ngAnimate - ngIf - Element not updated before exit animation

[colllin]

I made a demo:
http://jsfiddle.net/colllin/es569bqL/

The purple box has an ng-if attribute. When that expression evaluates to false, the element is removed, but neither the ng-class attribute nor the inner content is recomputed prior to the animation.

There is no issue when the ng-if expression switches to true and the element is added. The element is generated with the correct classes and content.

You can see that the text in the box should change at the same time as the "exit" animation is triggered. It never does.

For testing the ng-class attribute...
In this demo there are 4 possible combinations to move the box in and then out.
These 2 combos succeed:
In Left -> Out Left
In Right -> Out Right
These 2 combos fail:
In Left -> Out Right
In Right -> Out Left
This is because the ng-class doesn't get re-evaluated, which would have updated the CSS class that specifies the animation direction.

Is this a bug, or is there a nicer way to achieve this?

---

There's also a weird artifact if you quickly animate the box out and then back in -- there are 2 boxes rendered on-screen during the overlap of the exit/enter animations. Should I create a separate issue about that?

[colllin]

I found a temporary workaround. I can modify the $scope variables that trigger ng-class and content changes first, then wrap a $timeout around the part that triggers the ng-if.

Demo: http://jsfiddle.net/colllin/hpju0gyb/

Of course, the inner ng-if attributes are tied to the same $scope variable as the outer ng-if, so those still aren't working correctly. There's probably no workaround for that (other than creating a separate variable).

[matsko]

Hey @colllin. I think this issue may be related to: #10067. I know how to fix it. The fix itself will be a rather large refactoring, but I intend of fixing it this week.

[colllin]

@matsko Beautiful. Thank you so much. Look forward to it.

Not sure if it matters, but regarding the regression tag -- I first discovered the bug on 1.2.x and then switched to 1.3.x to see if it was fixed. So if it's a regression, it's an old one.

[petebacondarwin]

@matsko - are you able to take a look at your big refactoring this week?

[matsko]

This is still an issue in 1.4 even after the refactor. I think that we need to do something about the scope being destroyed too early.

[Narretz]

If it's related to scopes being destroyed too early, I assume it's similar to #12080