6. Coroutine Context

Written by Filip Babić

You’re getting pretty handy with coroutines, aren’t ya? In the first section of this book you’ve seen how you can start coroutines, bridge between threads in coroutines to return values, create your own APIs and much more. In the second section, you’ll focus on the internal coroutine concepts. And the most important is the CoroutineContext. Even though it’s at the core of every coroutine, you’ll see that it’s fairly simple after you take a look at its implementation and usage.

Contextualizing coroutines

Each coroutine is tied to a CoroutineContext. The context is a wrapper around a set of CoroutineContext.Elements, each of which describes a vital part that builds up and forms a coroutine: like the way exceptions are propagated, the execution flow is navigated, or just the general lifecycle.

These elements are:

• Job: A cancellable piece of work, which has a defined lifecycle.
• ContinuationInterceptor: A mechanism which listens to the continuation within a coroutine and intercepts its resumption.
• CoroutineExceptionHandler: A construct which handles exceptions in coroutines.

So, when you run launch(), you can pass it a context of your own choice. The context defines which elements will fit into the puzzle. If you pass in another Job, which also implements CoroutineContext.Element, you’ll define what the new coroutine’s parent is. As such, if the parent job finishes, it will notify all of its children, including the newly created coroutine.

If, however, you pass in an exception handler, another CoroutineContext.Element, you give the coroutine a way to process errors if something bad happens.

And the last thing you can pass in is a ContinuationInterceptor. These constructs control the flow of each coroutine-powered function, by determining which thread it should operate on and how it should distribute work.

The problem is, you wouldn’t want to provide a full implementation that manually handles continuations. If you want something else to do that part for you, while also being a CoroutineContext.Element, you have to provide a coroutine dispatcher.

You’ve used some of them before — like Dispatchers.Default. So the key to understanding ContinuationInterceptor usage is by learning what a dispatcher really is, which you’ll do in “Chapter 7: Context Switch and Dispatching”. For now, you’ll focus on combining and providing CoroutineContexts.

Using CoroutineContext

To follow the code in this chapter, import this chapter’s starter project using IntelliJ by selecting Import Project. Then navigate to the coroutine-context/projects/starter folder, selecting the coroutine-context project.

Even though you haven’t gone too deep into it, you’ve already used CoroutineContext extensively. Every time you’ve created a coroutine, from a CoroutineScope, you’ve passed in the scope’s CoroutineContext to the builders. Take the following snippet for an example:

GlobalScope.launch {
  println("In a coroutine")
}

You don’t see it, but there’s work done around the CoroutineContext for this simple snippet of code. Once again, if you look at the definiton of the launch, this is what you can see:

public fun CoroutineScope.launch(
    context: CoroutineContext = EmptyCoroutineContext,
    start: CoroutineStart = CoroutineStart.DEFAULT,
    block: suspend CoroutineScope.() -> Unit
): Job

You can see that the default context is the EmptyCoroutineContext. This basically means it’s going to use the most default behavior - no special lifecycle, no exception handling from within the coroutine, and most importantly - no custom threading. Futher along, launch() calls newCoroutineContext(context), to build up a full context for the coroutine. The code underneath is a bit complex, but in essence if the context is fully empty, it adds the Dispatchers.Default to it, adding default background worker threading.

So even though you don’t see it, the API itself uses contexts to achieve at least the default behavior. But you should always strive to explicitly and clearly provide what you want to happen.

Combining contexts can lead to very powerful mechanisms, so let’s see what it’s about.

