Summary: This change expands the limits to support a greater variety of scenarios.
Reviewed By: PeteTheHeat
Differential Revision: D12911841
fbshipit-source-id: a7c8eb6fece49dfe47b3ada98f55a02b43396ce8
Summary:
This PR increases the speed at which cached images are loaded and displayed on the screen. Images are currently cached in memory using RCTImageCache, but each time they are loaded, a round trip through RCTNetworking happens before RCTImageCache is even checked. This is likely so that RCTNetworking can handle the caching behavior required by the HTTP headers. However, this means that at the very least, images are read from disk each time they're loaded.
This PR makes RCTImageLoader check RCTImageCache _before_ sending a request to RCTNetworking. RCTImageCache stores a bit of information about the response headers so that it can respect Cache-Control fields without needing a roundtrip through RCTNetworking.
Here are a couple of graphs showing improved loading times before this change (blue) and after (red) with SDWebImage (yellow) as a baseline comparison. The increase is most evident when loading especially large (hi-res photo size) images, or loading multiple images at a time.
https://imgur.com/a/cnL47Z0
More performance gains can potentially be had by increasing the size limit of RCTImageCache: https://github.com/facebook/react-native/blob/1a6666a116fd8b9e8637956de2b41a1c315dd470/Libraries/Image/RCTImageCache.m#L39 but this comes at the tradeoff of being more likely to run into OOM crashes.
Pull Request resolved: https://github.com/facebook/react-native/pull/20356
Reviewed By: PeteTheHeat
Differential Revision: D12909121
Pulled By: alsun2001
fbshipit-source-id: 7f5e21928c53d7aa53f293b7f1b4ec5c99b5f0c2
Summary:
If a view inside of an inline view became dirty (e.g. its top/left prop changed), its position would not update on screen. This is because Yoga didn't know the view needed to be relaid out because Yoga's dirty signal didn't propagate all the way up to the root.
The problem is that inline views don't have a parent in the Yoga tree causing Yoga's dirtiness signal propagation to get cut off early. The fix is, when an inline views gets dirty, mark the parent Text's Yoga node as dirty. This will cause Yoga's dirtiness signal to propagate all the way up to the root node.
Yoga has a hook to inform you when your node is marked as dirty: `YGNodeSetDirtiedFunc`. We leverage this to find out when an inline view's Yoga node gets dirtied.
React Native almost handled this case. Everything worked fine as long as the inline view was nested inside of a virtual text node like this:
```
<Text>
<Text>
<InlineView />
</Text>
</Text>
```
However, the bug repros when the inline view is nested in a non-virtual text node:
```
<Text>
<InlineView />
</Text>
```
The fix is to move the special dirtiness propagation logic from `RCTVirtualTextShadowView` to `RCTBaseTextShadowView`.
**Test Plan**
Created an inline view. Tested the following kinds of updates on the inline view's content:
- Moved the content
- Removed the content
- Added the content
Tested this for an inline view that is directly inside of a text node as well as one that is nested under a virtual text node.
Here's the code I used for the inline view that moved its content after 2 seconds:
```
const RN = require('react-native');
const React = require('react');
export default class InlineView extends React.Component {
constructor(props, context) {
super(props, context);
this.state = { posBottom: false };
}
componentDidMount() {
super.componentDidMount && super.componentDidMount();
setTimeout(() => { this.setState({ posBottom: true }); }, 2000);
}
render() {
const pos = this.state.posBottom ? 25 : 0;
const color = this.state.posBottom ? 'pink' : 'green';
return (
<RN.View style={{ width: 50, height: 50, backgroundColor: 'steelblue'}}>
<RN.View style={{ width: 25, height: 25, top: pos, left: pos, backgroundColor: color }} />
</RN.View>
);
}
}
```
**Release Notes**
[IOS] [BUGFIX] [Text] - Fix case where content of inline views didn't get relaid out
Adam Comella
Microsoft Corp.
Pull Request resolved: https://github.com/facebook/react-native/pull/21968
Differential Revision: D12873795
Pulled By: shergin
fbshipit-source-id: bbc9f5d3ef25063b0015cec8c4aaf2e41ecd60a8
Summary:
When we use product flavor (e.g. develop/production) with Android Gradle Plugin 3.2,
javascript bundle is not copied due to that merged asset path point to incorrect location.
related #21408
<!--
Thank you for sending the PR! We appreciate you spending the time to work on these changes.
Help us understand your motivation by explaining why you decided to make this change.
If this PR fixes an issue, type "Fixes #issueNumber" to automatically close the issue when the PR is merged.
