Coroutines are the part of Kotlin I reach for whenever I need to deal with asynchronous work. They keep the code reading top-to-bottom, even when the underlying work is non-blocking, which makes concurrency a lot less painful than juggling raw threads or callbacks.

What are Coroutines?

Coroutines are light-weight threads that allow you to write asynchronous code in a sequential style. Unlike traditional threads, thousands of coroutines can run on a single thread without significant overhead.

Setup

Add the coroutines dependency to your project:

// Gradle (Kotlin DSL)
dependencies {
    implementation("org.jetbrains.kotlinx:kotlinx-coroutines-core:1.7.3")
    // For Android
    implementation("org.jetbrains.kotlinx:kotlinx-coroutines-android:1.7.3")
}

Basic Concepts

Suspend Functions

Functions that can pause execution without blocking the thread:

suspend fun fetchUser(): User {
    delay(1000)  // Non-blocking delay
    return User("John")
}

suspend fun longRunningTask(): Long {
    val time = measureTimeMillis {
        println("Starting...")
        delay(2000)  // Suspends for 2 seconds
        println("Done!")
    }
    return time
}

runBlocking

Bridges blocking and non-blocking worlds (mainly for main functions and tests):

fun main() = runBlocking {
    val time = longRunningTask()
    println("Execution time: $time ms")
}

launch

Starts a new coroutine that doesn’t return a result:

fun main() = runBlocking {
    launch {
        delay(1000)
        println("World!")
    }
    println("Hello,")  // Prints immediately
}
// Output: Hello, World!

async

Starts a coroutine that returns a result via Deferred:

fun main() = runBlocking {
    val deferred = async {
        delay(1000)
        "Hello, Async!"
    }

    println("Waiting...")
    val result = deferred.await()  // Suspends until result is ready
    println(result)
}

Coroutine Context and Dispatchers

Dispatchers

Control which thread(s) the coroutine runs on:

launch(Dispatchers.Default) {
    // CPU-intensive work (background thread pool)
}

launch(Dispatchers.IO) {
    // I/O operations (larger thread pool)
}

launch(Dispatchers.Main) {
    // UI updates (Android main thread)
}

launch(Dispatchers.Unconfined) {
    // Runs in caller thread until first suspension
}

Common Pool Pattern

fun main() = runBlocking {
    val time = async(Dispatchers.Default) { longRunningTask() }

    println("Printed after async launch")

    val result = time.await()
    println("Time taken: $result ms")
}

Structured Concurrency

Coroutine Scope

Coroutines are scoped, ensuring proper cancellation and resource cleanup:

class MyViewModel : ViewModel() {
    private val scope = CoroutineScope(Dispatchers.Main + Job())

    fun loadData() {
        scope.launch {
            val data = fetchData()
            updateUI(data)
        }
    }

    override fun onCleared() {
        scope.cancel()  // Cancel all coroutines
    }
}

coroutineScope

Creates a scope that waits for all children to complete:

suspend fun fetchTwoPosts(): Pair<Post, Post> = coroutineScope {
    val post1 = async { fetchPost(1) }
    val post2 = async { fetchPost(2) }
    Pair(post1.await(), post2.await())
}

Exception Handling

try-catch

launch {
    try {
        riskyOperation()
    } catch (e: Exception) {
        handleError(e)
    }
}

CoroutineExceptionHandler

val handler = CoroutineExceptionHandler { _, exception ->
    println("Caught $exception")
}

val scope = CoroutineScope(Dispatchers.Default + handler)

scope.launch {
    throw RuntimeException("Boom!")
}

async Exception Handling

Note: Exceptions in async are only thrown when await() is called:

val deferred = async {
    throw RuntimeException("Error!")
}

try {
    deferred.await()  // Exception thrown here
} catch (e: Exception) {
    println("Caught: ${e.message}")
}

Cancellation

Checking for Cancellation

suspend fun longTask() = coroutineScope {
    repeat(1000) { i ->
        if (!isActive) return@coroutineScope  // Check cancellation
        // or use ensureActive()
        delay(100)
        println("Processing $i")
    }
}

