Does this equivalence table make sense?

# coroutines

[tulio]

Does this equivalence table make sense?

-.kt

[streetsofboston]

For many use-cases, this makes sense.
However, don’t count out `Channel`s entirely 🙂

[streetsofboston]

Flow ‘handles’ back-pressure, by not having any 🙂
When calling emit in a flow, you can only do that on the scope of the collector, you can’t switch scope, e.g. you can’t do launch { emit(data) }. If you need to be able to do this, you’d need to create a channelFlow or callbackFlow, which introduce a Channel that has back-pressure

[tulio]

Interesting @streetsofboston. Thank you! 🙂

[Vsevolod Tolstopyatov [JB]]

Flow ‘handles’ back-pressure, by not having any 🙂

Suspension is a back-pressure in the same manner as Channel() (with default zero capacity). Flow effectively handles back-pressure, and if you need more than one in-flight element, buffer(n) can be used

[streetsofboston]

Ah… I didn’t consider suspension to be back-pressure; it just suspends. But, yes, it does effectively acts as one with zero-capacity.

@Vsevolod Tolstopyatov [JB] I found one thing with a flow without a buffer (just suspending).
When running the flow.collect on the Dispatchers.Main and the flow itself on <http://Dispatchers.IO|Dispatchers.IO>, by using flowOn, the flow effectively got a buffer size > 0…

[tulio]

Still about the table… I should use Flow for cold observables and `Channel`s for hot observables, right?

[tulio]

Or I could achieve both behaviours using only Flow?

[streetsofboston]

You have to call collect to make a flow ‘active’.

[streetsofboston]

When you call collect the lambda you provided to flow, channelFlow, etc, is called.
The collect suspends until that lambda has finished.
The flow’s lifecycle and the collector’s lifecycle are tied together like this.

I consider a Flow to be a special kind of suspend fun. Whereas a suspend fun only returns one result in a non-blocking way, a Flow can ‘return’ a stream of values in a non-blocking way. 🙂