Summary: public
The `bridge.modules` dictionary provides access to all native modules, but this API requires that every module is initialized in advance so that any module can be accessed.
This diff introduces a better API that will allow modules to be initialized lazily as they are needed, and deprecates `bridge.modules` (modules that use it will still work, but should be rewritten to use `bridge.moduleClasses` or `-[bridge moduleForName/Class:` instead.
The rules are now as follows:
* Any module that overrides `init` or `setBridge:` will be initialized on the main thread when the bridge is created
* Any module that implements `constantsToExport:` will be initialized later when the config is exported (the module itself will be initialized on a background queue, but `constantsToExport:` will still be called on the main thread.
* All other modules will be initialized lazily when a method is first called on them.
These rules may seem slightly arcane, but they have the advantage of not violating any assumptions that may have been made by existing code - any module written under the original assumption that it would be initialized synchronously on the main thread when the bridge is created should still function exactly the same, but modules that avoid overriding `init` or `setBridge:` will now be loaded lazily.
I've rewritten most of the standard modules to take advantage of this new lazy loading, with the following results:
Out of the 65 modules included in UIExplorer:
* 16 are initialized on the main thread when the bridge is created
* A further 8 are initialized when the config is exported to JS
* The remaining 41 will be initialized lazily on-demand
Reviewed By: jspahrsummers
Differential Revision: D2677695
fb-gh-sync-id: 507ae7e9fd6b563e89292c7371767c978e928f33
Summary: public
We were calling constantsToExport twice for every ViewManager, and including two copies of the values in __fbBatchedBridgeConfig. This diff removes the copy from UIManager and then puts it back on the JS side.
Reviewed By: tadeuzagallo
Differential Revision: D2665625
fb-gh-sync-id: 147ec4bfb404835e3875964476ba233d619c28aa
Summary: public
In iOS < 9, inserting a nil object into NSMutableDictionary crashes. It is valid for come components to return a nil shadowView (e.g. ART nodes), and this was crashing on iOS 8.
Reviewed By: jspahrsummers
Differential Revision: D2658309
fb-gh-sync-id: 7abf9273708cc03c3b6307b69ba11c016b471fbe
Summary: This queue processes layout and user interface updates, so it should have as high a quality-of-service/priority as possible.
public
Reviewed By: javache
Differential Revision: D2641837
fb-gh-sync-id: 934686f7969b43101af183148d67ff7be4bdf660
Summary: public
Rename it to `RCT_PROFILE_(BEGIN|END)_EVENT` to make it clearer that it's a macro,
since it has special behaviours.
Reviewed By: jspahrsummers
Differential Revision: D2631542
fb-gh-sync-id: 629c139462c4aa3582f719b14482017d13676e33
Summary: public
There were some old markers that are now automatically inject and now are no longer necessary (+ one that was missing an end call :( ))
Reviewed By: javache
Differential Revision: D2625901
fb-gh-sync-id: 4c4c9d6b4e8e2b4bdb9c64fde01000b0ca2e9f47
Summary: public
Add RCTFatal for reporting fatal runtime conditions. This centralizes failure handling to one function and allows you to customize how they should be handled. RCTFatal will be logged to the console and as a redbox and will also be triggered by fatal exceptions coming from RCTExceptionsManager.
Note that there is no RCTLogFatal, since just logging the fatal condition does not allow us to handle it consistently.
Reviewed By: nicklockwood
Differential Revision: D2615490
fb-gh-sync-id: 7d8e134419e10a8fb549297054ad955db3f6bee0
Summary: public
Added lightweight genarics annotations to make the code more readable and help the compiler catch bugs.
Fixed some type bugs and improved bridge validation in a few places.
Reviewed By: javache
Differential Revision: D2600189
fb-gh-sync-id: f81e22f2cdc107bf8d0b15deec6d5b83aacc5b56
Summary: public
The UIManager had a lock around the enqueued ui blocks, but now all the operations
should happen on the shadow thread, so I added assertions to it and removed the
locks.
