Summary:
Put `prepareForRecycle` to last before enqueue to recycle pool, ensure only call it when count lower than RCTComponentViewRegistryRecyclePoolMaxSize.
cc. shergin .
[General] [Changed] - Reorder prepareForRecycle before adding recycle pool
Pull Request resolved: https://github.com/facebook/react-native/pull/24025
Differential Revision: D14536843
Pulled By: shergin
fbshipit-source-id: 82a816e2c0afb5a6bb72637d7d55d6a4fda708af
Summary: Several things need to ironed out: 1) LocalState in Fabric C++, 2) setting dimensions of BottomSheet component to 0,0 for parent.
Reviewed By: shergin
Differential Revision: D14426167
fbshipit-source-id: 45a90a7971c87672872108a9e360926b4a6095f0
Summary:
This fixes a few memory leaks in fabrics handling of colors for borders, when using CGColorRef's we must be diligent about releasing the memory back.
[iOS] [Fixed] - Border style memory leaks
Pull Request resolved: https://github.com/facebook/react-native/pull/23815
Differential Revision: D14431250
Pulled By: shergin
fbshipit-source-id: dc663c633ae24809cb4841800d31a6ac6eeb8aa5
Summary:
I measured the performance of UIMananger methods and found that `create` method spends a significant amount of time on preallocating views (around 10 ms on my device). Interestingly, this time was actually spent on dispatching operations on the main thread, not on performing those operations. That makes me think about think about the preallocation problem one more time.
(Here and later I discuss preemptive optimistic preallocation views on iOS, but the issues are more-less applicable for all platforms.)
BTW, what's view preallocation? – In React Native we believe that we can get significant TTI wins instantiating platform views ahead of time while JavaScript thread is busy doing reconciliation and the main one is idle.
So, seems the current approach has those downsides:
* We spend too much time on dispatching only (jumping threads, creation Objective-C blocks, managing them in GCD, incrementing atomic refcountes inside `ShadowView` and so on). That's wasteful.
* We don't have much information during preallocations that can help prepare something to save time in the future. We don't know the future size of the component, so we cannot e.g. pre-draw a border or even allocate memory for a future drawing.
* We pre-allocate to much. At the moment of component creation, we don't know will this component be filtered out by view-flattening infra or not, so we always allocate a view for it. That's ~30% more views that we actually need.
* We don't stop allocating (or dispatching instructions for that) after the recycle pool is already filled in. That's also wasteful.
Why is this so bad and how can we fix it properly?..
I thought about this problem for months. And I think the answer is trivial: We don't know the exact shape of the interface until we finish creating it, and there is no way to overcome this problem on the client side. In the ideal world, the server size (or hardcoded manifest somewhere) tells the renderer (at the moment where we started navigating to it) how many views of which type will some surface use.
And until our world is not so perfect, I think we can get a significant performance improvement doing simple preallocation of 100 regular views, 30 text views, and 30 image views during rendering the first React Native surface. So I hardcoded that.
Reviewed By: JoshuaGross
Differential Revision: D14333606
fbshipit-source-id: 96beeb58b546258de1b8fd58550e0ae412b78aa9
Summary:
Right now we rely on the Paper UIManager to update animated node graphs - this hooks us into `RCTSurfacePresenter` in the same way so we are no longer reliant on Paper. Should also help with complex ordering corner cases with pre vs. post operations and restoring defaults when nodes are removed. More info:
https://github.com/facebook/react-native/pull/11819/files
Note that we don't have a way to differentiate animation nodes related to fabric views vs. paper views, so if paper and fabric are both rendering updates simultaneously it's possible they could get processed by the wrong callback. That should be very rare, rarely cause problems even if it does happen, and won't be a problem at all in a post-Paper world.
