Summary:
@public
Moves all Java classes used from C++ to their own header + implementation file. In `YGJNI.cpp`, we only call methods on these objects deriving from `jni::JavaClass`.
This is only a bit of cleanup, no functional change.
Reviewed By: SidharthGuglani
Differential Revision: D14261043
fbshipit-source-id: 2db1d81cb7f56a4cdc24b182b2166e1d7e24ba3c
Summary: `YGConfigContext` was using `global_ref` instances for the config, leading to the config object never being freed. Since we no longer need it after getting rid of cloning, we can make the context a pointer to a `global_ref` to the logger.
Reviewed By: SidharthGuglani
Differential Revision: D14258571
fbshipit-source-id: cce632499839a680eef00a3854f61ab74ae2a87a
Summary:
We set flags to true when setting margin, padding, border, position individually
Doing the same for batching API
Reviewed By: davidaurelio
Differential Revision: D14207550
fbshipit-source-id: ddfdcd5056bea0dd76bd6762f47e90370e26c9e1
Summary:
Added a new API `YGNodeSetLayoutStyleInputs` to pass layout style inputs from java to native code.
All the style inputs are passed in a float array in [key1, key2, value2, key3, value3a, value3b .....] format over JNI layer.
There are three types of style inputs
- do not need any value to be passed along with them like WidthAuto, HeightAuto
- need one value to be passed like Width, Height
- need two values to be passed like Margin, Padding (edge value and actual margin, padding value)
Reviewed By: davidaurelio
Differential Revision: D14166948
fbshipit-source-id: 4bea64d6a429959c3962c87e337914dcd99199fd
Summary:
We have now create a yoga node context which consists of weak reference to java yoga node object and flag which tells us whether margin, padding , border are set or not.
This flag was initially in java layer and we have moved this to native layer as it will help us with enabling the bacthing API for setting style inputs.
Reviewed By: davidaurelio
Differential Revision: D14243378
fbshipit-source-id: fed935ef18c1abf2b07e5d69d9ca79ced51699f2
Summary: We are now not setting flags when we set style inputs margin, padding, border and position on yoga node.
Reviewed By: davidaurelio
Differential Revision: D14224000
fbshipit-source-id: deef4c1ab1a60fbc4909183bc2aa59fa23939d43
Summary:
@public
The cloning features of YogaNode don’t seem to be used. Let’s remove them.
Reviewed By: SidharthGuglani
Differential Revision: D14165624
fbshipit-source-id: 5b710964a4abf1b35f3bcc25b143ffc719a03cec
Summary: "Critical" or "Fast" JNI methods are enticing by their name, but carry dangers that are not trivially visible.
Reviewed By: davidaurelio
Differential Revision: D14184560
fbshipit-source-id: 89ec70f53bb2cb89ff568d8b1fe222ede86c9824
Summary:
@public
If `jni_YGNodeFree` is called while GC is running, the weak reference table lock is held by the GC, leading to deadlock.
Here, we revert the method to being a regular native method, solving that problem.
Reviewed By: SidharthGuglani
Differential Revision: D14184220
fbshipit-source-id: 2882fa10586617cea2df99550a7dd8885376d11e
Summary:
@public
`setBaseLine` was the only place where we used a capitalised *L.* Fixed here.
Reviewed By: SidharthGuglani
Differential Revision: D14152320
fbshipit-source-id: abf54fe7d6088e03775968baa8421c4bf43d6a6e
Summary:
@public
Adds the ability to opt into avoiding global weak JNI refs via `YogaConfig`.
Note that only homogeneous trees are supported, i.e. **mixing weak-ref and non-weak-ref nodes will break!**
Not using JNI refs hopefully will help with avoiding JNI reference table overflows, and will help creating trees on multiple threads, as no lock has to be acquired at any time.
Reviewed By: SidharthGuglani
Differential Revision: D14151037
fbshipit-source-id: 56d94713d39aee080d54be4cb4cdf5e3eccb473a
Summary:
@public
Here, we extract an abstract class from `YogaNode`, in order to allow for experimentation with different implementations.
The reason for not choosing an interface is to keep ABI compatibility for `YogaNode.create()`.
Reviewed By: pasqualeanatriello
Differential Revision: D14044990
fbshipit-source-id: f8eb246338b55f34f0401198c0655abfcb7c9f37
Summary:
@public
Context-aware logging functions are an internal Yoga feature that will be used for Yoga’s JNI code.