_Pull requests that expand test coverage are more likely to get reviewed. Add a test case whenever possible!_
-->
Pull Request resolved: https://github.com/facebook/react-native/pull/21782
Differential Revision: D12854056
Pulled By: hramos
fbshipit-source-id: d1e6a395e762cc4a4f133977d54d0f469fa66b8c
Summary: This function sometimes causes an "Unable to find node on an unmounted component" crash during pagination for reasons that still need to be investigated; in the meanwhile, wrap this in a try-catch block to mitigate the crash.
Reviewed By: sahrens
Differential Revision: D12829971
fbshipit-source-id: bc9fe5b9b8c03430ff890bfbb27c39aa270c9eb7
Summary:
Adds the displayName prop to `View` and `Text` components. Because these now use `React.forwardRef`, they were showing as `Component` instead of their actual names.
Thanks to ljharb for helping to pinpoint the source of the issue!
Fixes#21937
Pull Request resolved: https://github.com/facebook/react-native/pull/21950
Differential Revision: D12827060
Pulled By: TheSavior
fbshipit-source-id: d812cae14d53ad821ab5873e737db63ad1a989e3
Summary: When running a trace, reduce the noise from the `__DEV__`-only `ForwardRef(View)` elements.
Reviewed By: ejanzer
Differential Revision: D9445865
fbshipit-source-id: 7cfe87bab6dd62d3800d2ca239724b5063c55c89
Summary:
Fix build errors when path to $NODE_BINARY contains spaces:
error: Can't find '/Path/With Spaces/To/node' binary to build React Native bundle
Why would $NODE_BINARY contain spaces? In my case, I am using sentry-cli which wraps the NODE_BINARY in it's own executable. The local path to the project, and thus the $NODE_BINARY, contains a space on my GoCD build agent.
'/Users/go/Library/Application Support/Go Agent/pipelines/my-ios-app/node_modules/sentry/cli/sentry-cli
See https://github.com/getsentry/react-native-sentry/issues/484https://github.com/getsentry/react-native-sentry/issues/389
Pull Request resolved: https://github.com/facebook/react-native/pull/21383
Differential Revision: D10851141
Pulled By: yungsters
fbshipit-source-id: f46853ac8b57957864e0d1a76b8513403223fccb
Summary:
This diff fixes an issue that caused the problem with `regeneratorRuntime` last Friday (more info: https://fb.facebook.com/groups/frontendsupport/permalink/2427883350560427/).
The root issue is that both `InitializeCore` and `FBInitializeCore` are included in the same bundle, this fix just prevents the bundle from being invalid once this happens..
*copied from: https://our.intern.facebook.com/intern/diff/D10444264/?transaction_id=595485237532887*
The way that `regeneratorRuntime` is polyfilled is not correct:
```
polyfillGlobal('regeneratorRuntime', () => {
// The require just sets up the global, so make sure when we first
// invoke it the global does not exist
delete global.regeneratorRuntime;
require('regenerator-runtime/runtime');
return global.regeneratorRuntime;
});
```
Since a `require`d module is only evaluated once (no matter how many times it's required), defining (and calling) a getter for `global.regeneratorRuntime` twice will end up storing `undefined` to `global.regeneratorRuntime`.
There were no issues before this diff because the ordering of requires made things work by coincidence, e.g the following code will work:
```
// Set up regenerator for the first time
polyfillGlobal('regeneratorRuntime', () => {
delete global.regeneratorRuntime;
require('regenerator-runtime/runtime');
return global.regeneratorRuntime;
});
// Set up regenerator for the second time. This will just override the previous getter (which has not even got executed so
// the `regenerator-runtime/runtime` module has not been evaluated yet.
polyfillGlobal('regeneratorRuntime', () => {
// The require just sets up the global, so make sure when we first
// invoke it the global does not exist
delete global.regeneratorRuntime;
require('regenerator-runtime/runtime');
return global.regeneratorRuntime;
});
// Now access regenerator
global.regeneratorRuntime;
```
But the following code won't work:
```
// Set up regenerator for the first time
polyfillGlobal('regeneratorRuntime', () => {
delete global.regeneratorRuntime;
require('regenerator-runtime/runtime');
return global.regeneratorRuntime;
});
// Access regenerator. This will cause the previous getter to be called, and the `regenerator-runtime/runtime` module will get evaluated.
