Cancelling coroutines when they are no longer needed can be a task easy to forget, it’s monotonous work and adds a lot of boilerplate code.
viewModelScope contributes to structured concurrency by adding an extension property to the ViewModel class that automatically cancels its child coroutines when the ViewModel is destroyed.
Scopes in ViewModels
A CoroutineScope keeps track of all coroutines it creates. Therefore, if you cancel a scope, you cancel all coroutines it created. This is particularly important if you’re running coroutines in a ViewModel. If your ViewModel is getting destroyed, all the asynchronous work that it might be doing must be stopped. Otherwise, you’ll waste resources and potentially leaking memory. If you consider that certain asynchronous work should persist after ViewModel destruction, it is because it should be done in a lower layer of your app’s architecture.
Add a CoroutineScope to your ViewModel by creating a new scope with a SupervisorJob that you cancel in the
onCleared() method. The coroutines created with that scope will live as long as the ViewModel is being used. See following code:
The heavy work happening in the background will be cancelled if the ViewModel gets destroyed because the coroutine was started by that particular
But that’s a lot of code to be included in every ViewModel, right?
viewModelScope comes to simplify all this.
viewModelScope means less boilerplate code
AndroidX lifecycle v2.1.0 introduced the extension property
viewModelScope to the ViewModel class. It manages the coroutines in the same way we were doing in the previous section. That code is cut down to this:
All the CoroutineScope setup and cancellation is done for us. To use it, import the following dependency in your
Let’s take a look at what’s happening under the hood.
Digging into viewModelScope
The code is publicly available in AOSP.
viewModelScope is implemented as follows:
The ViewModel class has a
ConcurrentHashSet attribute where it can store any kind of object. The CoroutineScope is stored there. If we take a look at the code, the method
getTag(JOB_KEY) tries to retrieve the scope from there. If it doesn’t exist, then it creates a new CoroutineScope the same way we did before and adds the tag to the bag.
When the ViewModel is cleared, it executes the method
clear() before calling the
onCleared() method that we would’ve had to override otherwise. In the
clear() method the ViewModel cancels the Job of the
viewModelScope. The full ViewModel code is also available but we are just focusing on the parts we are interested in:
The method goes through all the items in the bag and calls
closeWithRuntimeException that checks if the object is of type
Closeable and if so, closes it. In order for the ViewModel to close the scope, it needs to implement the
Closeable interface. That’s why
viewModelScope is of type
CloseableCoroutineScope that extends
CoroutineScope overriding the
coroutineContext and implements the
Dispatchers.Main as default
Dispatchers.Main is set as the default CoroutineDispatcher for
val scope = CloseableCoroutineScope(SupervisorJob() + Dispatchers.Main)
Dispatchers.Main is a natural fit for this case since ViewModel is a concept related to UI that is often involved in updating it so launching on another dispatcher will introduce at least 2 extra thread switches. Considering that suspend functions will do their own thread confinement properly, going with other Dispatchers wouldn’t be an option since we’d be making an assumption of what the ViewModel is doing.
Unit Testing viewModelScope
Dispatchers.Main uses the Android
Looper.getMainLooper() method to run code in the UI thread. That method is available in Instrumented Android tests but not in Unit tests.
org.jetbrains.kotlinx:kotlinx-coroutines-test:$coroutines_version library to replace the Coroutines Main Dispatcher by calling
Dispatchers.setMain(dispatcher: CoroutineDispatcher) with a TestCoroutineDispatcher that is available in
TestCoroutineDispatcher is a dispatcher that gives us control of how coroutines are executed, being able to pause/resume execution and control its virtual clock. It was added as an experimental API in Kotlin Coroutines v1.2.1. You can read more about it in the documentation.
Dispatchers.Unconfined as a replacement of
Dispatchers.Main, it will break all assumptions and timings for code that does use
Dispatchers.Main. Since a unit test should run well in isolation and without any side effects, you should call
Dispatchers.resetMain() and clean up the executor when the test finishes running.
You can use this JUnitRule with that logic to simplify your code:
Now, you can use it in your Unit Tests.
Testing coroutines using Mockito
Do you use Mockito and want to
verify that interactions with an object happen? Note that using Mockito’s
verify method is not the preferred way to unit test your code. You should check app-specific logic such as an element is present rather than verifying that interactions with an object happen.
Before checking that the interaction with an object happened, we need to make sure that all coroutines launched have finished. Let’s take a look at the following example.
In the test, we call the
runBlockingTest method inside the
TestCoroutineDispatcher that the rule creates. Since that Dispatcher overrides
Dispatchers.Main, MainViewModel will run the coroutine on that Dispatcher too. Calling
runBlockingTest will make that coroutine to execute synchronously in the test. Since our
verify Mockito call is inside the
runBlockingTest block, it will be called after the coroutine finishes and the interaction will have happened at that moment.
For another example, check out how we added this kind of Unit tests to the Kotlin Coroutines codelab in the following PR:
If you are using architecture components, ViewModel and coroutines, use
viewModelScope to let the framework manage its lifecycle for you. It’s a no brainer!
The Coroutines codelab has been already updated to use it. Check it out to find out more about Coroutines and how to use them in an Android app.