It will be possible to specify a context when calculating layout, which will be passed on to baseline and measure functions. This will be a private feature.
Reviewed By: SidharthGuglani
Differential Revision: D14123482
fbshipit-source-id: 8ba3b6c493bf79fe09831f22d2b6da44f09e3d95
Summary:
@public
Makes logging implementation internal to Yoga.
Breaking changes: removed `YGLog` and `YGLogWithConfig`.
The upcoming changes to the JNI layer (removal of weak global refs for each node) requires adding additional parameters to the logging functions that will only be available when calculating layout.
Reviewed By: SidharthGuglani
Differential Revision: D14123390
fbshipit-source-id: 468e4a240c190342868ffbb5f8beb92324cdfdd6
Summary:
@public
Removes `YGNodeInsertSharedChild` / `addSharedChildAt`.
This functionality is unused, and can cause memory leaks.
Reviewed By: SidharthGuglani
Differential Revision: D13711105
fbshipit-source-id: 86206c05393b3f1a497e6b046006f94ead88c6ce
Summary:
@public
Remove ability to configure Yoga to run with/without JNI fast calls on dalvik / art.
This switches to always run with fast calls.
Reviewed By: astreet
Differential Revision: D13144652
fbshipit-source-id: 091aab0cd1290d46346323d3e26a11dd0bb17187
Summary:
@public
Restores the yearless format of the MIT license.
Reviewed By: SidharthGuglani
Differential Revision: D13082510
fbshipit-source-id: f5a849b06652cedf68547d4a7963398b2627915f
Summary:
@public
Currently only parent can tell the layout to align its children based on baseline. But if one of the children is a column or row then basealign does not work as expected.
We have added an api setReferenceBaseline which when set to true would mean that it's baseline would be considered as the reference baseline for parent amongst its siblings. If there are more than one siblings with referenceBaseline set, the first one would be considered.
Reviewed By: davidaurelio
Differential Revision: D12883323
fbshipit-source-id: 19beccfc47d98bb38f81f5b66ba764e83680f821
Summary:
Changes all possible native JNI methods to critical methods.
For now, this only changes Android up and including v7. In order to be prepared for the `CriticalNative` annotation in Android v8, the following restrictions apply:
- Qualifying methods must be static (this is also enforced for Dalvik / Android v4)
- Method implementations can only consume primitive JNI types (`boolean`, jchar`, etc.)
Reviewed By: priteshrnandgaonkar
Differential Revision: D9943868
fbshipit-source-id: 728817eb37822b717fd3daf94cd9f02b42c17db6
Summary:
Moves binding of native methods into a separate native method that can be parameterized.
This will be used to experiment with JNI-related technology.
Reviewed By: priteshrnandgaonkar
Differential Revision: D9943870
fbshipit-source-id: 661f15537d5bbf7a3eef7717e3d99fed2de23904
Summary: This change drops the year from the copyright headers and the LICENSE file.
Reviewed By: yungsters
Differential Revision: D9727774
fbshipit-source-id: df4fc1e4390733fe774b1a160dd41b4a3d83302a
Summary:
@public
This reverts the Yoga/Java storage experiment. I will follow up with any learnings.
Reviewed By: pasqualeanatriello
Differential Revision: D9168405
fbshipit-source-id: fb227fb9353bd4c4e3bebbe9b04eec1132e532e8
Summary:
@public
Adds another version of property storage for `YogaNode`, using `sun.misc.Unsafe`.
Adopts the stub concept from Litho for `Unsafe`, as it is hidden by the Android SDK.
Reviewed By: pasqualeanatriello
Differential Revision: D9140103
fbshipit-source-id: a4b376eca341b724a00f873467ae8bf8eaac69f4
Summary:
- adds more property assignments
- reduces the number of layout roots that exist simultanously
Reviewed By: pasqualeanatriello
Differential Revision: D8989389
fbshipit-source-id: 6a0ac800a4caad61a2f4bf98caa314855b70875f
Summary:
@public
Adds an implementation of `YogaNodeProperties` that sets style properties using a `ByteBuffer` rather than JNI calls.
We hope for a speed improvement.
Reviewed By: pasqualeanatriello
Differential Revision: D9042225
fbshipit-source-id: c7f2b24eaeddd1190755bec85a5034079bd2f492
Summary:
@public
Adds an implementation of `YogaNodeProperties` that accesses style and layout properties using a `ByteBuffer` rather than JNI calls.
