This is a small library that provides the Kotlin Coroutines suspending extension Call.await()
for Retrofit 2
Based on kotlinx.coroutines implementation.
New version of library (after 1.0.0) support only Kotlin 1.3
Kotlin 1.2 and experimental coroutines are not supported anymore, but you can use version 0.13.0
for old projects.
Download the JAR:
Gradle:
implementation 'ru.gildor.coroutines:kotlin-coroutines-retrofit:1.1.0'
Maven:
<dependency>
<groupId>ru.gildor.coroutines</groupId>
<artifactId>kotlin-coroutines-retrofit</artifactId>
<version>1.1.0</version>
</dependency>
NOTE: All examples in this README use
runBlocking
to build coroutine but it is only useful for testing or examples.For a real application you probably want to use some other coroutines builder that does not block a thread, for example
launch
from kotlinx.coroutines.If you want to use this library for UI please also check the Guide to UI programming with coroutines
There are three suspending extensions:
Common await API that returns a result or throws an exception
fun Call<T>.await(): T
In case of an HTTP error or an invocation exception await()
throws an exception
// You can use retrofit suspended extension inside any coroutine block
fun main(args: Array<String>): Unit = runBlocking {
try {
// Wait (suspend) for result
val user: User = api.getUser("username").await()
// Now we can work with result object
println("User ${user.name} loaded")
} catch (e: HttpException) {
// Catch http errors
println("exception${e.code()}", e)
} catch (e: Throwable) {
// All other exceptions (non-http)
println("Something broken", e)
}
}
Common await API that returns a Response or throws an exception
fun Call<T>.awaitResponse(): Response<T>
In case of an invocation exception awaitResponse()
throws an exception
// You can use retrofit suspended extension inside any coroutine block
fun main(args: Array<String>): Unit = runBlocking {
try {
// Wait (suspend) for response
val response: Response<User> = api.getUser("username").awaitResponse()
if (response.isSuccessful()) {
// Now we can work with response object
println("User ${response.body().name} loaded")
}
} catch (e: Throwable) {
// All other exceptions (non-http)
println("Something broken", e)
}
}
API based on sealed class Result:
fun Call<T>.awaitResult(): Result<T>
fun main(args: Array<String>): Unit = runBlocking {
// Wait (suspend) for Result
val result: Result<User> = api.getUser("username").awaitResult()
// Check result type
when (result) {
//Successful HTTP result
is Result.Ok -> saveToDb(result.value)
// Any HTTP error
is Result.Error -> log("HTTP error with code ${result.error.code()}", result.error)
// Exception while request invocation
is Result.Exception -> log("Something broken", e)
}
}
Also, Result
has a few handy extension functions that allow to avoid when
block matching:
fun main(args: Array<String>): Unit = runBlocking {
val result: User = api.getUser("username").awaitResult()
//Return value for success or null for any http error or exception
result.getOrNull()
//Return result or default value
result.getOrDefault(User("empty-user"))
//Return value or throw exception (HttpException or original exception)
result.getOrThrow()
//Also supports custom exceptions to override original ones
result.getOrThrow(IlleagalStateException("User request failed"))
}
All Result
classes also implemented one or both interfaces: ResponseResult
and ErrorResult
You can use them for access to shared properties of different classes from Result
fun main(args: Array<String>): Unit = runBlocking {
val result: User = api.getUser("username").awaitResult()
//Result.Ok and Result.Error both implement ResponseResult
if (result is ResponseResult) {
//And after smart cast you now have an access to okhttp3 Response property of result
println("Result ${result.response.code()}: ${result.response.message()}")
}
//Result.Error and Result.Exception implement ErrorResult
if (result is ErrorResult) {
// Here yoy have an access to `exception` property of result
throw result.exception
}
}
To prevent unexpected behavior with a nullable body of response Call<Body?>
extensions .await()
and .awaitResult()
are available only for
non-nullable Call<Body>
or platform Call<Body!>
body types:
fun main(args: Array<String>): Unit = runBlocking {
val user: Call<User> = api.getUser("username")
val userOrNull: Call<User?> = api.getUserOrNull("username")
// Doesn't work, because User is nullable
// userOrNull.await()
// Works for non-nullable type
try {
val result: User = user.await()
} catch (e: NullPointerException) {
// If body will be null you will get NullPointerException
}
// You can use .awaitResult() to catch possible problems with nullable body
val nullableResult = api.getUser("username").awaitResult().getOrNull()
// But type of body should be non-nullable
// api.getUserOrNull("username").awaitResult()
// If you still want to use nullable body to clarify your api
// use awaitResponse() instead:
val responseBody: User? = userOrNull.awaitResponse().body()
}
By wrapping call with kotlinx.coroutines
async(), you may run a few requests parallelly without waiting for the previous request.
fun main(args: Array<String>): Unit = runBlocking {
val users = listOf("user1", "user2", "user3")
.map { username ->
// Pass any coroutine context that fits better for your case
// Coroutine Dispatcher also controls parallelism level
// for CommonPool parallelism is `availableProcessors - 1`
// But you can use any custom dispatcher with any parallelism strategy
async(CommonPool) {
// Send request. We use `awaitResult()` here to avoid try/catch,
// but you can use `await()` and catch exceptions
api.getUser(username).awaitResult()
}
}
// Handle results
// in this example we get result or null in case of error and filter all nulls
.mapNotNull {
// Wait (suspend) for result of `async()` and get result of request
// We must call first `await()` only when all `async` blocks are created for parallel requests
it.await().getOrNull()
}
}
You can read more about concurrent usage of async in the kotlinx.coroutines guide