Reviewed By: shergin
Differential Revision: D14336760
fbshipit-source-id: 1c6a72fa67d5fedbaefb21cd4d7e5d75484f4fae
Summary: Use the codegen for the Slider component with the new `inferfaceOnly` option
Reviewed By: TheSavior
Differential Revision: D14295981
fbshipit-source-id: 0482572892fbcffada43c7c6fbf17e70546300b8
Summary: This diff removes the "isLayoutable" parameter from SchedulerDelegate.schedulerDidRequestPreliminaryViewAllocation. This now can be infered from the shadowView parameter
Reviewed By: shergin
Differential Revision: D14296481
fbshipit-source-id: b200504f9c2bef41f0a70257f1f5a274fbe97cbb
Summary:
Update Props during pre-allocation avoiding re-updating the same props during mounting
MobileLab test showed an improvement of:
MARKETPLACE_YOU_TTI_SUCCESS: -3.34% = 58ms
Reviewed By: shergin
Differential Revision: D14252170
fbshipit-source-id: 1f4e9ad5dcecbc06651fa065135ffeed4892d984
Summary:
We are now generating the native cpp files for ActivityIndicatorView via Buck.
Deleting the hand written files and switching over.
Reviewed By: TheSavior
Differential Revision: D14247446
fbshipit-source-id: 63a6df3254e4184de6c8abb9ea2c89654ad54398
Summary:
This is a back-out of D14214844, we noticed that this regressed TTI for Marketplace You screen running in Fabric
Original commit changeset: b81005f2bf49
Reviewed By: JoshuaGross
Differential Revision: D14247897
fbshipit-source-id: de0cea92b437b2fbcd075f0d6a0066156800e3f0
Summary: The changes allows to get the State object on mounting layer and initiate the updates.
Reviewed By: mdvacca
Differential Revision: D14217185
fbshipit-source-id: 370644e06e350932e93c39adbe46544b071c28b3
Summary: That's bummer that we have to do it, but it's actually reasonable. Files in `core` and `events` depend on each other creating circular dependencies and other similar hard problem.
Reviewed By: mdvacca
Differential Revision: D14195022
fbshipit-source-id: 96a44ae28631cc9ccd7d7de72a94526f9e0dd12a
Summary:
This diff changes the pre-allocation of Images update the props on the ImageViews during the creation of ShadowNodes instead of during rendering.
The purpose of this change is to optimize TTI.
Reviewed By: shergin
Differential Revision: D14214844
fbshipit-source-id: b81005f2bf494f62f421dc24846e1561e13b9a87
Summary:
Not super clean, but not terrible.
Unfortunately this still relies on the old Paper UIManager calling delegate methods to flush the operations queues. This will work for Marketplace You since Paper will be active, but we need to fix this, along with Animated Events which don't work at all yet.
Random aside: it seems like taps are less responsive in fabric vs. paper, at least on iOS. There is a sporadic delay between the touches event coming in nativly to the JS callback invoking the native module function to start the animation - this will need some debugging.
Reviewed By: shergin
Differential Revision: D14143331
fbshipit-source-id: 63a17eaafa1217d77a532a2716d9f886a96fae59
Summary:
CALayer will crash if we pass NaN or Inf values.
It's unclear how to detect this case on cross-platform manner holistically, so we have to do it on the mounting layer as well.
NaN/Inf is a kinda valid result of some math operations. Even if we can (and should) detect (and report early) incorrect (NaN and Inf) values which come from JavaScript side, we sometimes cannot backtrace the sources of a calculation that produced an incorrect/useless result.
Besides that, I will investigate why the crash is actually happening, so we might need to fix something in layout engine. But, it general, we cannot capture all errors like that, so we need to have it here anyway.
Reviewed By: JoshuaGross, mmmulani
Differential Revision: D14126058
fbshipit-source-id: 807e5a223bdef48af9a3b7210803863431e8c507
Summary: Use the new copyright header format used elsewhere in the React Native repository.
Reviewed By: shergin
Differential Revision: D14091706
fbshipit-source-id: b27b8e6bcdf2f3d9402886dbc6b68c305150b7d5
Summary:
Sometimes, when we deal with ImageRequest and ImageResponseObserverCoordinator we subscribe for status (or access the coordinator) without owning an ImageRequest. In those cases, we have to retain the coordinator explicitly.
For those cases, ImageRequest now exposes `ImageResponseObserverCoordinator` as a `std::shared_ptr`.
