[ASCollectionView] Fix index space translation of Flow Layout Delegate methods. #467
Conversation
…e methods. This includes a few other cleanups, including overflow of signed integer indices.
Source/ASCollectionView.mm
Outdated
|
|
||
| for (ASCollectionElement *element in self.dataController.pendingMap.itemElements) { | ||
| ASCellNode *node = element.node; | ||
| if (node.shouldUseUIKitCell) { |
There was a problem hiding this comment.
-
It's probably inconsequential, but maybe we should just ignore calls where the layout isn't a flow layout, rather than using size 0?
-
Instead of building a giant array – I don't really like the
itemElementsproperty and may propose removing it – it's better to enumerate the map directly andif (element.supplementaryElementKind) { continue; }Note to self: add aUICollectionElementCategory elementCategoryproperty to ASCollectionElement to simplify this distinction. -
Shouldn't we also refetch header/footer reference sizes here? Meshes well with 2
There was a problem hiding this comment.
@Adlai-Holler some merge conflicts have arisen, and I'm afraid my time availability is nearing a local minimum. :( Do you think this PR is acceptable to land as-is if the conflicts are fixed?
- I think it's better to have the 0 default, but indeed maybe no one will ever know :)
- The iteration suggestion sounds like an improvement for sure. I'm not sure I can commit to doing it only because I would have to get some test cases running to exercise the change.
- This seems potentially important, and indeed might motivate me to do Test PR #2.
Could you give me an Accept if you are comfortable with this, and then if I do 2 or 3, I can land without waiting?
There was a problem hiding this comment.
@Adlai-Holler It took me a while, but I finally got around to implementing your advice — and went quite a bit further. Thanks for suggesting these things, as they are improvements.
In particular, I was intentionally skipping over handling the update of supplementary element sizes, which is not great. That's now fixed and the code is better factored.
If you or @nguyenhuy can give me a sanity check review soon, I'd appreciate that - would like to merge this in pretty soon.
| } | ||
| return ASFlowLayoutDefault(l, minimumInteritemSpacing, 10.0); // Default is documented as 10.0 | ||
| } | ||
|
|
||
| - (CGFloat)collectionView:(UICollectionView *)cv layout:(UICollectionViewLayout *)l | ||
| minimumLineSpacingForSectionAtIndex:(NSInteger)section | ||
| minimumLineSpacingForSectionAtIndex:(NSInteger)section |
| #define ASDisplayNodeCAssertPositiveReal(description, num) ASDisplayNodeCAssert(num >= 0 && num <= CGFLOAT_MAX, @"%@ must be a real positive integer.", description) | ||
| #define ASDisplayNodeCAssertInfOrPositiveReal(description, num) ASDisplayNodeCAssert(isinf(num) || (num >= 0 && num <= CGFLOAT_MAX), @"%@ must be infinite or a real positive integer.", description) | ||
| #define ASDisplayNodeCAssertPositiveReal(description, num) ASDisplayNodeCAssert(num >= 0 && num <= CGFLOAT_MAX, @"%@ must be a real positive integer: %f.", description, (CGFloat)num) | ||
| #define ASDisplayNodeCAssertInfOrPositiveReal(description, num) ASDisplayNodeCAssert(isinf(num) || (num >= 0 && num <= CGFLOAT_MAX), @"%@ must be infinite or a real positive integer: %f.", description, (CGFloat)num) |
|
🚫 CI failed with log |
…nsure delegate invalidation re-fetches supplementary sizes too.
appleguy
force-pushed
the
ASCV_Passthrough
branch
from
3e4da91
to
e2a9ffb
Compare
…eForUIKitCellWithKind:atIndexPath: method.
|
Hmm, this erroneous test failure is striking again. ASCollectionViewTests |
|
Thanks for reviewing and merging! |
…e methods. (TextureGroup#467) * [ASCollectionView] Fix index space translation of Flow Layout Delegate methods. This includes a few other cleanups, including overflow of signed integer indices. * [ASCollectionView] Improve code sharing of UIKit size method calls; ensure delegate invalidation re-fetches supplementary sizes too. * [ASCollectionView] Final method ordering and doc-comment for new _sizeForUIKitCellWithKind:atIndexPath: method.
This includes a few other cleanups, including overflow of signed integer indices.
It will take a couple PRs to fully sync both upstream and downstream in this case, but I am happy to integrate feedback along the way. Just a heads up that it may be easier to do in subsequent PRs, depending on the importance of changes (stylistic vs functional).