Just a cautionary tale, when using `AnimatedContent` (and its variants); if the `targetState` change...

# compose

[Zoltan Demant]

Just a cautionary tale, when using AnimatedContent (and its variants); if the targetState changes at all, even if it doesnt result in a new animation running, there will be additional recompositions as a result (I was seeing 2-9, instead of 1). I learned this the hard way, Id normally specify what I wanted to render as the targetState and use transition.contentKey to key it correctly; stripping out the stuff not directly related to the animation made an immense difference. Ill share some more details in the thread for anyone interested, I dont think this is a bug, but rather me misunderstanding how it works - hopefully it can save you from doing the same mistake as I did!

[Zoltan Demant]

What I wound up doing is strictly using targetState for things that instruct which animation should be used.

AnimatedContent(
    targetState = size,
    transitionSpec = Transition,
    content = { size ->
        // Content for size
    }
)

And if theres any other data associated with size in this case, Id store it elsewhere outside of the targetState... e.g. in a remembered value, etc.

[Illustrator]

You mean you won't use the size given from the lamda?

[Zoltan Demant]

Ill still use the size in the lambda, but consider that theres a ton of other data accompanying that specific size (which really doesnt matter to the animation itself); thats what Im referring to as not being directly included in the AnimatedContent at all.

[Chris Sinco [G]]

cc @Doris Liu