Eg, concretely in the code, `completionBlock` and `progressBlock` copied a raw pointer to the observer inside which can lead to a crash when ImageRequest is being deallocated before we received an image data.
Reviewed By: JoshuaGross
Differential Revision: D14072079
fbshipit-source-id: e10120bc05bf685e288f7b3d69092714dcd91d43
Summary: Apparently, I haven't modify all files in D14018903. Here is the last (I hope) chunk.
Reviewed By: JoshuaGross
Differential Revision: D14058288
fbshipit-source-id: b21950cdd1aa9aa55b0c72fac0f8b44e4a7d131c
Summary:
Trivial.
If you have troubles with rebasing on top of this revision, run this on your diff:
$ find */*.h */*.mm */*.cpp */*.m -exec clang-format -style=file -i {} \;
Reviewed By: JoshuaGross
Differential Revision: D14018903
fbshipit-source-id: fd0ce2da0e11954e683385402738c701045e727c
Summary:
There is no reason to allocate views ahead of time on the main thread.
There is a chance that this view will not be mounted and we are not saving any time because it's a sequential process anyway (because we are doing it on the main thread). Moreover, the switching context can only slowdown JS execution.
Reviewed By: JoshuaGross
Differential Revision: D14019433
fbshipit-source-id: 83ac05a91e4b70cb382a55d6687752480984404e
Summary: Recycling and dealloc were not implemented at all before for Slider, so I've taken a first stab at it. It's a little more complex than I initially thought, due to things I don't 100% understand about UISlider as well as Fabric, so I've left a TODO note to fix this at some point. We should be aware that view recycling doesn't appear to be working the way I would expect currently though.
Reviewed By: shergin
Differential Revision: D13965475
fbshipit-source-id: fd18a219cead770b63b514fdc868e23214e735b7
Summary: Don't use shared_ptr in this case, it's not needed.
Reviewed By: shergin
Differential Revision: D13965413
fbshipit-source-id: ec98c13f53c7d558a0cb68cea0f97568dd202cd8
Summary: The biggest change is that (1) the image proxy/observer code from the Image component has been generalized, (2) the four image props for the Slider component are fully supported, (3) a handful of props that were ignored or buggy on iOS now perform as expected.
Reviewed By: shergin
Differential Revision: D13954892
fbshipit-source-id: bec8ad3407c39a1cb186d9541a73b509dccc92ce
Summary:
This diff adds performance loggers for Fabric in Android to be able to compare current version or RN with Fabric
This is the summary of Points and Annotations:
- **UIManager_CommitStart**: time that React starts the commit (react tree is ready to start rendering in native)
- **UIManager_LayoutTime**: this is the time it takes to calculate layout in yoga
- **UIManager_FabricFinishTransactionTime**: Time it takes transform "C++ mutationInstructions" into "Java MountItems" and cross boundaries from C++ to Java (including serialization of data) (THIS IS ONLY FABRIC)
- **UIManager_DispatchViewUpdates**: time right before RN moves the mount operations to the Queue that is going to be processed in the next tick UI thread
- **UIManager_BatchRunStart**: time right before the mountItems are going to be process in the UI Thread
- **UIManager_BatchedExecutionTime**: time it took to run batched mountItems (usually layout and prop updates on views)
- **UIManager_NonBatchedExecutionTime**: time it took to run non-batched mountItems (usually creation of views)
Reviewed By: fkgozali
Differential Revision: D13838337
fbshipit-source-id: 0a707619829e7d95ce94d9305ff434d1224afc46
Summary: Folly promises/futures have been replaced by an observer model which keeps track of loading state. This resolves at least one crash that I can no longer repro and simplifies the code a bit (IMO).
Reviewed By: shergin
Differential Revision: D13743393
fbshipit-source-id: 2b650841525db98b2f67add85f2097f24259c6cf
Summary: Objective-C side of the Fabric-compatible slider component for iOS.
Reviewed By: mdvacca
Differential Revision: D13745263
fbshipit-source-id: 647631d6fc86f81a5d4f735c507636ed9c468093
Summary: This diff open sources Fabric Android implementation and it extracts ComponentDescriptorFactory into a function that can be "injected" per application
Reviewed By: shergin
Differential Revision: D13616172
fbshipit-source-id: 7b7a6461216740b5a1ad5ebbead9e37de4570221
Summary: We are now generating the native cpp files for Switch via Buck. Deleting the hand written files and switching over.