Combining different contexts

Another interesting aspect to coroutine contexts is the ability to compose them and combine their functionality. Using the +/plus operator, you can create a new CoroutineContext from the combination of the two. Since you know each coroutine is composed of several objects, like the continuation interceptor for the flow, exception handler for errors and a job for lifecycle, there has to be a way to create a new coroutine with all these pieces of the puzzle. And this is where summing contexts comes in handy. You can do it as simply as this:

fun main() {
  val defaultDispatcher = Dispatchers.Default

  val coroutineErrorHandler = CoroutineExceptionHandler { context, error ->
    println("Problems with Coroutine: ${error}") // we just print the error here
  }

  val emptyParentJob = Job()

  val combinedContext = defaultDispatcher + coroutineErrorHandler + emptyParentJob

  GlobalScope.launch(context = combinedContext) {
    println(Thread.currentThread().name)
  }

  Thread.sleep(50)
}

The code above is an example of how to create a context from both the Dispatchers.Default and the error handler. As such, you can add more functionality to a context all at once, effectively building up all of the features your coroutine should use — error handling, threading and lifecycle. So, if you’ve built a complex CoroutineContext for error handling, it would be cool to be able to use it in every coroutine you create. The same goes for the lifecycle of coroutines and their threading mechanisms.

By summing CoroutineContexts, you combine all of their CoroutineContext.Elements, creating a union of their functionality. However, there are some things which don’t make sense, like combining two different Dispatchers. This would mean the second dispatcher’s threading will override the first one’s. If you try to do that, the compiler will even give you a message saying it doesn’t make sense.

Providing contexts

When it comes to software, you usually want to build it in a way that abstracts away the communication between layers. With threading, it’s useful to abstract the way you switch between different threads. You can abstract this by attaching a thread provider, providing both main and background threads. It’s no different with coroutines! Since the threading mechanism is abstracted with CoroutineContext objects, and their respective CoroutineDispatcher instances, you can build a provider that you’d use to delegate which context should be used every time you build coroutines. Usually, these providers have a declared interface, which gives you the main and background threads or schedulers, since that’s what’s important in applications with user interfaces.

Let’s see how you’d build such a provider.

Building the ContextProvider

You’ve already learned which CoroutineContext objects exist and what their behavior is. To build the provider, you first have to define an interface, which provides a generic context, which you’ll run the expensive operations on. Note that this Provider interface is not part of Coroutines but will help us abstract out the main and background contexts. The interface would look like this:

interface CoroutineContextProvider {

  fun context(): CoroutineContext
}

This way, you can build many different CoroutineContextProviders, each for a specific use case you may have in mind. In the implementation, you can pass in the required CoroutineContext to the constructor, abstracing away the information in a factory function, or your dependency injection graph.

class CoroutineContextProviderImpl(
    private val context: CoroutineContext
) : CoroutineContextProvider {

  override fun context(): CoroutineContext = context
}

This way, whenever you’re building coroutines, you can use the context provider:

GlobalScope.launch(context = provider.context()) {
}

Because of this you’re able to depend on the abstract provider of contexts, rather than manually writing all of the contexts you use. Additionally, when you build the provider, you could pass in any context you want, effectively switching out the pools of threads that let the event of work swim in. To create such a provider, you can do the following:

val backgroundContextProvider = 
  CoroutineContextProviderImpl(Dispatchers.Default)

You could go another step forward, and make the context provider provide not only thread-based contexts, but also error handling contexts and a lifecycle-related context. Or virtually any combination of those coroutine context elements.

This is extremely useful if you want to abstract away the contexts you’re using often. Additionally, it can help you with testing, which you’ll see in the “Chapter 16: Testing coroutines”.

You can check out these examples by importing this chapter’s final project, using IntelliJ, and selecting Import Project, and navigating to the coroutine-context/projects/final folder, selecting the coroutine-context project.

Key points

• All the information for coroutines is contained in a CoroutineContext and its CoroutineContext.Elements.
• There are three main coroutine context elements: the Job, which defines the lifecycle and can be cancelled, a CoroutineExceptionHandler, which takes care of errors, and the ContinuationInterceptor, which handles function execution flow and threading.
• Each of the coroutine context elements implements CoroutineContext.
• ContinuationInterceptors, which take care of the input/output of threading. The main and background threads are provided through the Dispatchers.
• You can combine different CoroutineContexts and their Elements by using the +/plus operator, effectively summing their elements.
• A good practice is to build a CoroutineContext provider, so you don’t depend on explicit contexts.
• With the CoroutineContextProvider you can abstract away complex contexts, like custom error handling, coroutine lifecycles or threading mechanisms.