Summary:
Here we refine the `ShadowNode::BaseTraits` static method. Before this change, ConcreteViewShadowNode<> enforces Layoutable and YogaLayoutable traits. This change moves that responsibility to LayoutableShadowNode and YogaLayoutableShadowNode which makes overall logic more coherent.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052027
fbshipit-source-id: fd25264204b0232b1dbbff6f9dfefd9fbcb146c4
Summary:
Now we can move custom to `YogaLayoutableShadowNode` code from `ConcreteViewShadowNode<>` template to `YogaLayoutableShadowNode` itself reducing the amount of templated code and reduce overall complexity.
Note, here we have to do `dynamic_cast` for now, we will address that in coming diffs.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: mdvacca
Differential Revision: D20052021
fbshipit-source-id: dac5969a97b75e54c7728a1ca8161922bd2245ca
Summary:
The fact that `LayoutableShadowNode` now inherits `ShadowNode` allows us to de-virtualize `getLayoutableChildNodes` and move that to `LayoutableShadowNode`. Less code, less virtual dispatch, less complexity.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052032
fbshipit-source-id: 580e86b5a746028e470788e00027f247bf77126c
Summary:
D19963353 mentioned the infrastructure that re-routes methods calls related to adding and cloning children between YogaLayoutableShadowNode and ShadowNode. `cloneAndReplaceChild` is exactly this. It was implemented as a virtual method that is called from `ConcreteViewShadowNode`. The whole process requires building a list of children of some class and passing that as a list of pointers. Now we don't need it all that because we can call directly and statically. That change will allow us to simplify that infra even more in the future diffs.
With all previous changes, now we can implement `getYogaLayoutableChildren` inside `YogaLayoutableShadowNode` and call that statically. Eventually, that will allow us to remove templated `getChildrenSlice`. Previously the call for that method must be in `ConcreteViewShadowNode`, now it's not true anymore which we will use later to even better goodness.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: mdvacca
Differential Revision: D20052020
fbshipit-source-id: e5c819a4d21b2dbcd08f3439e1783e3a9cba5ef4
Summary:
D19963353 mentioned the infrastructure that re-routes methods calls related to adding and cloning children between YogaLayoutableShadowNode and ShadowNode. `cloneAndReplaceChild` is exactly this. It was implemented as a virtual method that is called from `ConcreteViewShadowNode`. The whole process requires building a list of children of some class and passing that as a list of pointers. Now we don't need it all that because we can call directly and statically. That change will allow us to simplify that infra even more in the future diffs.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052022
fbshipit-source-id: ddf341c112edd8a2f79eaf74465a9a360a168541
Summary:
This is another perf-critical place where using traitCast should help with perf.
After this change we practically will not have any `dynamic_cast`s (except those which will be delited (with methods where they are used) in the future diffs).
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052029
fbshipit-source-id: 63b589d3c3322f0afd72f18b4701fe31d0e05c2f
Summary:
We use this function in Diffing, and Diffing is one of the hottest pieces of Fabric. Removing `dynamic_cast` here should improve perf. See the previous diff for more details.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052023
fbshipit-source-id: d6427ebf076d7f8244f0aeaf16407efa2368957e
Summary:
The Diffing is one of the hottest pieces of Fabric. Removing dynamic_cast here should improve perf. See the previous diff for more details.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052030
fbshipit-source-id: 27fd9f34f2a1f9d22b9da6b1e3c1a2982045c07a
Summary:
`traitCast` is a special form of static_cast that checks additional requirements (similar to `dynamic_cast`) before performing the cast. We will use that in many places in the coming diffs.
Restructuring the class hierarchy in the previous diff finally allows us to do static casts among ShadowNode and LayoutableShadowNode and other classes (previously it wasn't allowed because of lack of common base class).
Why we don't want to use `dynamic_cast`:
* It's expensive (and we need to do the cast on the hottest fragments of the framework). (See: (1) http://www.stroustrup.com/fast_dynamic_casting.pdf and (2) https://www.youtube.com/watch?v=ARYP83yNAWk Herb Sutter & proposal of `down_cast`).
* It's code-size inefficient, whereas `static_cast` has zero runtime and code-size overhead.
* Removing `dynamic_cast` will allow us finally to opt-out `RTTI` (additional code size and perm wins).
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D20052024
fbshipit-source-id: d293c1cf80deb7817d333d5306d6b32bf3abdb27
Summary:
This is a very crucial change, everything else in this stack depends on it.
Here is the set of constraints that we have for ShadowNode-based class hierarchy:
* `ShadowNode` is a base class that defines basic operations on the nodes. It does not have virtual methods by design (virtual dispatch hurts performance; we want to limit and centralize that in `ComponentDescriptor`s).
