ios animating already transformed view - ios

Question

I have a collection view with a collection of images (which are views). The layout causes transformations to occur on every visible item. The item at the center is both zoomed and translated while the others are just translated.
The point is, the layout is setting the layer's transform property for each item.
Later, when the user touches an item, I want to animate the item using a keyframe animation.
The behavior I am experiencing is that the item seems to revert back to its untransformed state, the animation occurs on its untransformed state and then the item's layout transformation returns.
Why?
I've tried using the view's transform property, the backing layer's affine transform property, and the backing layer's 3D transform property all with this same behavior. The keyframe animation uses 3D transforms.
I clearly have a gap in my understanding of transforms and animations, though I have seen this question posed for platforms other than iOS without answers.
Layout transformation code (in a flow layout subclass):
// Calculate the distance from the center of the visible rect to the center of the attributes.
// Then normalize it so we can compare them all. This way, all items further away than the
// active get the same transform.
let distanceFromVisibleRectToItem = CGRectGetMidX(visibleRect) - attributes.center.x
let normalizedDistance = distanceFromVisibleRectToItem / ACTIVE_DISTANCE
let isLeft = distanceFromVisibleRectToItem > 0
var transform = CATransform3DIdentity
var maskAlpha: CGFloat = 0.0
if (abs(distanceFromVisibleRectToItem) < ACTIVE_DISTANCE) {
// We're close enough to apply the transform in relation to how far away from the center we are.
transform = CATransform3DTranslate(transform,
(isLeft ? -FLOW_OFFSET : FLOW_OFFSET) * abs(distanceFromVisibleRectToItem/TRANSLATE_DISTANCE),
0, (1 - abs(normalizedDistance)) * 40000 + (isLeft ? 200 : 0))
// Set the zoom factor.
let zoom = 1 + ZOOM_FACTOR * (1 - abs(normalizedDistance))
transform = CATransform3DScale(transform, zoom, zoom, 1)
attributes.zIndex = 1
let ratioToCenter = (ACTIVE_DISTANCE - abs(distanceFromVisibleRectToItem)) / ACTIVE_DISTANCE
// Interpolate between 0.0f and INACTIVE_GREY_VALUE
maskAlpha = INACTIVE_GREY_VALUE + ratioToCenter * (-INACTIVE_GREY_VALUE)
} else {
// We're too far away - just apply a standard perspective transform.
transform = CATransform3DTranslate(transform, isLeft ? -FLOW_OFFSET : FLOW_OFFSET, 0, 0)
attributes.zIndex = 0
maskAlpha = INACTIVE_GREY_VALUE
}
attributes.transform3D = transform
Animation code (note this is a Swift extension to the UIView class):
func bounceView(completion: (() -> Void)? = nil) {
var animation = bounceAnimation(frame.height)
animation.delegate = AnimationDelegate(completion: completion)
layer.addAnimation(animation, forKey: "bounce")
}
func bounceAnimation(itemHeight: CGFloat) -> CAKeyframeAnimation {
let factors: [CGFloat] = [0, 32, 60, 83, 100, 114, 124, 128, 128, 124, 114, 100, 83, 60, 32,
0, 24, 42, 54, 62, 64, 62, 54, 42, 24, 0, 18, 28, 32, 28, 18, 0]
var transforms: [AnyObject] = [NSValue(CATransform3D: self.layer.transform)]
for factor in factors {
let positionOffset = factor/256.0 * itemHeight
let transform = CATransform3DMakeTranslation(0, -positionOffset, 0)
transforms.append(NSValue(CATransform3D: transform))
}
let animation = CAKeyframeAnimation(keyPath: "transform")
animation.repeatCount = 1
animation.duration = CFTimeInterval(factors.count)/30.0
animation.fillMode = kCAFillModeForwards
animation.values = transforms
animation.removedOnCompletion = true // final stage is equal to starting stage
animation.autoreverses = false
return animation;
}
Note: I should state more simply that I want this process to begin and end in the collection view layout transformed state. I want the animation to occur on the layout transformed state as well (i.e. no reverting to the untransformed state at all).

Answer 1

It seems that in your animations you are not syncing your model with your presentation layer.
When you animate with Core Animation (CABasicAnimation, for instance) it only changes the presentation layer (the thing you see on the screen) but in fact the state your layers have is different from the visible one.
(You can read more about that in this objc.io issue)
So basically to fix this you need to update your model in your CAAnimation. You need to set a from and toValue, and then, after you set those, you update your transform so it can match the final state of your presentation layer.
EDIT
Now that we have some code I can be concrete
To sync the model layer to the presentation layer you need to set your model to the final state before adding the animation:
var animation = bounceAnimation(frame.height)
animation.delegate = AnimationDelegate(completion: completion)
layer.transform = CATransform3DMakeTranslation(0, 0, 0)
layer.addAnimation(animation, forKey: "bounce")
David Rönnqvist has a good rant about this issue in this gist.

Answer 2

SOLUTION:
The help Tiago Almeida offered was very useful to get me thinking about what the animation was actually doing (how it was actually working). My basic assumption that the animation would begin with the current model layer transform was incorrect. As Tiago points out, the presentation layer is a separate and distinct layer from the model layer.
Once I got that through my thick head, I realized that I had to manually concatenate the current model transform with the transform being built for the animation frame and viola! Things work as expected.
The updated animation code:
var transforms: [AnyObject] = []
for factor in factors {
let positionOffset = factor/256.0 * itemHeight
var transform = CATransform3DMakeTranslation(0, -positionOffset, 0)
transform = CATransform3DConcat(self.layer.transform, transform)
transforms.append(NSValue(CATransform3D: transform))
}