal object itself instead of its children.
This way, it is never null, and thus doesn't cause a false positive in the bailout condition.
* Add a comment about HostPortal in getHostSibling
* Revert the fix because I don't know why it worked
unmountHostComponents() should have worked despite finding the wrong parent because it should have changed the parent when pushing and popping the portals.
* Don't call commitDeletion recursively
This leads to a "Cannot find host parent" bug because commitDeletion() clears the return field.
When we're inside the loop, we assign node.sibling.return to node.return but by this moment node.return has already been nulled.
As a solution we inline code from commitDeletion() without the nulling.
It still fails tests but for a different reason (unrelated bug).
* Skip portal children in commitNestedUnmounts()
We are currently already visiting them in commitUnmount() portal case since it's recursive.
This condition avoids visting them twice.
* Set node.child.return before going deeper
It doesn't seem to influence existing tests but we have this protection in all other similar places.
It protects against infinite loops.
* Revert "Fix the bug when switching to a null portal child"
This reverts commit ed9747deed.
I'll solve this by using an array in place of null instead.
* Use [] for empty Portal pendingProps
This avoids a false positive bailout with pendingProps == null when portal is empty.
* Test that SVG elements get created with the right namespace
* Pass root to the renderer methods
* Keep track of host instances and containers
* Keep instances instead of fibers on the stack
* Create text instances in begin phase
* Create instance before bailing on offscreen children
Otherwise, the parent gets skipped next time.
We could probably create it later but this seems simpler.
* Tweak magic numbers in incremental tests
I don't understand why they changed but probably related to us moving some work into begin phase?
* Only push newly created nodes on the parent stack
Previously I was pushing nodes on the parent stack regardless of whether they were already in current or not.
As a result, insertions during updates were duplicated, and nodes were added to existing parents before commit phase.
Luckily we have a test that caught that.
* Fix lint
* Fix Flow
I had to wrap HostContext API into a closure so that it's parameterizeable with I and C.
* Use the same destructuring style in scheduler as everywhere else
* Remove branches that don't seem to run anymore
I'm not 100% sure this is right but I can't get tests to fail.
* Be explicit about the difference between type and tag
I was confused by th HACK comment so I learned how DOM and SVG work with casing and tried to write a more descriptive comment.
It also seems like passing fiber.type into finalizeInitialChildren() is a potential problem because DOM code assumes tag is lowercase.
So I added a similar "hack" to finalizeInitialChildren() that is identical to the one we have prepareUpdate() so if we fix them later, we fix both.
* Save and restore host context when pushing and popping portals
* Revert parent context and adding children in the begin phase
We can address this later separately as it is a more hot path.
This doesn't affect correctness of SVG container behavior.
* Add a test for SVG updates
This tests the "jump" reuse code path in particular.
* Record tests
* Read ownerDocument from the root container instance
This way createInstance() depends on the innermost container only for reading the namespace.
* Track namespaces instead of creating instances early
While we might want to create instance in the begin phase, we shouldn't let DOM guide reconciler design.
Instead, we are adding a new concept of "host context". In case of ReactDOMFiber, it's just the current namespace.
We are keeping a stack of host context values, ignoring those that are referentially equal.
The renderer receives the parent context and type, and can return a new context.
* Pop child context before reading own context and clarify API
It wasn't quite clear from the API which context was being returned by the renderer. Changed the API to specifically ask for child context, and thus to pop it before getting the current context.
This fixes the case with <foreignObject> to which I intended to give SVG namespace.
* Give SVG namespace to <svg> itself
* Don't allocate unnecessarily when reconciling portals
We create stacks lazily so that if portal doesn't contain <svg>s, we don't need to allocate.
We also reuse the same object for portal host context state instead of creating a new one every time.
* Add more tests for edge cases
* Fix up math namespace
* Maintain a separate container stack
* Fix rebase mistakes
* Unwind context on errors
* Reset the container state when reusing the object
* Add getChildHostContext() to ReactART
* Record tests
Solves a few things:
- Moves code out of performWork into workLoop so that it can
be optimized.
- errorLoop (now clearErrors) is no longer called recursively.
- Removes a while (true) loop inside performWork (not that it really
matters for optimization, since performWork contains a try block and
will be deopted, anyway).
If a ref throws while detaching, it should not prevent
componentWillUnmount from being called.
Additionally, errors thrown by removeChild should not be ignored, even
as the result of unmounting a failed subtree.
Subsequent failures should propagate the error to the next boundary.
The conceptual model for this is
try {
render();
} catch (error) {
setStateToRecover();
render();
}
Closes#8485
This uncovered a separate issue where some errors were being unscheduled
and rescheduled multiple times before flushing. Turns out we need to
check both a fiber and its alternate when determining if it represents a
failed unit of work. I didn't notice this before because we were
scheduling a new update on *every* boundary at the end of the commit
phase, not just the ones that captured an error during that commit. I
updated the unit tests to catch this in the future.
We still need tryComponentWillUnmount because deletion should be non-
interruptible.
Refactoring the commit phase to better handle errors also allows us
to switch between fast and slow versions of the work loop. The slower
version runs whenever there are captured errors; it must check each
unit of work to see if it has failed. The faster version runs in the
normal case where there are no errors. Whenever an error is thrown, we
switch from the fast work loop to the slow work loop.
Andrew Clark
Andrew Clark
Brian Vaughn
Brian Vaughn
Dan Abramov
Ben Alpert
Sebastian Markbage
Sebastian Markbåge
Ben Alpert