* `ConcreteShadowNode<>` template *statically* wires particular `ShadowNode` and particular `Props` establishing a type-safe relationship between them ensured on compile time.
* Not all `ShadowNode`s are "layoutable", not all "layoutable" `ShadowNode`s use Yoga to do layout.
* Layotability (and YogaLayotability) feature is implemented as two classes `LayoutableShadowNode` and `YogaLayoutableShadowNode`. These classes essentially need to know something about the ShadowNode nature of the object (get the list of children or replace some children).
Before the change, `LayoutableShadowNode` and `YogaLayoutableShadowNode`classes did not inherit `ShadowNode` because we don't want to use *virtual inheritance* for `ShadowNode`: `ConcreteShadowNode<>` already inherits `ShadowNode`, so if we make `LayoutableShadowNode` inherits `ShadowNode`, we will have to somehow flatten this inheritance hierarchy (and the virtual inheritance is an answer to that). (Yes, C++ supports multiple inheritance and virtual inheritance.)
Before this change, we solved this dilemma this way: We have a subclass-template `ConcreteViewShadowNode<>` that inherits `ConcreteShadowNode<>` and `YogaLayoutableShadowNode`. Then, we had a bunch of methods that implement the rerouting of some functionality from `YogaLayoutableShadowNode` to `ShadowNode` and vise-versa. (See the diagram "Before".)
That worked fine, except the caveats:
* That wiring is nasty, complex, hard to reason about and overall limiting.
* There is no way to statically cast `LayoutableShadowNode` to `ShadowNode` (because there is no common base class). That forces us to use dynamic_cast on some perf critical paths (including layout, diffing and so on).
* Adding features that rely on interop between `LayoutableShadowNode` and `ShadowNode` is a nightmare.
It should be a better way to deal with this dilemma, and this diff implements a different approach: We can have a base class of `ConcreteShadowNode<>` as a template parameter. With this approach, we can make `LayoutableShadowNode` inherit `ShadowNode` and when we need to instantiate a `ConcreteShadowNode<>` that needs to be layoutable, we can just specify `YogaLayoutableShadowNode` as a base class. (See the diagram "After".)
This simple change will allow us to simplify a lot of things. The rest of the stack is about getting rid of unnecessary moving parts. Which will finally allow us to build "Inline Views" feature.
```
╭──────────────────────╮
│ ◎ ○ ○ ░░░░░░░░░░░░░░░│
├──────────────────────┤
│ │
│ │
│ Before │
│ │ ┌────────────────────────────┐ ┌────────────────────────────┐
│ │ │ │ │ │
│ │ │ ShadowNode │ │ LayoutableShadowNode │
└──────────────────────┘ │ │ │ │
└────────────────────────────┘ └────────────────────────────┘
▲ ▲
│ │
╔════════════════════════════╗ ┌────────────────────────────┐
║ ║ │ │
┌────────�║ ConcreteShadowNode<> ║ │ YogaLayoutableShadowNode │
│ ║ ║ │ │
│ ╚════════════════════════════╝ └────────────────────────────┘
│ ▲ ▲
│ │ │
│ │ │
│ │ │
│ │ ╔════════════════════════════╗ │
│ │ ║ ║ │
│ └─────║ ConcreteViewShadowNode<> ║─┘
│ ║ ║
│ ╚════════════════════════════╝
│ ▲
│ │
│ ┌──────────────┴───────────────┐
│ │ │
│ │ │
│ │ │
┌────────────────────────────┐ ┌────────────────────────────┐ ┌────────────────────────────┐
│ │ │ │ │ │
│ TextShadowNode │ │ ViewShadowNode │ │ ParagraphShadowNode │
│ │ │ │ │ │
└────────────────────────────┘ └────────────────────────────┘ └────────────────────────────┘
╭──────────────────────╮
│ ◎ ○ ○ ░░░░░░░░░░░░░░░│
├──────────────────────┤
│ │ ┌────────────────────────────┐
│ │ │ │
│ After │ │ ShadowNode │
│ │ │ │
│ │ └────────────────────────────┘
│ │ ▲
└──────────────────────┘ ┌────────────────────────────────────┤
│ │
│ ╔════════════════════════════╗
┌────────────────────────────┐ ║ ConcreteShadowNode ║
│ │ ║ <Base: ShadowNode> ║
│ LayoutableShadowNode │ ║ ║
│ │ ╚════════════════════════════╝
└────────────────────────────┘ ▲
▲ │
│ ┌────────────────────────────┐
┌────────────────────────────┐ │ │
│ │ │ TextShadowNode │
│ YogaLayoutableShadowNode │ │ │
│ │ └────────────────────────────┘
└────────────────────────────┘
▲
│
╔════════════════════════════╗
║ ConcreteShadowNode ║
║ <Base: ║
║ YogaLayoutableShadowNode> ║
╚════════════════════════════╝
▲
│
╔════════════════════════════╗
║ ║
║ ConcreteViewShadowNode<> ║
║ ║
╚════════════════════════════╝
▲
├─────────────────────────────────────┐
│ │
┌────────────────────────────┐ ┌────────────────────────────┐
│ │ │ │
│ ParagraphShadowNode │ │ ViewShadowNode │
│ │ │ │
└────────────────────────────┘ └────────────────────────────┘
```
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19963353
fbshipit-source-id: b65c8a5064bdb54ab64f08a8e546aa9e2b5a486b
Summary:
For backwards-compatibility with Paper, we're implementing a feature in Fabric that will allow TextInputs to use the default padding from the theme in Android.