Timeout

val result = withTimeout(1000) {
    fetchData()  // Throws TimeoutCancellationException if takes > 1s
}

// Or with null on timeout
val result = withTimeoutOrNull(1000) {
    fetchData()
}

Flow (Cold Streams)

For asynchronous data streams:

fun numbers(): Flow<Int> = flow {
    for (i in 1..3) {
        delay(100)
        emit(i)
    }
}

fun main() = runBlocking {
    numbers()
        .filter { it % 2 == 1 }
        .map { it * it }
        .collect { println(it) }  // 1, 9
}

Channels (Hot Streams)

For communication between coroutines:

val channel = Channel<Int>()

launch {
    for (x in 1..5) channel.send(x * x)
    channel.close()
}

launch {
    for (y in channel) println(y)  // 1, 4, 9, 16, 25
}

Time Measurement

import kotlin.system.measureTimeMillis

val elapsed = measureTimeMillis {
    runBlocking {
        delay(1000)
    }
}
println("Took $elapsed ms")

Kotlin Native Concurrency

For Kotlin Multiplatform, there are special considerations:

Frozen Objects

In Kotlin/Native, objects passed between threads are frozen (immutable):

// Objects are frozen by default
object DataManager {
    // Use @ThreadLocal for mutable state
    @ThreadLocal
    var mutableData: String? = null
}

// Global properties are accessible only from main thread
// unless marked @SharedImmutable or @ThreadLocal

Worker API

// Background work in Kotlin/Native
val worker = Worker.start()
val future = worker.execute(TransferMode.SAFE, { data }) {
    // This runs in worker thread
    processData(it)
}
val result = future.result  // Get result (freezes if needed)

Best Practices for Native

  1. Use ensureNeverFrozen() to catch unexpected freezing
  2. Understand that freeze() applies to object graph
  3. Be careful with closures capturing outer scope
  4. Use @ThreadLocal for thread-specific mutable state

Unit Testing Coroutines

dependencies {
    testImplementation("junit:junit:4.13.2")
    testImplementation("org.mockito:mockito-core:4.0.0")
    testImplementation("org.jetbrains.kotlin:kotlin-test-junit:$kotlin_version")
    testImplementation("org.jetbrains.kotlinx:kotlinx-coroutines-test:1.7.3")
}

@Test
fun `test coroutine`() = runTest {
    val result = async { fetchData() }.await()
    assertEquals("expected", result)
}

// With TestDispatcher for time control
@Test
fun `test with delay`() = runTest {
    var result = ""
    launch {
        delay(1000)
        result = "done"
    }
    advanceTimeBy(1000)
    assertEquals("done", result)
}

Best Practices

  1. Use structured concurrency: Always scope coroutines properly
  2. Prefer suspend functions: Make asynchronous operations explicit
  3. Choose the right dispatcher: IO for network/disk, Default for CPU
  4. Handle exceptions properly: Use try-catch or exception handlers
  5. Cancel when appropriate: Clean up resources and stop unnecessary work
  6. Avoid GlobalScope: Use proper scoping for lifecycle management

Common Patterns

Sequential by Default

suspend fun loadContent() {
    val user = fetchUser()       // Waits
    val posts = fetchPosts(user) // Then executes
    display(user, posts)
}

Concurrent with async

suspend fun loadContent() = coroutineScope {
    val user = async { fetchUser() }
    val posts = async { fetchPosts() }
    display(user.await(), posts.await())  // Parallel execution
}

Fire and Forget with launch

fun logAnalytics(event: String) {
    scope.launch(Dispatchers.IO) {
        analyticsService.log(event)  // Don't wait for result
    }
}

Wrapping up

Once you get comfortable with scopes, dispatchers, and exception handling, coroutines tend to fade into the background — they just stop being the part of the code you worry about. Flow and channels round out the toolkit for streaming and inter-coroutine communication, and structured concurrency keeps lifecycles tidy without much thought.