Make 'stretch' aligned children fill cross-axis in containers of undefined size in the cross-axis #171
Conversation
|
Looks good, just needs the suggested cleanups. Please communicate about these changes here in #145 as this is based on his work. |
jsendros
force-pushed
the
stretch-cross-axis
branch
from
cd9fa3b to
77a547a
Compare
|
The reason all the tests change in LayoutCachingTest.java is because they all have roots which are all default styled (flexDirection: column, alignItems: stretch). The width of the root changes, so the width of the leaves must change. |
jsendros
force-pushed
the
stretch-cross-axis
branch
from
77a547a to
601add6
Compare
|
Looks good but you should also land the generated code for each language in the repo. |
|
Never mind, the commit code does have the generated code. One C# test is failing though. |
jsendros
force-pushed
the
stretch-cross-axis
branch
2 times, most recently
from
639bd7b to
f29af8a
Compare
…fined size in the cross-axis Summary: Fetched and rebased facebook#145 and updated it to do 2 things: 1) Make 'stretch' aligned children grow to fill their containers in the cross-axis if the cross-axis size is undefined. The added test might be the best explanation of this. 2) Make sure this doesn’t decrease the frequency of the “simpleStackCross” optimization in a relatively complex sample environment. Test Plan: Added a test: Container with 3 row directioned children and no explicit height set. One child is of height 0, one of 0 height with stretch alignment, and one of 150 height. Expected the 0 height child remains of height 0, stretch grows to 150, and 150 stays the same.
jsendros
force-pushed
the
stretch-cross-axis
branch
from
f29af8a to
3f0f4be
Compare
| @@ -155,7 +155,7 @@ public void testInvalidatesFullTreeWhenParentWidthChanges() | |||
| root.calculateLayout(); | |||
| markLayoutAppliedForTree(root); | |||
|
|
|||
| c0.Height = 200; | |||
| c0.Width = 200; | |||
There was a problem hiding this comment.
This right here... This is why I wish there wasn't code(or test) duplication.
|
Landed in #145 |
| child->layout.position[trailing[crossAxis]] -= getTrailingMargin(child, crossAxis) + | ||
| getRelativePosition(child, crossAxis); | ||
|
|
||
| layoutNode(child, maxWidth, maxHeight, direction); |
There was a problem hiding this comment.
This causes us to call layoutNode recursively a second time, which has side-effects. The bug we're seeing because of this is that items that are centered get their leading dimension incremented twice.
Summary: Fetched and rebased
#145 and updated it to do 2
things:
cross-axis if the cross-axis size is undefined. The added test might be
the best explanation of this.
“simpleStackCross” optimization in a relatively complex sample
environment.
Test Plan: Added a test: Container with 3 row directioned children and
no explicit height set. One child is of height 0, one of 0 height with
stretch alignment, and one of 150 height. Expected the 0 height child
remains of height 0, stretch grows to 150, and 150 stays the same.