Note that this uses some pretty ugly hacks that probably shouldn't be used inside of components at all: looking directly at rawProps, overriding props/Yoga styles in the component descriptor, etc. I would (personally) really like to kill this feature entirely unless and until we can find a more elegant solution.
Changelog: [Internal]
TextInputs are still not pixel-perfect with Paper, but they're much closer, and the underline visual glitchiness is no longer an issue.
Reviewed By: mdvacca
Differential Revision: D20109605
fbshipit-source-id: 543282843e0a9f03a504d72d7a014431099bd64c
Summary:
Changelog: [Internal]
As agreed in https://fb.quip.com/Oh2mAaTAbBj6, `findNodeAtPoint` now calls callback with `shadowNode` instead of `instanceHandle`.
Reviewed By: shergin
Differential Revision: D20097269
fbshipit-source-id: bd2a1bcd26ab2510f16c3e73f628be4b1f7dacfc
Summary:
Changelog: [Internal]
# Problem
Calling `UIManager::updateState` does not increment state revision because it calls `ConcreteComponentDescriptor::createState` which creates new state with state revision 1.
# How did this propagate?
This error propagated itself in TextInput when trying to input a value, you would be only allowed to type in 1 character.
Reviewed By: JoshuaGross
Differential Revision: D20072844
fbshipit-source-id: 37b8173307e1d91d6e9c41b5ff2e185dde31cc38
Summary:
We are moving towards 100%-prettified files. That's the first step when we apply Clang Format for `ReactCommon`.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D20110895
fbshipit-source-id: 0a0ce4997cf1c3721b0b07ef78c1a57ce87d20f9
Summary:
Until now, there were two measure functions that differ in only one parameter (rootTag). The rootTag is used to use the context associated to the tag as part of the calculation of layout, otherwise it just uses the ReactApplicationContext.
This diff unifies both method into an unique method that
changelog: [internal]
Reviewed By: JoshuaGross
Differential Revision: D20081281
fbshipit-source-id: b1f6a6cedbf78f36f36fd0f93407c23c6996d76b
Summary:
We recently updated React Native's docs site to have its own domain reactnative.dev and needed to update the URLs in the source code
CHANGELOG:
[INTERNAL]
Reviewed By: hramos
Differential Revision: D20072842
fbshipit-source-id: 1970d9214c872a6e7abf697d99f8f5360b3b308e
Summary:
Changelog: [internal]
Repro steps:
1. Go to fbsource/fbobjc.
2. Run following focus command P124860753.
3. Enable Address Sanitizer in Xcode.
3. Run the app from Xcode.
5. Make sure `react_fabric.enable_marketplace_home_ios` and `react_fabric.enabled_state_reconciliation_ios` Mobile Config flags are set to true.
6. Navigate to Marketplace Home and observe use of freed memory exception. You have to wait until second page of marketplace home feed is loaded.
{F228880980}
# Root cause
`ChangedShadowNodePairs` being a vector of pairs of references.
Those `shared_ptr` do not outlive the the run of `reconcileStateWithTree` (that's what I originally thought when I made them references).
This was introduced in D19814500.
Reviewed By: JoshuaGross
Differential Revision: D19977784
fbshipit-source-id: 8c4a381137baa08a6af9d3c1cb295fe6c2b2bf95
Summary:
Changelog: [Internal]
This allows to block queue until image is fetched.
In paper, the photo of dog will not appear unless the network request is "fast enough".
