Using `ColumnScope.AnimatedVisibility` gives this ...
# compose
s
Using
ColumnScope.AnimatedVisibility
gives this annoying problem of having to do double check for the state, since the content lambda does not give the current state. Like:
Copy code
AnimatedVisibility(model != null) {
  if (model != null) {
    ActualComposable(model)
    Spacer(Modifier.height(20.dp))
  }
}
Since I am not sure if doing
!!
in there is safe, as the old content may still be animating out, and it will call
!!
on the null item to try and render the item as it’s leaving. What do you folks usually do for this? Do I just have to do this double null check or is there something I can do instead, some API I may be missing?
I can see one can do something like
Copy code
AnimatedContent(
  targetState = model,
  label = "AnimatedContent",
  transitionSpec = {
    ContentTransform(
      targetContentEnter = fadeIn() + expandVertically(),
      initialContentExit = fadeOut() + shrinkVertically(),
    )
  },
) { model ->
  if (model != null) {
    ActualComposable()
  }
}
to use AnimatedContent which does in fact give me the current item inside the lamvda, but then I need to specify the ContentTransform myself and so on, which I do like that AnimatedVisibility on a Column puts a sane default there.
e
s
Thanks, I +1 that one
a
That wouldn’t apply in this case though, it’s definitely possible that the content inside
AnimatedVisibility
is getting called, even if
model
is
null
. This is a case where the strict data flow of Compose makes the required code a bit more explicit for what’s actually happening. Let’s say that the animation for
AnimatedVisibility
takes 10 minutes to animate out (the world’s slowest remove animation) Even if you clear the model, something somewhere still needs to hold onto it to continue to render the UI for it for the next 10 minutes. The user could be doing other actions, other things could be changing, and the content inside the
AnimatedVisibility
is still being composed and updated.
s
Hmm that’s a very valid point, and the example of having the animation take 10 minutes really makes this obvious. I really wonder what I should be doing there then. My content in there takes in this entire model which has quite a lot of state, so I can’t “fake” it just while it’s animating out. In theory, what I want in there is for the last known state to be saved as a screenshot, and have that thing animate out, as I don’t need it to be interactive, update and so on. But I know I can’t quite do that. I guess there isn’t a straight-forward solution to this as I see it hmm
a
The main documentation for
AnimatedVisibility
has a sample for managing a list where items can be removed with an animation out which might be interesting to take a look at: https://developer.android.com/reference/kotlin/androidx/compose/animation/package-summary#AnimatedVisibility(androidx[…]lin.String,kotlin.Function1) The core idea is that an item is only completely removed from everywhere once it has finished animated out.
In theory, what I want in there is for the last known state to be saved as a screenshot
What you could maybe sort of do is “cache” with
remember
the item to still render with, and keep that around even if the item is deleted higher up. Or have two lists: the “logical” list, which are the items that will be in the list if all animations resolve, and the “display” list, which might have additional items that are in the process of animating out.
s
Yeah that solution from the link is interesting, but would definitely not want to introduce all this extra code to make this animation look a bit better, at least not in what I am working on rn which doesn’t need the animation to look that perfect. I like the idea of temporarily remembering the item for a bit longer, but again I’m sure if I look back at this in 1 month from now I’ll start wondering what the hell I was doing with this extra logic in the UI 😅 Thing is with all these, in a more animation-rich screen I’d probably justify doing it, but now I was trying to get an easy win by animating some portion of a column out instead of having it just disappear. But this “easy win” doesn’t end up being that easy after all unfortunately.
Thanks a lot for helping me by explaining the options I have btw, I appreciate it a lot, despite me saying that I am most likely going to decide not to use any of those at least right now. I’ll probably need it in the future 😊
Hey you know what, after a good night’s sleep, the solution became super obvious. I just pass a nullable model to
ActualComposable(model)
, and I handle the null case in there. In my case I can basically show most of what I was showing already (it’s a card) with the same height and all, so the animation won’t be jumpy. This model was needed for some of the text only in this case (“Foo” in the video), so the bad thing is just that that text shows nothing instantaneously when it starts leaving instead of also animating out, but I get this animation without adding any extra logic just for this animation like remebering stuff and so on. So that’s much better than what I first had anyway. Thanks so much for helping me out!
TIL
expandVertically
and
shrinkVertically
can take in
expandFrom = Alignment.CenterVertically
and
shrinkTowards = Alignment.CenterVertically
. Every time I do anything with the animation APIs it’s always so pleasant to work with to do these little touches that take like no effort, it’s insane.
e
if I look back at this in 1 month from now I’ll start wondering what the hell I was doing
Thats what comments are for 😅
s
Thats what comments are for
The comments will start lying before you know it. They’re sneaky like that 😅