Reviewed By: JoshuaGross, mdvacca
Differential Revision: D13666672
fbshipit-source-id: 72cf6f6af9374511f2742f8f0d996fa52e1bff5b
Summary:
shergin mentioned that he'd like to move away from RTTI a bit and use explicit key strings for context container instances rather than relying on the `typeid`, so this does this.
We also fatal with a useful error message if we get a collision, rather than failing silently.
Reviewed By: shergin
Differential Revision: D13384308
fbshipit-source-id: 0b06d7555b082be89e8f130c23e94be99749a7a3
Summary:
`RCTSurfaceHostingProxyRootView` surfaces are still automatically started right after the initialization to match `RCTRootView` interface, but `RCTSurfaceHostingView` must be started explicitly now. Also fixed some internal stuff so start and register are clear and distinct.
Background / initial motivation:
One tricky bit - we render the template as part of init`ing the rootView, so we don't know what the surfaceId will be before hand to register the UITemplate. Two possible solutions:
1) Require start be called explicitly after initializing the rootView, and setup the context in between.
2) Do something like "setUITemplateConfigForNextSurface" before creating the rootView, and have some hook when the surfaceId is assigned that associates the surfaceId with that "next" UITemplate stuff before.
(1) seems a lot cleaner, but it requires ever user of rootView to explicitly call start on it - how do you feel about that? Seems like we could also use that start call to decide if the initial render should be synchronous or not? start vs. startSync?
Reviewed By: mdvacca
Differential Revision: D13372914
fbshipit-source-id: 6db297870610e6c231f8a78c0dd74d584cb64910
Summary: We need a way for different apps to inject dependencies or additional functionality into Fabric - ReactNativeConfig might be a special case, but I think this could clean up it's integration nicely, and I'm using this for a uitemplate cache system so we can use CompactDisk or other storage systems for caching depending on the app.
Reviewed By: mdvacca
Differential Revision: D13407287
fbshipit-source-id: 45481908434e6235850aa4d2d6b2bfb936a23be7
Summary:
So, it does not start itself automatically right after instantiation.
(Classic RCTSurface still kinda start itself automatically but only because start/stop concept is not implemented for this yet.)
Reviewed By: sahrens
Differential Revision: D13461294
fbshipit-source-id: 05430688f69a0d9bf75d03e6d25f02ccd5d3176a
Summary: Each app may provide different impl for its runtime specific behaviors, then Fabric and other new infra can share the same config instance to configure stuffs.
Reviewed By: sahrens
Differential Revision: D13290319
fbshipit-source-id: 30e3eeedc6ff6ef250ed233b27e38cb7c1062b55
Summary:
ShadowView, ShadowViewMutation, and Differentiator were decoupled to separate module.
That enables us to use ShadowView more widely without facing a circular dependency problem.
Reviewed By: mdvacca
Differential Revision: D13205229
fbshipit-source-id: 7373864bf153a7813c2f97edb263a41454ce0b88
Summary:
Previously, we stored a pointer to ShadowNode inside NSAttributedString's attributes to make possible retrieving an EventEmitter associated with some text fragment.
That worked fine besides only one caveat: the internal implementation of NSAttributedString is quite strange and that causes a memory leak. Because of some reason, NSAttributedString does not release stored attributes after own destruction (maybe OS uses some kind of caching).
So, now, instead of storing a strong pointer to ShadowNode inside NSAttributedString, we store a weak pointer to EventEmitter. Storing a weak pointer is okay because a desired lifetime of EventEmitter is guaranteed by LocalData stored inside a View. Storing a weak EventEmitter instead of weak ShadowNode will also help us with migration to ShadowView (we cannot store ShadowView weakly because it's a stack allocated object).
