Is there any performance or concurrency difference in using a MutableSharedFlow (with replay 1, buff...

# coroutines

[Daniele Segato]

Is there any performance or concurrency difference in using a MutableSharedFlow (with replay 1, buffer 1 and drop oldest) over a MutableStateFlow?

MutableSharedFlow<Something?>(
    replay = 1,
    extraBufferCapacity = 1,
    onBufferOverflow = BufferOverflow.DROP_OLDEST
)

vs

MutableStateFlow<Something?>(null)

[Dominaezzz]

If you mean, buffer 0, then state flow might perform ever so slightly better but I don't think it's significant enough to let that dictate your decision.

[Daniele Segato]

I've updated my question

[Dominaezzz]

Oh, I'm pretty sure there's a behaviour difference there. (State flow is still better)

[Daniele Segato]

I'm trying to understand when I should use a MutableSharedFlow and why :)

[Nick Allen]

To clarify:
MutableSharedFlow maintains a buffer of size replay + extraBufferCapacity (it's "extra" for a reason".
MutableStateFlow has additional behavior in that it discards duplicates.
A MutableStateFlow is equivalent to MutableSharedFlow(replay = 1, extraBufferCapacity = 0, onBufferOverflow = BufferOverflow.DROP_OLDEST) , where you always read from the shared flow with sharedFlow.distinctUntilChanged().

[Nick Allen]

While MutableStateFlow is good for managing state that can change, MutableSharedFlow is better suited for broadcasting event objects, where collectors are essentially event listeners. If something happens twice, you don't necessarily want the second event to be thrown away. The "event" in this case could be a sealed class of different event types, or could just be Unit to act as a simple signal.

[Daniele Segato]

Thanks @Nick Allen very clear.