How does Circuit handle Effect(one-time events, e.g. Toast, screen transitions)?

[The-Kan]

Circuit is returning the State value of the screen from the Presenter. If the State contains an Effect, it seems that the Effect may be executed repeatedly rather than once in a situation such as recomposition (such as a Toast appearing multiple times). In that case, how do you handle things like Effects?

[ZacSweers]

You essentially acknowledged it in your description - if it's in state, it'll repeat. As such, effects aren't part of state :).

There's two things you're asking about — effects and transitions.

Effects are implemented like other compose effects, we even have an effects extension artifact with one for toasts here.

Transitions are done with Compose's native transition APIs and we have docs here.

[The-Kan]

Thanks for the comment. The reason I was wondering is that in the MVI pattern, you can pass state and effects separately to the ui. I wonder if Circuit can also handle effects separately.

[ZacSweers]

There is no separate built-in API for sending data from the presenter to the UI like you're asking for, no. You can only pass state.

[The-Kan]

Yes, thank you. I will handle the toast and screen movement as you told me.

[easyhooon]

#1876

I also asked a similar question before, but at that time I didn't know about the existence of toastEffect(only knew about ImpressionEffect for logging and analytics from the documentation), so I handled it by distinguishing a separate SideEffect state and processing it as state. Now I think I can use toastEffect instead. Thank you for the answer!

I'm also curious if snackbar (and other one-time events as well) can be handled by implementing them like toastEffect.

[easyhooon]

Well, it seems toastEffect has some constraints. As shown in the example, when a button click triggers a toast call (where UI is the initiator), there's no problem.

However, when calling toast as a result of an API call (where Presenter is the initiator), we can't trigger toastEffect from within the presenter, so it seems we still need to introduce separate state to handle this anyway.