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:
GCC internally aliases isnan and a dozen other math.h primitives to their `__builtin_` counterparts, which compile straight to assembler. Clang does not do this, so calls to `isnan` compile to a function call. The Android libc does not define `isnan` because it assumes all application code compiled `__builtin_isnan` and has no need for the function `isnan`.
This is a temporary kludge: Starting with NDK r16, the math.h header has appropriate `#defines` to alias primitives to their `__builtin_` counterparts. Upon upgrade to NDK r16, this can be reverted.
Reviewed By: passy
Differential Revision: D6527499
fbshipit-source-id: 65d9e43c7e7d973a6c92e9863ebe469b35a24ac8
Summary:
Use clang-5.0 -Oz as default NDK compiler for android.
It does 3 good things:
* brings us into compliance with deprecation of gcc for android NDKs since r13
* yields slightly smaller native code, on the order of many-10s-to-small-100s of KB
* slight improvement in startup performance, on the order of 10s of ms
Reviewed By: mzlee
Differential Revision: D6155682
fbshipit-source-id: 2f64e742f4dc44be171274c2ad0a41fb1e7079d7
Summary:
ComponentKit requires all size geometry provided to be rounded to pixel grid. This ensures
that layout will not truncate provided sizes.
Reviewed By: cuva
Differential Revision: D6466875
fbshipit-source-id: 62a480411e7d2354a6879917c42bbf0ee96cdadd
Summary:
Fix min/max percentage constraints on measured nodes. Currently we passed in the available size instead of the parent size.
Fixes#611
Closes https://github.com/facebook/yoga/pull/649
Reviewed By: priteshrnandgaonkar
Differential Revision: D6408019
Pulled By: emilsjolander
fbshipit-source-id: 18be5056dbc0dc179970ec231ab8b4e2cdba65c5
Summary:
Suppresses the unused warnings on ```YGDefaultLog```. It uses a more generic macro which simply casts the variable to ```(void)```. This is the simples approach to make mutiple different compilers happy. Fixes#650
Closes https://github.com/facebook/yoga/pull/651
Differential Revision: D6407999
Pulled By: emilsjolander
fbshipit-source-id: 19fd78dd8b84eafdbb48875dd003d506a98c4807
Summary:
If you have a measurable node and set ```marign-left: auto``` + ```align-item:stretch``` on it, it won't get measured and they get a width/height of ```-(nan)```. This change fixes that behaviour. Fixes#644.
Closes https://github.com/facebook/yoga/pull/645
Differential Revision: D6413512
Pulled By: emilsjolander
fbshipit-source-id: 755febeb33bb0d4520ca6b3c28d56ac333e4a14d
Summary:
Adds new ```space-evenly``` for ```justify-content```.
Also adds a typofix in one of the other justify-content tests.
Fixes#657
Closes https://github.com/facebook/yoga/pull/658
Differential Revision: D6407996
Pulled By: emilsjolander
fbshipit-source-id: cc837409e1345624b4bd72c31e25fe68dcb0f6a3
Summary:
This fixes shrinking of elements which are in a non stretch alignment, but keeps the legacy stretch behaviour in place. Additionally this adds a testcase for ```useLegacyStretchBehaviour```
Fixes#633
Closes https://github.com/facebook/yoga/pull/635
Differential Revision: D6408037
Pulled By: emilsjolander
fbshipit-source-id: 377ab0308dd3a75a148a0af31ab5eb3ffb5b5d83
Summary:
If we have a values already set to undefined and set it to undefined again, we invalidate the layout. This change takes this case into account and keeps the layout valid.
Fixes#630
Closes https://github.com/facebook/yoga/pull/648
Differential Revision: D6408013
Pulled By: emilsjolander
fbshipit-source-id: dc2a848d84d3de9f4650fac9e41d7c8169446406
Summary:
Expected to be able to call `YGNodeLayoutGetMargin(node, YGEdgeEnd)`, but instead, the program aborts with `"Cannot get layout properties of multi-edge shorthands"`.
This bug seems to incorrectly prevent properties from YGEdgeEnd for all Layout properties.