Reviewed By: nicklockwood
Differential Revision: D2605760
fb-gh-sync-id: e1bc649f759502e7e9fd059932e0cba38dba05bf
Summary: There is no point in dispatching to main thread if there is nothing to do there.
This place gets called basically any time a repeating js timer fires, which doesn't imply UI changes (although usually that's why people setup timers).
Combined with previous diffs that makes us not generate empty blocks (nil instead), this could be minor perf win in some rare cases.
This also changes semantic of `reactBridgeDidFinishTransaction` call a bit. Previously it was done no matter if UI has changed or not.
I think it should be safe, since seems like callees really care only about views being laid out.
Depends on D2571166. (not strictly speaking)
public
Reviewed By: jspahrsummers, nicklockwood
Differential Revision: D2571188
fb-gh-sync-id: 02d52e4615475072c3c27226e67c431a667ec990
Summary: Same as in previous diffs. Gets us into a better place to know if we really have UI updates and it's marginally more efficient.
Depends on D2571143. (not really)
public
Reviewed By: nicklockwood
Differential Revision: D2571166
fb-gh-sync-id: e8f34521ec2e12156a49f1cd655e92df1db34fca
Summary: Previously `_bridgeTransactionListeners` were informed about `reactBridgeDidFinishTransaction` inside of one of the UI blocks.
That seems pretty arbitrary, doesn't really mean a "transaction" is really over (assuming transaction means all UI updates) and even when that block does nothing we still need to call these listeners, since there could be other UI blocks generated somewhere else!
So I've moved this call to a place that seemed better (=after all UI blocks are done), since all listeners are interested in knowing when layout has happened.
public
Reviewed By: nicklockwood
Differential Revision: D2571122
fb-gh-sync-id: 62be03ebc4353d6f6318c9765079b87b07483be2
Summary: When you reload and create a new bridge, one of the things that happens during setup is that the RCTAccessibilityManager fires a notification. The old bridge would receive this notification from the new bridge's RCTAccessibilityManager, which we don't want, especially because the two are running on different shadow queues.
I believe this led to a gnarly crash in NSConcreteTextStorage because RCTMeasure in RCTShadowText.m was getting called for the old RCTText (getting destroyed) from a notification fired from the new shadow queue. The fix is for the UIManager to handle notifications only from its bridge's RCTAccessibilityManager. See #2001 for the kinds of crashes we were seeing.
Closes https://github.com/facebook/react-native/pull/3279
Reviewed By: @svcscm
Differential Revision: D2521652
Pulled By: @nicklockwood
fb-gh-sync-id: a4ffe3ef8304667727e573e2a2e8b716e1d2f3e1
Summary:
We currently wait until after views have been updated on the main thread before sending layout events. This means that any code that relies on those events to update the UI will lag the atual layout by at least one frame.
This changes the RCTUIManager to send the event immediately after layout has occured on the shadow thread. This noticably improves the respinsiveness of the layout example in UIExplorer, which now updates the dimension labels immediately instead of waiting until after the layout animation has completed.
Summary:
Currently, the system for mapping JS event handlers to blocks is quite clean on the JS side, but is clunky on the native side. The event property is passed as a boolean, which can then be checked by the native side, and if true, the native side is supposed to send an event via the event dispatcher.
This diff adds the facility to declare the property as a block instead. This means that the event side can simply call the block, and it will automatically send the event. Because the blocks for bubbling and direct events are named differently, we can also use this to generate the event registration data and get rid of the arrays of event names.
The name of the event is inferred from the property name, which means that the property for an event called "load" must be called `onLoad` or the mapping won't work. This can be optionally remapped to a different property name on the view itself if necessary, e.g.
RCT_REMAP_VIEW_PROPERTY(onLoad, loadEventBlock, RCTDirectEventBlock)
If you don't want to use this mechanism then for now it is still possible to declare the property as a BOOL instead and use the old mechanism (this approach is now deprecated however, and may eventually be removed altogether).
