Was looking at the benchmark code in the coroutine project Ran into this ```fun concat first Flow lt Int gt second Flow lt Int gt Flow lt Int gt = flow first collect value gt return collect emit value second collect value gt return collect emit value ``` What is the point of the `return collect`

fun concat(first: Flow<Int>, second: Flow<Int>): Flow<Int> = flow {
    first.collect { value ->
        return@collect emit(value)
    }

    second.collect { value ->
        return@collect emit(value)
    }
}

What is the point of the

return@collect

octylFractal

05/28/2020, 8:16 PM

there is no void in kotlin so in java that looks something like:

second.collect(value -> {
  return emit.invoke(value);
});

but if that was not added, it would be

second.collect(value -> {
  emit.invoke(value);
  return Unit.INSTANCE;
});

which can potentially be less optimal I presume

taer

05/28/2020, 8:18 PM

so when I remove the

return@collect

it still works identically.

octylFractal

05/28/2020, 8:19 PM

yes, it's still the same in theory

but I'm not a JVM optimization expert. for all I know this actually has a significant effect especially in the case of coroutines, I think it does because of the extra case where it returns

COROUTINE_SUSPENDED

. I recall seeing this kludge in the flow code somewhere, let me see if I can find it

taer

05/28/2020, 8:20 PM

is it shortcutting something? I saw stuff like that in the stdlib coroutine too

I recall something about tailrec in one I looked at

but this one doesn't seem like it's related

Dominaezzz

05/28/2020, 8:22 PM

Are you talking about tail call optimisation?

taer

05/28/2020, 8:23 PM

yeah.. I saw something in the stdlib mentioning that w/ a similar looking return

octylFractal

05/28/2020, 8:24 PM

https://github.com/Kotlin/kotlinx.coroutines/blob/1eeed509abbe8e542124841ce14884202463460b/kotlinx-coroutines-core/common/src/flow/operators/Emitters.kt#L40

and of course the relevant issue, https://youtrack.jetbrains.com/issue/KT-28938

taer

05/28/2020, 8:24 PM

yup.. That was it.

octylFractal

05/28/2020, 8:24 PM

though it appears to have been fixed

taer

05/28/2020, 8:25 PM

awesome. So that's the same concept in that concat method I assume

octylFractal

05/28/2020, 8:25 PM

yes

taer

05/28/2020, 8:26 PM

Awesome.. Thanks!

octylFractal

05/28/2020, 8:29 PM

can confirm that on recent versions (1.3.61), this compiles the same https://gist.github.com/octylFractal/34ae00f1436d5391c7d427b74ab1850f

so that's unnecessary now

taer

05/28/2020, 8:31 PM

heh.. I just did the same thing. 🙂

2 Views

#coroutines Join Slack