Closes https://github.com/facebook/yoga/pull/632
Differential Revision: D6408060
Pulled By: emilsjolander
fbshipit-source-id: 4ab3b2ffb2f1bb6fd3a27f780caf0123abcdb230
Summary:
The min/max inner width shouldn't take the margins into account.
Adds a test for both cases.
Fixes#664
Closes https://github.com/facebook/yoga/pull/665
Differential Revision: D6407982
Pulled By: emilsjolander
fbshipit-source-id: ffa549a06f802263e3b8488e90756aa3f722d52d
Summary: Refactored the `YGNodeToString` function from c to c++.
Reviewed By: emilsjolander
Differential Revision: D6397372
fbshipit-source-id: 79e701efe7f19db6dac1aea6328ebf0ac84a7ac3
Summary: Changed the extensions of c files to cpp and made the project build successfully
Reviewed By: gkassabli
Differential Revision: D6271299
fbshipit-source-id: 66c0e54ccf019d72d1fd0b4d117826e4e84fdc89
Summary: Removed YGSetMemoryFuncs, as it is not used internally and barely externally.
Reviewed By: emilsjolander
Differential Revision: D6374147
fbshipit-source-id: 8c896dce23571268a1a52df5cc72486af9d66df7
Summary: We'll need these libraries for rainbow, and none of them exist in fbcode.
Reviewed By: mzlee
Differential Revision: D5897801
fbshipit-source-id: c379b76fef2ab204555ceded0834efc6ba932f90
Summary: YGRoundValueToPixelGrid does not handle float point values correctly. It causes layout bugs in arm 32 devices. The way values are rounded to pixel grid is the following: if value is close to the left pixel - discard error and set it equal to left pixel. If value is close to right pixel - set it equal to the right pixel. Otherwise if value is closer to the left - set it equal to the left, closer to the right - set it equal to the right. The problem is when values are close to the middle. Old implementation could produce either left or right boundaries. This patch moves all values that are close to the middle to right boundary. This way results are consistent.
Reviewed By: emilsjolander
Differential Revision: D6202484
fbshipit-source-id: bb80235452208caec388775574cc4f07bfd649c4
Summary: Yoga has an edge case that cause layout to return NaN. This happened when we used absolute position with percentage and infinite constraint in that dimension. This diff fixes that case to default to border+padding
Reviewed By: emilsjolander
Differential Revision: D6199731
fbshipit-source-id: f40ddf00614b2f507848fb35f348a9dfef14b323
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
Summary:
This is meant to show a possible route format for a persistent form of Yoga. Where previous layouts can remain intact while still taking advantage of incremental layout by reusing previous subtrees.
```c
YGNodeRef YGNodeClone(const YGNodeRef node);
```
The core of this functionality is a new API to clone an existing node. This makes a new detached node with all the same values as the previous one. Conceptually this makes the original node "frozen" from that point on. It's now immutable. (This is not yet enforced at runtime in this PR but something we should add.)
Since the original is frozen, we reuse the children set from the original node. Their parent pointers still point back to the original tree though.
The cloned node is still mutable. It can have its styles updated, and nodes can be inserted or deleted. If an insertion/deletion happens on a cloned node whose children were reused, it'll first shallow clone its children automatically.
As a convenience I also added an API to clear all children:
```c
void YGNodeRemoveAllChildren(const YGNodeRef node);
```
During insert/delete, or as a result of layout a set of reused children may need to be first cloned. A kind of copy-on-write. When that happens, the host may want to respond. E.g. by updating the `context` such as by cloning any wrapper objects and attaching them to the new node.
```c
typedef void (*YGNodeClonedFunc)(YGNodeRef oldNode,
YGNodeRef newNode,
YGNodeRef parent,
int childIndex);
void YGConfigSetNodeClonedFunc(YGConfigRef config,
YGNodeClonedFunc callback);
```
This PR doesn't change any existing semantics for trees that are not first cloned.
It's possible for a single node to exist in two trees at once and be used by multiple threads. Therefore it's not safe to recursively free a whole tree when you use persistence. To solve this, any user of the library has to manually manage ref counting or tracing GC. E.g. by replicating the tree structure in a wrapper.
In a follow up we could consider moving ref counting into Yoga.