{F228670428}
Reviewed By: shergin
Differential Revision: D19907906
fbshipit-source-id: 98d68abf1255388f9b97fb9da367d8f583a4220d
Summary:
Having the overridden function that returns a different type is apparently not a good idea (and might cause bugs and unexpected behavior), so it was renamed. The function also got a new return type (`const &` instead of `std::shared_ptr`) for simplicity, better performance, and smaller code size.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: JoshuaGross
Differential Revision: D19837694
fbshipit-source-id: b7a96424bd040409371724907b3fb3931cd8a2e8
Summary:
With new logic we save one attributed string copy operation.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19844371
fbshipit-source-id: 62f6b0f9c8514a1b55224ccaa52c8fa89c06c9e7
Summary:
Changelog: [Internal]
There shouldn't be any logical changes to the code.
Reviewed By: JoshuaGross
Differential Revision: D19814500
fbshipit-source-id: 2be492eebb284c7e241662d9d78117badcb2eee4
Summary:
# Problem
We calculate `stateRevision_` in constructor from children, but children change after shadowNode is initialised and before it is sealed. So the `stateRevision_` we calculate in constructor can be incorrect.
# Solution
Recalculate `stateRevision_` whenever children change. This can happen in two methods `ShadowNode::replaceChild` and `ShadowNode::appendChild`. This diff implements this solution.
Changelog: [Internal]
Reviewed By: JoshuaGross
Differential Revision: D19813840
fbshipit-source-id: 8fc3b6601e4618f4ee5b322eebc230e0bbb92e3a
Summary:
Before this change the Element<> API didn't allow to specify state objects and crashed during instantiation of non-null states.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: JoshuaGross
Differential Revision: D19816358
fbshipit-source-id: 95ba5e03ea98c0260593462146e8523c95245e2b
Summary:
This pull request fixes a few `#include`s in the Fabric source:
* Changes `<Glog/logging.h>` to `<glog/logging.h>` in two files, which was an issue for case-sensitive file systems
* In `ParagraphProps.cpp`, changes the include of `attributedstring/primitives.h` from a relative style to a more absolute style.
## Changelog
[Internal] [Fixed] - Fabric: Fix various includes
Pull Request resolved: https://github.com/facebook/react-native/pull/27947
Test Plan: Fabric compiles perfectly, and the Fabric test suite passes.
Reviewed By: shergin
Differential Revision: D19788148
Pulled By: sammy-SC
fbshipit-source-id: 30a297e8aea50c7c7fd5f48745edf04811a6bdbd
Summary:
This pull request fixes the CXX stub for `TextLayoutManager`. The signature for `TextLayoutManager::measure` changed slightly, and this patch updates the stub to reflect that.
## Changelog
[Internal] [Fixed] - Fabric: Fix CXX Stub for TextLayoutManager
Pull Request resolved: https://github.com/facebook/react-native/pull/27948
Test Plan: Fabric compiles, and the Fabric test suite passes.
Reviewed By: shergin
Differential Revision: D19788207
Pulled By: sammy-SC
fbshipit-source-id: 8f42cca0be99022b038742728ad4fec6c04d5cf0
Summary:
This change removes the concept of `StateTarget`, replacing its role with `ShadowNodeFamily`.
When `StateTarget` was built, we didn't have a concept of `Family`, and when we added it we introduced a retain-cycle: ShadowNode -> Family -> StateTarget -> ShadowNode. This diff fixes that.
This change does not change conceptually how the state behaves, it just adjusts internal machinery.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19799013
fbshipit-source-id: c1360bfbf6b8ac34e2a856a40047eafeb50ed070
Summary:
Before this change, the concrete component-specific data/payload object associated with a State was stored inside a templated subclass as a normal instance variable; after the change, it's stored as a shared pointer inside the base class. The original motivation was that storing that inside subclass saves us one shared pointer and one heap allocation.
This approach overcomplicated a lot of things and all possible savings are probably compensated with additional complexity (we have to have templated state-update lamdas in subclasses and so on). And to update the data in the previous approach we need to create a shared pointer to data anyway.
This change will allow future improvements in the coming diff.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: JoshuaGross
Differential Revision: D19799014
fbshipit-source-id: 287ed939353ba58d9e434d1502ecfbb208c6daa5
Summary:
Only use State Reconciliation during certain commits from ReactJS, not during State commits or others
The idea is that this will have better perf, since many types of commits don't need State Reconciliation at all.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D19794999
fbshipit-source-id: 336325f225d993a6aae9f55cb95a7a2b38a6d853
Summary:
This implements proposal #2 in our State architecture doc: https://fb.quip.com/bm2EAVwL7jQ5
Problem description: see the text in the comment of TreeStateReconciliation.h
Solution: see also comments in TreeStateReconciliation.h.