Summary:
Our events all follow a common pattern, so there's no good reason why the configuration should be so verbose. This diff eliminates that redundancy, and gives us the freedom to simplify the underlying mechanism in future without further churning the call sites.
Summary:
Moved the view creation & property binding logic out of RCTUIManager into a separate RCTComponentData class - this follows the pattern used with the bridge.
I've also updated the property binding to use pre-allocated blocks for setting the values, which is more efficient than the previous system that re-contructed the selectors each time it was called. This should improve view update performance significantly.
Summary:
Dynamic Text Sizes for Text component.
Text gains new prop - allowFontScaling (false by default).
There is also AccessibilityManager module that allows you to tune multipliers per each content size category.
Summary:
The bridge implementation on React Android does not currently support boxed numeric/boolean types (the equivalent of NSNumber arguments on iOS), nor does Java support Objective-C's nil messaging system that transparently casts nil to zero, false, etc for primitive types.
To avoid platform incompatibilities, we now treat all primitive arguments as non-nullable rather than silently converting NSNull -> nil -> 0/false.
We also now enforce that NSNumber * objects must be explicitly marked as `nonnull` (this restriction may be lifted in future if/when Android supports boxed numbers).
Other object types are still assumed to be nullable unless specifically annotated with `nonnull`.
Summary:
RCTNetworkImageView and RCTStaticImage had significant overlap in functionality, but each had a different subset of features and bugs.
This diff merges most of the functionality of RCTNetworkImageView into RCTStaticImage, eliminating some bugs in the former, such as constant redrawing when properties were changed.
I've also removed the onLoadAbort event for now (as it wasn't implemented), and renamed the other events to match the web specs for `<img>` and XHMLHttpRequest. The API is essentially what Adobe proposed here: http://blogs.adobe.com/webplatform/2012/01/13/html5-image-progress-events/
The following features have not yet been ported from RCTNetworkImageView:
- Background color compositing. It's not clear that this adds much value and it increases memory consumption, etc.
- Image request cancelling when images are removed from view. Again, it's not clear if this is a huge benefit, but if it is it should be combined with other optimisations, such as unloading offscreen images.
(Note that this only affects the open source fork. For now, internal apps will still use FBNetworkImageView for remote images.)
Summary:
These are the changes needed for full interop with the (as yet unreleased) new
version of React Devtools.
- the on-device inspector is minimized when devtools is open
- devtools highlight -> device and device touch -> devtools select works
- editing react native styles :)
Summary:
Dynamic Text Sizes for Text component.
Text gains new prop - allowFontScaling (true by default).
There is also AccessibilityManager module that allows you to tune multipliers per each content size category, but predefined multipliers are there.
This could potentially break some apps so please test carefully.
Summary:
Remove layout-only views. Works by checking properties against a list of known properties that only affect layout. The `RCTShadowView` hierarchy still has a 1:1 correlation with the JS nodes.
This works by adjusting the tags and indices in `manageChildren`. For example, if JS told us to insert tag 1 at index 0 and tag 1 is layout-only with children whose tags are 2 and 3, we adjust it so we insert tags 2 and 3 at indices 0 and 1. This keeps changes out of `RCTView` and `RCTScrollView`. In order to simplify this logic, view moves are now processed as view removals followed by additions. A move from index 0 to 1 is recorded as a removal of view at indices 0 and 1 and an insertion of tags 1 and 2 at indices 0 and 1. Of course, the remaining indices have to be offset to take account for this.
The `collapsible` attribute is a bit of a hack to force `RCTScrollView` to always have one child. This was easier than rethinking out the logic there, but we could change this later.
Nick Lockwood
Pieter De Baets
Justin Spahr-Summers
Tadeu Zagallo
Pawel Sienkowski
Martin Kralik
Ryan Olson
James Ide
Matt Mahoney
Vladislav Alexeev
Alex Akers
Spencer Ahrens
Kevin Gozali
Jared Forsyth
Param Aggarwal