Closes https://github.com/facebook/yoga/pull/636
Reviewed By: emilsjolander
Differential Revision: D5941921
Pulled By: sebmarkbage
fbshipit-source-id: c8e93421824c112d09c4773bed4e3141b6491ccf
Summary:
The Android logger adds newline automatically, which makes the log ugly on Android.
To avoid that, wrap the android logger, print to a memory buffer. Call __android_log_write() when '\n' is detected.
Reviewed By: emilsjolander
Differential Revision: D5842889
fbshipit-source-id: 8ecc829dc43614934c47fd3a68317259e403325a
Summary:
This fixes shrinking of elements which are in a non stretch alignment
Fixes#633
Closes https://github.com/facebook/yoga/pull/634
Differential Revision: D5874862
Pulled By: emilsjolander
fbshipit-source-id: 1426aa6b60f6ba42c2be702e6f24cea935ab7acb
Summary:
Add a new test named YGZeroOutLayoutRecursivlyTest, which verifies that padding and margin of a child node should be set to 0 when the display is set to none.
Modify YGZeroOutLayoutRecursivly function, use memset to clear the layout struct of the node.
Reviewed By: emilsjolander
Differential Revision: D5842351
fbshipit-source-id: d2c2b6f829588a200525b7ccbdb63003cb4dcc4e
Summary:
This PR fixes a declaration mismatch for `YGNodeCanUseCachedMeasurement` where the last argument is declared non `const` in `.h` and `const` in `.c`.
Additionally it uses explicit `float` for fraction calculation do avoid usage of `double` assignment.
Closes https://github.com/facebook/yoga/pull/607
Differential Revision: D5677931
Pulled By: emilsjolander
fbshipit-source-id: 502da957089e4439ed956987ff8dec10bd033ba3
Summary: At least one compiler seems to care about it.
Reviewed By: emilsjolander
Differential Revision: D5675518
fbshipit-source-id: 72a6e208263dde0b6bb46a78fedb2796d0e0a600
Summary:
@public
== Before ==
- Aspect ratio would do its best to fit within it's parent constraints
- Aspect ratio would prioritize `alignItems: stretch` over other sizing properties.
== After ==
- Aspect ratio is allowed to make a node grow past its parent constraints. This matches many other aspects of flexbox where parent constraints are not treated as hard constraints but rather as suggestions.
- Aspect ratio only takes `alignItems: stretch` into account if no other size definition is defined. This matches the interaction of other properties with `alignItems: stretch`.
== Updating your code ==
**You probably don't need to do anything** but in case something does break in your product it should be as easy as adding `{width: '100%', height: '100%', flexShrink: 1}` to the style declaring the `aspectRatio`.
Reviewed By: gkassabli
Differential Revision: D5639187
fbshipit-source-id: 603e8fcc3373f0b7f2461da2dad1625ab59dcb19
Summary: There was an uncovered edge case where number close to the whole was forced to round down because it was considered non-whole and had forced flooring. This diff covers that + adds a bunch of test cases to cover rounding function
Reviewed By: emilsjolander
Differential Revision: D5465632
fbshipit-source-id: 57e11092a97eba5dd76daad15fa8619535ff9c1b
Summary: Yoga had a bug in Rounding calculation that caused text nodes dimensions to be rounded up even when the dimensions didn't need rounding
Reviewed By: emilsjolander
Differential Revision: D5406211
fbshipit-source-id: df1d54ed0805dfc3abbd8f0ceae30f6d8c26d61a
Summary:
This fixes the case where we change the layout, so that it doesn't overflow anymore.
This also improves the readability by using `|=` instead of referencing the value twice.
Closes https://github.com/facebook/yoga/pull/587
Differential Revision: D5388657
Pulled By: emilsjolander
fbshipit-source-id: ce1b1ded1feed7314a2c16bf695f62b866c19ea0
Summary:
If we have a fractional measure output which matches the subpixel rounding factor, we still should round both dimension into the same direction.
Fixesfacebook/yoga#580.
Closes https://github.com/facebook/yoga/pull/583
Reviewed By: marco-cova
Differential Revision: D5274212
Pulled By: emilsjolander
fbshipit-source-id: 1febf9194210437ab77f91319d10d4da9b284b79
Summary:
We need to provide Yoga clients with means of detecting overflow in flexbox layout. This information can be used later to identify a non-overflowing layout variant among others.