We hope for a speed improvement.
This needs further cleanup after experimenting, e.g. to codegen the offsets.
Reviewed By: pasqualeanatriello
Differential Revision: D8911723
fbshipit-source-id: 3c24b57eb545155878896ebb5d64d4553eb6bedc
Summary:
Here we introduce an abstraction over node property storage, in order to experiment with different approaches for Java/C integration.
- interface `YogaNodeProperties` as abstraction
- current JNI code factored into `YogaNodePropertiesJNI.java`
- `YogaNode` delegates all calls, no API changes
Reviewed By: astreet
Differential Revision: D8769448
fbshipit-source-id: e67327ce41fa047a51a986c652b3d59992a510e2
Summary:
Replaces two nested if-blocks with guards. This is intended to help with restructuring this function in follow-ups.
@public
Reviewed By: priteshrnandgaonkar
Differential Revision: D8785659
fbshipit-source-id: 7b9d63e9814b83b999397c016ad67ad348bb0f72
Summary:
This diff refactors the cloning mechanism for YogaNode used from Fabric UI renderer and RN iOS graphs.
Previously, we were cleaning the owner of the child's cloned node inside the C++ implementation of YogaNode. This was a mistake because this modified the last commited YogaTree, causing side effect in RN iOS graphs.
Reviewed By: shergin
Differential Revision: D8672627
fbshipit-source-id: c9902d00690e0361fd58aed84b506c42258bd995
Summary:
Sync fbjni with `libaries/fbjni`.
This includes a hack to fix a deadlock we found for cmake-debug variants only. This still gets us back in sync with other fbjni consumers which is going to save us trouble in the future if Yoga ever wants to make use of a newer feature.
Manual changes made in addition to the hand-crafted CMakeLists and cxx buck rules:
- `jni::isObjectRefType` has SDK lookup disabled.
- `fbjni/` path is changed back to `fb/` to retain compatibility with Yoga.
Reviewed By: priteshrnandgaonkar
Differential Revision: D8531991
fbshipit-source-id: 776f519e2e5f9bea37f55990348f7ed81c4860b4
Summary:
This diff includes the following changes:
1 ) I extended the Java implementation of YogaNode to be able to get a full copy of a YogaNode object without copying the List of children of the original YogaNode. In other words, the new copy of the YogaNode will have the same data of the original YogaNode, but with an empty list of children.
2 ) We created a new method in Yoga.cpp called YGNodeInsertSharedChild. This new method is going to be used by Fabric in order to temporarily share a YogaNode between two "Yoga Trees" (the "current Yoga" tree and a partial "clone of the current Yoga tree"). We exposed this new functionality in the java implementation of Yoga (method addSharedChildAt)
I'm including sebmarkbage for more context.
Reviewed By: emilsjolander
Differential Revision: D7245421
fbshipit-source-id: 72578c8261f29e4a12fc6c72a91f2f891cd58d48
Summary:
In the persistent version of Yoga, a YogaNode can be shared between two YogaTrees, that means that a YogaNode could have more than one Parent at one point in time. That's why the concept of Parent of a YogaNode is not a 1-1 relationship anymore.
This diff changes the semantic of Parent of a YogaNode to Owner of a Yoga Node. CC sebmarkbage and priteshrnandgaonkar for more context.
Technically this diff renames the field YogaNode.parent to YogaNode.owner (and every internal field, Getter and Setter that is related to parent)
Note that as part of this diff I also modified the CSSLayoutDEPRECATED version of Yoga in order to keep compatibility with the C++ implementation.
Reviewed By: priteshrnandgaonkar
Differential Revision: D7352778
fbshipit-source-id: dcf1af5e72bfc3063b5c4bda197d7952a9194768
Summary:
This diff exposes `shouldDiffLayoutWithoutLegacyStretchBehaviour` from YGConfig and `doesLegacyStretchFlagAffectsLayout` from YGLayout to YogaConfig and YogaNode respectively
Also added a positive test case. Didn't find the negative test case example. @[508947467:ianc] or @[759512522:emilsj], can you suggest a negative test case example.
