Background work

Threads

Coroutines

• Coroutines are asynchronous and non-blocking.
• Coroutines use suspend functions to make asynchronous code sequential.

Suspending

Suspend functions are only allowed to be called from a coroutine or another suspend function

suspend fun doSomethingLong() {
    // request server, DB, filesystem, ...
}

fun main() {
    doSomethingLong() //  KO

    GlobalScope.launch {
        doSomethingLong() //  OK
    }
}

suspend fun otherSuspendFunction() {
    doSomethingLong() //  OK
}

Couroutine Context

Coroutines always execute in some CoroutineContext: a set of various elements, mainly its Job and its CoroutineDispatcher

Scope

A coroutine's scope defines the context in which the coroutine runs.

• A scope combines information about a coroutine's Job and CoroutineDispatcher
• Scopes keep track of coroutines that are “in them”

actually just a wrapper around a CoroutineContext, can be seen as a "parent context"

ex: GlobalScope, MainScope, viewModelScope, lifeCycleScope

Job

Basically, a Job is anything that can be canceled

• Every coroutine has a Job, and you can use it to cancel the coroutine
• Jobs can be arranged into parent-child hierarchies
• Canceling a parent job immediately cancels all the job's children

fun main() {
    val job = GlobalScope.launch {
        // do something long
    }
    if (input == `^C`) job.cancel()
}

Dispatcher

The CoroutineDispatcher sends off coroutines to run on various threads

suspend fun fetch(url: String) {
    // start on Dispatchers.Main
    showLoading(true)
    val result = withContext(Dispatchers.IO) { // switch to IO for network call
        longNetworkCall(url)
    }
    // back on Dispatchers.Main
    showLoading(false)
    showResult(result)
}

Usage

val job = scope.launch { /* Do something long */ }
job.join() // wait for work to finish
job.cancel() // cancel work

val result: Deffered<SomeClass> = scope.async {
    mySuspendFun()
}
val result: SomeClass = result.await() // wait for result data

Usage on Android

// in Repository
suspend fun getData() : Int = withContext(Dispatchers.IO) {
    // execute long IO operation
}

// in ViewModel
viewModelScope.launch { // canceled when ViewModel is destroyed
    repository.getData()
}


// in Fragment or Activity
lifecycleScope.launch { // canceled when fragment is destroyed
    whenStarted { /* starts when fragment is in started state */ }
    // the rest executes after the whenStarted block
}
lifecycleScope.launchWhenStarted { /* launches when fragment is in started state */ }

Observer pattern

Design pattern that allows decoupling actions and data consumption by decoupling the observable (or subject) from the observers (or listeners):

val observable = Observable<Data>(initialData)

observable.notify(newData)

observable.observe { data -> /* use the value */ }

Reactive Streams

Represent data as an async sequence that can be observed : (pseudo-code)

val stream = Stream(null)

listOf("red", "white", "blue").forEach { color ->
    stream.send(color)
    wait(1000)
}

stream.filterNotNull()
      .map { it.toUpperCase() }
      .receiveEach { value -> log(value) }

Streams can be "hot" or "cold" (analogy: Radio // CD)

Flow

Implementation of reactive streams based on coroutines:

val flow: Flow<String> = listOf("red", "white", "blue").asFlow()
            .map { it -> it.ToUpperCase() }
            .flowOn(Dispatchers.Default)

scope.launch {
    flow.collect { response -> println(response) }
}

StateFlow

Special type of flow that always has a value:

// create and modify in ViewModel:
val userFlow = MutableStateFlow<NetworkUser?>(null)
suspend fun refreshUser() { userFlow.value = repository.fetchUser() }

// react in Activity/Fragment:
lifecycleScope.launch {
    userFlow.filterNotNull().collect { user -> nameTextView.text = user.name }
}

// or in Compose:
val user by userFlow.collectAsState()
Text(user.name)