// Here, `global.regeneratorRuntime` will have a correct value.
global.regeneratorRuntime;
// Set up regenerator for the second time. This will define a new getter for `global.regeneratorRuntime`, which will delete `delete global.regeneratorRuntime`
// and return undefined (note that `require('regenerator-runtime/runtime');` is a noop since the module has been already evaluated).
polyfillGlobal('regeneratorRuntime', () => {
// The require just sets up the global, so make sure when we first
// invoke it the global does not exist
delete global.regeneratorRuntime;
require('regenerator-runtime/runtime');
return global.regeneratorRuntime;
});
```
Reviewed By: fromcelticpark
Differential Revision: D10483975
fbshipit-source-id: 5b3ef6e11c4fc4f79e3857c1ade9e7bc2beb6a39
Summary: This diff fixes a crash that happens when a component throws an exception that contains a null message
Reviewed By: achen1
Differential Revision: D10429661
fbshipit-source-id: 9faed36906844b51f5a3dc2b3cbc80ede6b93457
Summary: Just a small diff to remove some unused parameters in some methods of the ReactInstnaceManager class
Reviewed By: achen1
Differential Revision: D10011168
fbshipit-source-id: 2247ecd01bfc63c27837c6abad4bff1b4e590ac0
Summary:
@public
If you call NetInfo.getCurrentConnectivity multiple times in succession, we'll create a bunch of callbacks but lose them in the ether.
With this fix, we'll unschedule them before creating a new one, which should resolve some crashes we're seeing.
Reviewed By: PeteTheHeat
Differential Revision: D10409486
fbshipit-source-id: 6065b09fa626f7f06aed9bf0e278c0a6a6169f58
Summary:
release method of local_ref and global_ref doesn't call deallocator, in fact, it leaves the caller responsible for deletion of the reference, while otherwise the reference is released on scope left.
Fixes#18292.
Pull Request resolved: https://github.com/facebook/react-native/pull/20913
Differential Revision: D9616237
Pulled By: hramos
fbshipit-source-id: 021aa3e4f039e6b7a98da3e4224c1ee49d5a4921
Summary: ListView is deprecated and SwipeableListView uses ListView. Thus, it is deprecated as well.
Reviewed By: RSNara
Differential Revision: D10437408
fbshipit-source-id: a08391d5b099e74b6ec179cd940ac404b2e702f4
Summary:
[RN] Relax the requirement that lazy module cannot be initialized on the main thread
I tried to understand the D5364734 that intoduced this, and I am not sure, but belive that asserting here is too strict. If you have a module that you want to lazily initialize, and module does not demand the main queue, it should be just a warning if you run on the main queue, not necessarily an error.
Reviewed By: mmmulani
Differential Revision: D10429880
fbshipit-source-id: 018c211d45b98dd8c552bf0289fe517d05e56d47
Summary:
Fixes#19069
The --simulator option for the run-ios command now can take an optional
iOS version between parenthesis to further match the desired simulator.
This is useful if you have installed simulators for different iOS
versions and you want to run the app in an especific one. Example:
react-native run-ios --simulator "iPhone 6s (9.3)"
Thank you for sending the PR! We appreciate you spending the time to work on these changes.
Help us understand your motivation by explaining why you decided to make this change.
Updated tests for the findMatchingSimulator function to include test cases specifying iOS version, and tested on the command line in my app to make sure it has the expected behavior.
[CLI] [ENHANCEMENT] [{/runIOS/findMatchingSimulator.js}] - run-ios command with the --simulator option now allows specifying the iOS version to run an specific simulator if you have multiple versions of the simulator installed. Example: `react-native run-ios --simulator "iPhone 6s (9.3)"`.
Pull Request resolved: https://github.com/facebook/react-native/pull/19079
Differential Revision: D10432487
Pulled By: hramos
fbshipit-source-id: efa50d798b79d83bfe357ee17967a56c7c003bee
Summary:
With this, we send the correct x position when using center or right aligned text. In order to accomplish this though, we have to pass the text alignment into the Layout object that we create.
Also update RNTester to allow us to try different alignments.
Reviewed By: sahrens
Differential Revision: D10316494
fbshipit-source-id: 11c7d2a59e636528f12211168acb46f16b54a126
Summary:
@public
Just movin' stuff around.
This shouldn't render any actual metrics yet on Android, need native changes for that.
Reviewed By: achen1
Differential Revision: D9584669
fbshipit-source-id: 4b6169b14d1f2053258191f67e1f361a4b714a8e
Summary:
@public
As we're doing in D9440914 (OSS 64a52532fe), send text metrics in an onTextLayout callback. These can be used by surrounding views for doing complicated layout like:
- displaying a cursor at the end of text
- vertical centering using capheight-baseline
This right now isn't very performant but is only done when `onTextLayout` is set. I plan to optimize it with a capheight and xheight cache in a follow up diff.
Reviewed By: achen1
Differential Revision: D9585613
fbshipit-source-id: aa20535b8371d5aecf15822d66a0d973c9a7eeda
Summary: This adds a callback for <Text> to get metrics about the rendered text. It's divided by line but that could be changed to "fragments" (which makes more sense for multi-lingual). Right now by line is convenient as you frequently want to know where the first and last line end (though we could make this work with fragments I suppose).
Reviewed By: shergin
Differential Revision: D9440914
fbshipit-source-id: bb011bb7a52438380d3f604ffe7019b98c18d978
Summary:
(cordova-node-)xcode 1.0.0 includes a fix for #13160 plus a few more
fixes. This should increase robustness in Xcode project handling
overall.
