Hi, I'm a bit confused when to use `LiveData`, `Channels`, `Flow`. I did some researches and if I go...

# coroutines

[Lilly]

Hi, I'm a bit confused when to use LiveData, Channels, Flow. I did some researches and if I got it right, LiveData is prefered in ViewModel because of it's lifecycle awareness and it runs in main thread while in data layer Flow is preferred. But what about Channels, where does it fit in? Or does Flow replace Channels in the long run?

[r4zzz4k]

Channel is nowadays just a low-level primitive for coroutines synchronization. There are still cases not covered by Flow, but there are becoming less of them with time.

[Lilly]

Sorry, but what means low-level?

[Zach Klippenstein (he/him) [MOD]]

Channels are basically an implementation of "blocking" queues (except they don't block, they suspend). If you need a blocking queue in a coroutine, use a channel. They're probably not the right tool for most other things though*.

* This is technically only true once SharedFlow and friends get released, which will hopefully be soon.

[r4zzz4k]

I mean in terms of abstraction level. Flow is more like a user-facing thing which would appear throughout your public API. Channels, on the other hand, would appear only internally, specifically when you implement some custom patterns of communication between your coroutines not supported otherwise.

[Lilly]

Thanks for the detailed answer. When I'm in a data layer (i.e. a class which acts like a repository) and I want to expose a connection state (i.e. an enum CONNECTED, DISCONNECTED, ESTABLISHED) should I use a channel or a flow?

[Lilly]

I guess Flow would better fit in here, right?

[Zach Klippenstein (he/him) [MOD]]

Yes, StateFlow would probably be the best fit.

[Lilly]

ok nice