Changelog: [internal]
Reviewed By: mdvacca
Differential Revision: D19617329
fbshipit-source-id: 845fb5fe27f2591be433b6d77799707b3516fb1a
Summary:
Systraces were disabled a year ago in D14019272 because we suspected they can negatively affect perf. We don't think this is the case anymore.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D19786284
fbshipit-source-id: 185ed45b134fdcadf131cfddfcf8faf37537a684
Summary:
These props are applied to `View`, therefore they most not be a part of base text attributes, so we reset them.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: mdvacca
Differential Revision: D19764144
fbshipit-source-id: a0d2900e3161c47b83114360c843fa85a4389f8a
Summary:
Of course, compare_exchange_strong didn't actually do what I wanted.
Using a mutex is simpler and actually has the semantics we want: atomically get the current value, compare, and bail if the value is the same, or swap and continue.
Changelog: [Internal]
Reviewed By: shergin
Differential Revision: D19754451
fbshipit-source-id: 6b0aef217b235959af683ec5e31b07a0dd7bb040
Summary:
Simple hack to prevent duplicate onLayout events from being emitted to JS.
Because of the addition of State Reconciliation to the Fabric lifecycle (see previous diff), in certain circumstances entire subtrees can be relayed-out many, many times even though they aren't changing. For JS product code that responds to every onLayout and forces a ReactJS tree commit (see: some usages of VirtualizedList) this can cause an async infinite loop of commits and layouts even though the tree and the LayoutMetrics aren't actually changing. Even though nothing is changing, it can still cause serious performance regressions and even some bugs since the internals of various state machines may assume onLayout won't be called many times.
Changelog: [Internal]
Reviewed By: shergin
Differential Revision: D19715280
fbshipit-source-id: d879e24f1c7b1f710ad430b7473aa9293d093dea
Summary:
Adding a `.clang-tidy` to a bunch of dirs under `react-native-github/ReactAndroid` and `react-native-github/ReactCommon`.
I don't want to add a single `.clang-tidy` at the root because we'll need more fine-grained control over what checks are enabled in different parts of the codebase; for example, fabric will and TM will probably have more checks enabled than older parts of the codebase that we're not actively modernizing, and the Hermes team probably wants its own config to be consistent with the rest of their codebase.
Starting off each `.clang-tidy` by only enabling clang-diagnostic; this is just to test that it's working. In the future, we'll work with the community to gradually enable more checks.
Changelog: [Internal]
Reviewed By: mdvacca
Differential Revision: D19705749
fbshipit-source-id: 979cea053b645ac4a9790340033bfcfb49ca0f97
Summary:
Changelog: [internal]
`ShadowNodeFamily` can be used as target in state updates instead of `ShadowNode`.
Reviewed By: shergin
Differential Revision: D19517428
fbshipit-source-id: 6831357e749239d5afec1dfd2d44a26ca6553e51
Summary:
Changelog: [internal]
Merges all of responsibilities of `StateCoordinator` into `ShadowNodeFamily`.
Reviewed By: shergin
Differential Revision: D19500104
fbshipit-source-id: f31ffded5a840e722fd898eef6a9f52cd2186df7
Summary:
This is a quite fragile and important part of the render engine. We need to dirty Yoga node only in cases where a change affects layout. In the case of over-dirtying, we can kill performance. In the case of under-dirtying, we can produce an incorrect layout.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19596279
fbshipit-source-id: 9f2ac67c44cb35c8ba44be1025b94b7921b74e17
Summary:
This is aligned with all simular changes that we do for `*Props` types (see D19583582 and D19390813).
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19596281
fbshipit-source-id: 283b89c65504e20a461b12c2d1218325c6621dd8
Summary:
Reasons:
* The name of the method now better represent what it's doing;
* It exposes information about "dirty" state of the node without opening actual `LayoutableShadowNode` protected APIs;
* It's a tiny bit faster now because it checks the flag before calling Yoga.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19596282
fbshipit-source-id: 3d87d9d5ba20bb8e360683f149b5ebf90beecd65
Summary:
It makes perfect sense because `Builder` builds totally new shadow trees, so those are not sealed or used by anyone yet.
Assigning it to const shared pointer will do logical sealing (and it does not requires const-cast).
Fewer const-casts in the code, fewer bugs.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19596284
fbshipit-source-id: 75d1c706034958ba7e4bc80a68af75a57c46eb6f
Summary:
Before this change `Element<X>` cannot have children of `Element<Y>` which was wrong because we don't have such limitation in Fabric.
Changelog: [Internal] Fabric-specific internal change.
Reviewed By: sammy-SC
Differential Revision: D19596283
fbshipit-source-id: 9002f5dd42b3d05e7cf492499499399c97b58152