Flexbox layout considered overflown if (this is *not* an exhaustive list):
- if any child node overflows
- [no-wrap] total flex basis of all child nodes is greater than available space in parent and there are no flexible children
- [no-wrap] after flexing there is still not enough space to layout all child nodes
Reviewed By: gkassabli
Differential Revision: D5336645
fbshipit-source-id: c4f87d1754d7bac848e8d347b31d619393b94d2c
Summary:
The only thing I found in the spec for this change is the following. Not exactly sure if this is the thing this PR is about:
> For each flex item, subtract its outer flex base size from its max-content contribution size. If that result is not zero, divide it by (if the result was positive) its **flex grow factor floored at 1** or (if the result was negative) by its scaled flex shrink factor, having **floored the flex shrink factor at 1**. This is the item’s max-content flex fraction.
But at least it seems a required change.
Fixesfacebook/yoga#566
Closes https://github.com/facebook/yoga/pull/572
Differential Revision: D5264388
Pulled By: emilsjolander
fbshipit-source-id: 0004d1c3b9bad070a98cd6766c1adc06a54475f8
Summary: If the width has an exact dimension, while the left and right edges have a decimal part, that means the decimal part is equal and we can round both down instead of rounding left down and right up. Same for top and bottom.
Reviewed By: emilsjolander
Differential Revision:
D5209073
Tags: accept2ship
fbshipit-source-id: a3a6a43767aa707ebfa5eee62a83adcdd88d7ce6
Summary:
This is a fix on top of 56b10fc. It takes the case into account were you have `wrap-reverse` and `align-items: flex-end` set.
Closes https://github.com/facebook/yoga/pull/568
Differential Revision: D5155521
Pulled By: emilsjolander
fbshipit-source-id: 7e5fcfa2fbb48b6c6279da46cc648a071ff2b079
Summary: Variable name suggest it does not constrain to min dimension but it did indeed. Changing that to match variable name. This required updating code to use fully constraint value again. gkassabli this seems wrong, do you know what is going on?
Reviewed By: gkassabli
Differential Revision: D5144146
fbshipit-source-id: a5d711f8ad7049b0d96ee0f0f3eef1165ab3b4e8
Summary: Previously pointScaleFactor was stored as a divider that opened way to accumulating difference. For instance in ScreenScale=3, pointScaleFactor = 0.3333343 (0.000001 error). When used for width = 300 that error was multiplied by the number of times pointScale contained in width (300 *3 = 900) and we had almost 0.001 error accumulated. With this change Yoga will avoid such issue
Reviewed By: shergin
Differential Revision: D5137923
fbshipit-source-id: 652b59bc3da3f35ee93ffa3695936f623298a023
Summary: There is a case when measure() function will be called with negative width/height because of margin/padding. This diff fixes that scenario
Reviewed By: emilsjolander
Differential Revision: D5111534
fbshipit-source-id: 99f09b85f0c6a0e5dec89a26baba8f9d560100da
Summary:
1, Change bottom to be based� on height of parent.
2, Respect margin value when layout with right/bottom.
Closes https://github.com/facebook/yoga/pull/552
Differential Revision: D5102090
Pulled By: emilsjolander
fbshipit-source-id: dca291413ffc2027d7628f4c8b8eeeb0b4823bc2
Summary: We should always try to make root node as small as possible, while previously this wasn't functioning this way
Reviewed By: emilsjolander
Differential Revision: D5071164
fbshipit-source-id: b8afef42477d0ed87d0c9fcfd26349e0a0babd6e
Summary: When config was added to YGNodeCanUseCachedMeasurement we didn't account for default 0 scale factor. This diff fixes that
Reviewed By: emilsjolander
Differential Revision: D5061016
fbshipit-source-id: 907ef987bd0ead29cf02f6945a3f03d4ffb58a2e
Summary: We want to be able to use YGConfig to account for possible rounding of width/height
Reviewed By: emilsjolander
Differential Revision: D5059560
fbshipit-source-id: d729e991758a8c668a4b373105b71337961875cd