Reviewed By: sahrens
Differential Revision: D13196886
fbshipit-source-id: f8714e4b3709765629d6456edf0c635bf5f7c53b
Summary: Over-retaining a LocalData object inside the View can cause a crash during tearing down JS VM because LocalData can indirectly retain EventEmitter objects which were not properly "disabled".
Reviewed By: sahrens
Differential Revision: D13196887
fbshipit-source-id: 001d9fadf775c89f750c84fe8da9b84d4636631c
Summary: RCTViewComponentView retains an EventEmitter, so we have to clear this up after we recyled the view.
Reviewed By: sahrens
Differential Revision: D13196884
fbshipit-source-id: e9f2e2400be864c5c6177227255012101ed8c4d1
Summary: View recycling can be pretty aggressive and memory intensive, so we can properly react on system memory-pressure notification.
Reviewed By: mdvacca
Differential Revision: D13176278
fbshipit-source-id: 38ea1b27da988aeaaa5db6ac0b94389a0bd2799e
Summary:
Apparently, the previous behavior brings more problems than some *possible-in-the-future* features and flexibility.
The new model allows us to easily implement "nested text" feature.
(We temporary hope the old behavious for Android only for compatibility reasons.)
Reviewed By: mdvacca
Differential Revision: D13176277
fbshipit-source-id: 01f7bfb3c2e70cc89d76ecb78add016ee91cbd63
Summary:
The whole mounting iOS infra now uses `ComponentHandle` instead of `std::string` as a reference to particular `ComponentView` implementation. All changes are pretty straightforward, we use a different thing/type to refer to the particular class; no changes in the logic besides a new `RCTComponentViewFactory` that serves the same role of classes registry as Objective-C runtime served previously.
That has several benefits:
* It should be slightly faster, mostly because we don't need to convert `char *` strings to `std::string` and then to `NSString *`.
* We don't need string-based component-name maps anymore (at least on this layer). We can call classes as we want and it will work because of classes are now explicit about which ShadowNodes they are compatible with.
* Most importantly, it's explicit now! That means that no runtime magic is involved anymore and we can rely on static linting tool now and not be afraid of improper code stripping/overoptimization.
Reviewed By: mdvacca
Differential Revision: D13130760
fbshipit-source-id: aadf70525a1335b96992443abae4da359efdc829
Summary: The new method in the protocol enforces view component classes to expose a component handle of the component that the view component represents. That will allow us to wire up those classes with shadow views in runtime explicitly and in a much more performant way than it is now.
Reviewed By: mdvacca
Differential Revision: D13114663
fbshipit-source-id: 853187d978aab200f85719d9c1d9fea2e3ad4e55
Summary: Future::then taking a value-taking function is deprecated and being deleted. This cleans up a few more callsites.
Reviewed By: yfeldblum, shergin
Differential Revision: D13163277
fbshipit-source-id: 98d1f78c5b34ca34603cc24d79157a4718573576
Summary: That's a temporary change that useful only while we don't have deeper/proper intergration with UIKit's navigation infra.
Reviewed By: mdvacca
Differential Revision: D13104172
fbshipit-source-id: 024e71e04470d56d2c5e9b3f6c3392555ce50b63
Summary:
Apparently, if a gesture recognizer got 'reset', OS does not call `touchesCancelled:` method, so we have to do it manually.
To implement this we have to split `_dispatchTouches:eventType:` into two methods: the first converts Objective-C touches to C++ touches, the second consumes that. We have to do this because during reset we don't have a collection of UIKit touches.
Reviewed By: mdvacca
Differential Revision: D13072807
fbshipit-source-id: 677e2febf9f96dcdfaadfadf5b9ab3893f93e796
Summary: Every C++ engineer (except me several months ago) knows that `operator []` can mutate the collection (Yeah! Don't ask), so this is especially dangerous if your hash function is broken (see the previous diff).
Reviewed By: mdvacca
Differential Revision: D13072805
fbshipit-source-id: 4436a8ff12fb27a57bfb6ee0ff986d7a9a825549
zhongwuzw
Joshua Gross
ericlewis
Valentin Shergin
Spencer Ahrens
Rick Hanlon
David Vacca
Héctor Ramos
Eli White
Doug Russell
Kevin Gozali
Lee Howes