can anybody using coroutine 1.4.0-M1 confirm that a ```MutableSharedFlow(replay = 0)``` doesn't emit...

# coroutines

[nitrog42]

can anybody using coroutine 1.4.0-M1 confirm that a

MutableSharedFlow(replay = 0)

doesn't emit values ? 😕 the same code works if replay = 1 (but I don't want to replay first value each time 😢 )

[Marc Knaup]

Works for me. Prints “hello”.

[nitrog42]

ok I just find out it's because I use tryEmit 😢

[nitrog42]

ahhh I think it needs a buffer

[nitrog42]

ok that's it, just needs a buffer value, good to know

[Marc Knaup]

Hmm. I don’t think it should require a buffer

[nitrog42]

here's my sample

[nitrog42]

fun main() {
    val mSharedFlow = MutableSharedFlow<Int>(extraBufferCapacity = 10)
    val sharedFlow: SharedFlow<Int> = mSharedFlow.asSharedFlow()


    val scope = CoroutineScope(Dispatchers.Default)
    val job = scope.launch {
        sharedFlow.collect {
            println("got $it")
            if (it == 0) {
                cancel()
            }
        }
    }
    runBlocking {
        mSharedFlow.emit(1)
        mSharedFlow.emit(2)
        mSharedFlow.emit(0)
    }
}

[nitrog42]

if you use tryEmit, with a buffer of 0:

[nitrog42]

nothing is printed 😄

[nitrog42]

with a buffer of one :
got 1

[nitrog42]

it works if you use delay between the tryEmit

[Marc Knaup]

Yeah according to the code it requires either buffer size > 0 or replay size > 0.
The latter is confusing

[nitrog42]

of course if I change the bufferOverflow policy to drop it works (with either the latest or oldest, it actually logic now I get the buffer part

[Marc Knaup]

I think it cannot know whether it would suspend in that case so it just assumes that it can’t.

[nitrog42]

val mSharedFlow = MutableSharedFlow<Int>(extraBufferCapacity = 1, onBufferOverflow = BufferOverflow.DROP_LATEST)

[Marc Knaup]

There are many options. What do you want to achieve?

[nitrog42]

I was just playing with it, I just wanted a simple equivalent of a PublishSubject (if you know Rx)

[Marc Knaup]

Isn’t that like StateFlow and replays?

[nitrog42]

a PublishSubject has no replay, and stateflow unfortunatelly reemit the first value, but sharedflow works exactly like I want now

[Marc Knaup]

So you have no initial value either?

[nitrog42]

yes
it is used to transmit events usually (edited) 
things you want to handle only once like errors

[Marc Knaup]

So SharedFlow with no replay makes most sense. What should happen on backpressure?

[nitrog42]

I usually don't need backpressure but here we can often drop oldest

[nitrog42]

depends on the case, right no wi was just trying to display the latest error, but I can just use emit which will work too

[Marc Knaup]

tryEmit with a DROP_LATEST buffer should be fine.

A shared flow configured with a BufferOverflow strategy other than SUSPEND (either DROP_OLDEST or DROP_LATEST) never suspends on emit, and thus tryEmit to such a shared flow always returns true.

[nitrog42]

yes it should work
I'm pretty happy about this new coroutine addition anyway !