Reviewed By: emilsjolander
Differential Revision: D7272067
fbshipit-source-id: e67e82eb057e4c7124904c715f9dca4dcfea21ea
Summary:
Changed NaN with large number to support `-ffast-math` compiler flag.For `-ffast-math` to work, all floating point numbers should be finite. Reason for not going with `FLT_MAX`, is that, it may cause number overflow during math operations. So thats why I opted for big number smaller than `FLT_MAX`. Earlier we used NaN, while NaN is involved in comparision the comparision operator behaves differently, it always returns false. Also operators like, fmaxf,fminf etc. have wierd beahviours. This diff takes care of those things as far as possible, and all tests are passing.
Running ./instrumentation_tests/run instrumentation_tests/com/facebook/feed/ctacoalescing:ctacoalescing --class AttachmentCallToActionSelectorBenchmarkTest --benchmark --extra-arg iterations=100 shows the perf gain of 13-15%
Reviewed By: emilsjolander
Differential Revision: D6969537
fbshipit-source-id: bdc09eaf703e0d313ca65c25a4fb44c99203d9bf
Summary:
Includes React Native and its dependencies Fresco, Metro, and Yoga. Excludes samples/examples/docs.
find: ^(?:( *)|( *(?:[\*~#]|::))( )? *)?Copyright (?:\(c\) )?(\d{4})\b.+Facebook[\s\S]+?BSD[\s\S]+?(?:this source tree|the same directory)\.$
replace: $1$2$3Copyright (c) $4-present, Facebook, Inc.\n$2\n$1$2$3This source code is licensed under the MIT license found in the\n$1$2$3LICENSE file in the root directory of this source tree.
Reviewed By: TheSavior, yungsters
Differential Revision: D7007050
fbshipit-source-id: 37dd6bf0ffec0923bfc99c260bb330683f35553e
Summary:
This diff exposes the YogaNode clone operation to JNI in order to be able to clone Java YogaNode objects.
The clone method performs a shallow copy of the java YogaNode.
I made YogaNode to implement Cloneable, I know that this might not be a good idea but in this case it simplifies the cloning mechanism. I am open to suggestions.
IMPORTANT NOTES:
- The current implementation IS NOT making a deep copy of the mData instance variable.
- The mParent Java instance variable will reference the parent of the original Java YogaNode, is that ok sebmarkbage?
Reviewed By: priteshrnandgaonkar
Differential Revision: D6935971
fbshipit-source-id: a2008f1eb849b5074585b48699b7de56d5ac90d4
Summary: This diff extends the JNI version of yoga in order to allow Java instances of the YogaConfig class to receive a callback when a Yoga node is cloned.
Reviewed By: priteshrnandgaonkar
Differential Revision: D6918605
fbshipit-source-id: e424c78680c04e21154ebe21405671c4e90f6529
Summary: Expose a function which marks all descendants dirty
Reviewed By: emilsjolander
Differential Revision: D6911869
fbshipit-source-id: e0a3abcf5653f921297edfdca473d83b947cc627
Summary: Previously the logging logic assumed fixed number of characters in the string to be logged. With this diff the logging logic is made dynamic, catering to variable length of the string to be logged
Reviewed By: emilsjolander
Differential Revision: D6784491
fbshipit-source-id: 26e4520a84be355ff992b808297ce7a95b3d09e3
Summary: Moved c implementation of `YGNode` to C++ struct. Not moving to C++ class as the React Classes dependent on `Yoga.h` assume it to be C. Thats why keeping `Yoga.h` C compatible. Sorry for the long diff, didn't thought that it will turn out to be this much big.Will keep an eye on number of lines next time 😉
Reviewed By: emilsjolander
Differential Revision: D6592257
fbshipit-source-id: 641e8b9462ad00731a094511f9f5608b23a6bb21
Summary:
More fine grained dirty marking
Currently a node's dirty flag propagates to the root of the tree ensuring that when any node is invalidated its whole subtree will be re-calculated. This is often times not needed. There are many properties which only effects a node's children and would not need to propagate all the way to the root such as align-items. Also in cases where the style does change layout it may not need to propagate all the way to the root but can often stop at the nearest position: absolute parent.
This change has the potential of greatly improving performance of re-calculating a tree.
This might require adding a second dirty flag named hasDirtyDescendants ensuring that traversal still works even though a parent is not marked as dirty.
Reviewed By: emilsjolander
Differential Revision: D6134754
fbshipit-source-id: bbcfee14058140b946401de756a3f130de0f51cd
