StateFlow vs SharedFlow vs Channel

Comparison

	StateFlow	SharedFlow	Channel	LiveData
Hot / cold	hot	hot	hot	hot
Has current value	yes	no (replay cache)	no	yes
Receivers	all collectors	all collectors	each value to one receiver	all observers
Duplicates	skipped	delivered	delivered	delivered
Platform	any (KMP)	any	any	Android only
Best for	UI state	events to many listeners	one-time events / work queues	older Java code

Choosing

Screen state (loading, list, error): StateFlow<UiState>.
One-time UI events (navigate, show snackbar) that must not be lost while the screen is in the background: Channel(Channel.BUFFERED) exposed with receiveAsFlow(). Many teams instead store the event in the state and clear it after it is shown.
Broadcast to several parts of the app (logout happened, cart changed): SharedFlow.

Exercise: Create a MutableStateFlow holding a data class Filter(val veg: Boolean, val maxPrice: Int). Combine it with a flow of dishes and print the filtered list each time the filter changes.

Show answer

import kotlinx.coroutines.*
import kotlinx.coroutines.flow.*

data class Dish(val name: String, val veg: Boolean, val price: Int)
data class Filter(val veg: Boolean = false, val maxPrice: Int = 1000)

fun main() = runBlocking {
    val dishes = MutableStateFlow(listOf(Dish("Paneer Tikka", true, 280), Dish("Chicken 65", false, 320), Dish("Dosa", true, 90)))
    val filter = MutableStateFlow(Filter())
    val job = launch {
        combine(dishes, filter) { list, f -> list.filter { (!f.veg || it.veg) && it.price <= f.maxPrice }.map { it.name } }
            .collect { println(it) }
    }
    delay(10); filter.update { it.copy(veg = true) }
    delay(10); filter.update { it.copy(maxPrice = 100) }
    delay(10); job.cancel()
}