Fixes#13160.
-------------------------
Thank you for sending the PR! We appreciate you spending the time to work on these changes.
Help us understand your motivation by explaining why you decided to make this change.
If this PR fixes an issue, type "Fixes #issueNumber" to automatically close the issue when the PR is merged.
_Pull requests that expand test coverage are more likely to get reviewed. Add a test case whenever possible!_
Pull Request resolved: https://github.com/facebook/react-native/pull/21766
Reviewed By: codytwinton
Differential Revision: D10384245
Pulled By: hramos
fbshipit-source-id: 488156be67cfc2d99ca81d7cb82747bab35984a7
Summary:
1. The user inserts a character ('0') at index 0. Because the range matches 0, 0, predictedText is set to that character that was inserted.
2. In textInputDidChange, it discovers a mismatch between the rendered text ('1234') and predicted text ('0')
3. This triggers textInputShouldChangeTextInRange to be called again with the 'new' text that it thinks was just added ('1234')
4. It goes to insert this text, but runs into the maxLength limit, so it gets truncated and then inserted.
(I'm not totally sure why only happens if maxLength is set - I need to look into that.)
One fix for this is to just get rid of the range check, but that'll regress #18374. I decided to just check and see if the rendered text is empty instead of checking the range where text could be inserted, since that seems like it should properly handle both cases.
Reviewed By: shergin
Differential Revision: D10392176
fbshipit-source-id: 84fb3b6cac9b0aa25b3c1a127d43f9cdc5a1c6a8
Summary:
Original PR: https://github.com/facebook/react-native/pull/21340
The original PR fixed a bug with border radius on Android 7.0, but caused a new bug where no `ReactART` components would render if they were nested within a View with border radius set.
Differential Revision: D10409352
fbshipit-source-id: 0b4e36ae9d4c3bc2f4884a43909f8efdeeb41b74
Summary:
Related to https://github.com/facebook/react-native/pull/21488
Disclaimer: I made this PR.
I think there's some requestAnimationFrame events that are not cleared on unmount because of bad use of `splice` method.
- All flow tests succeed.
- RNTester: iOS (this change should only affect iOS because calculateChildFrames is iOS only)
Show perf monitor, show ListView* screen, start scrolling. UI frame Rate is used at the beginning. When scrolling there is no drop in FPS rate.
- TODO: I'll write a load test for ListView
[GENERAL] [ENHANCEMENT] [ListView.js] - rm TimerMixin
Pull Request resolved: https://github.com/facebook/react-native/pull/21802
Differential Revision: D10391812
Pulled By: RSNara
fbshipit-source-id: 49f0b0a4641ec29bcb4cc04bd3bafb42b3842b69
Summary:
Related to #21485
- Remove TimerMixin from ListView
- All flow tests succeed.
- RNTester: <ListView> iOS (this change should only affect iOS because calculateChildFrames is iOS only)
Show perf monitor, show ListView* screen, start scrolling. UI frame Rate is used at the beginning. When scrolling there is no drop in FPS rate.
TODO: I think a load test would be more relevant:
- Update props multiple times and scroll
[GENERAL] [ENHANCEMENT] [ListView.js] - rm TimerMixin
Pull Request resolved: https://github.com/facebook/react-native/pull/21488
Differential Revision: D10219088
Pulled By: RSNara
fbshipit-source-id: 946e4fc1319324c5bf4947a2060b18bebb6fc493
Summary: Currently the pull to refresh icon on browse feed is super cut off on Android. Expose the progressViewOffset prop from FBPullToRefresh to support offsetting the PTR component to make the component more visible.
Reviewed By: yungsters
Differential Revision: D10274679
fbshipit-source-id: 1735c4d2d98523ccc3d1ec3733465028ae33df7b
Summary:
DynamicFromMap internally uses SimplePool object to recycle dynamic prop objects. But the pool is not multi-thread safe. Currently the most used dynamic props are size props such as left, paddingVertical, marginTop and so on. These props are only accessed from the layout thread so the pool works fine. If a dynamic prop is needed in UI thread, then the two threads can access the same pool object and cause random errors. This PR make the pool object thread local to avoid synchronization. After this change there are two pool objects created in the process.
Tested in official Airbnb app after updating accessibilityComponentType to be dynamic.
Once this is merged, I'll send another PR to support "disabled" state in `accessibilityComponentType`.
[ANDROID] [BUGFIX] [DynamicFromMap] - Fix a crash caused by dynamic props.
Pull Request resolved: https://github.com/facebook/react-native/pull/17842
Differential Revision: D10374238
Pulled By: hramos
fbshipit-source-id: 7ebf89c5abf06bd5fb43b205348ba4dc7e19517d
