Async All the Way

Asynchronous code reminds me of the story of a fellow who mentioned that the world was suspended in space and was immediately challenged by an elderly lady claiming that the world rested on the back of a giant turtle. When the man enquired what the turtle was standing on, the lady replied, “You’re very clever, young man, but it’s turtles all the way down!” As you convert synchronous code to asynchronous code, you’ll find that it works best if asynchronous code calls and is called by other asynchronous code—all the way down (or “up,” if you prefer). Others have also noticed the spreading behavior of asynchronous programming and have called it “contagious” or compared it to a zombie virus. Whether turtles or zombies, it’s definitely true that asynchronous code tends to drive surrounding code to also be asynchronous. This behavior is inherent in all types of asynchronous programming, not just the new async/await keywords.

“Async all the way” means that you shouldn’t mix synchronous and asynchronous code without carefully considering the consequences. In particular, it’s usually a bad idea to block on async code by calling Task.Wait or Task.Result. This is an especially common problem for programmers who are “dipping their toes” into asynchronous programming, converting just a small part of their application and wrapping it in a synchronous API so the rest of the application is isolated from the changes. Unfortunately, they run into problems with deadlocks. After answering many async-related questions on the MSDN forums, Stack Overflow and e-mail, I can say this is by far the most-asked question by async newcomers once they learn the basics: “Why does my partially async code deadlock?”

Figure 3 shows a simple example where one method blocks on the result of an async method. This code will work just fine in a console application but will deadlock when called from a GUI or ASP.NET context. This behavior can be confusing, especially considering that stepping through the debugger implies that it’s the await that never completes. The actual cause of the deadlock is further up the call stack when Task.Wait is called.

Figure 3 A Common Deadlock Problem When Blocking on Async Code

public static class DeadlockDemo
{
  private static async Task DelayAsync()
  {
    await Task.Delay(1000);
  }
  // This method causes a deadlock when called in a GUI or ASP.NET context.
  public static void Test()
  {
    // Start the delay.
    var delayTask = DelayAsync();
    // Wait for the delay to complete.
    delayTask.Wait();
  }
}

The root cause of this deadlock is due to the way await handles contexts. By default, when an incomplete Task is awaited, the current “context” is captured and used to resume the method when the Task completes. This “context” is the current SynchronizationContext unless it’s null, in which case it’s the current TaskScheduler. GUI and ASP.NET applications have a SynchronizationContext that permits only one chunk of code to run at a time. When the await completes, it attempts to execute the remainder of the async method within the captured context. But that context already has a thread in it, which is (synchronously) waiting for the async method to complete. They’re each waiting for the other, causing a deadlock.

Note that console applications don’t cause this deadlock. They have a thread pool SynchronizationContext instead of a one-chunk-at-a-time SynchronizationContext, so when the await completes, it schedules the remainder of the async method on a thread pool thread. The method is able to complete, which completes its returned task, and there’s no deadlock. This difference in behavior can be confusing when programmers write a test console program, observe the partially async code work as expected, and then move the same code into a GUI or ASP.NET application, where it deadlocks.

The best solution to this problem is to allow async code to grow naturally through the codebase. If you follow this solution, you’ll see async code expand to its entry point, usually an event handler or controller action. Console applications can’t follow this solution fully because the Main method can’t be async. If the Main method were async, it could return before it completed, causing the program to end. Figure 4 demonstrates this exception to the guideline: The Main method for a console application is one of the few situations where code may block on an asynchronous method.

Figure 4 The Main Method May Call Task.Wait or Task.Result

class Program
{
  static void Main()
  {
    MainAsync().Wait();
  }
  static async Task MainAsync()
  {
    try
    {
      // Asynchronous implementation.
      await Task.Delay(1000);
    }
    catch (Exception ex)
    {
      // Handle exceptions.
    }
  }
}

Allowing async to grow through the codebase is the best solution, but this means there’s a lot of initial work for an application to see real benefit from async code. There are a few techniques for incrementally converting a large codebase to async code, but they’re outside the scope of this article. In some cases, using Task.Wait or Task.Result can help with a partial conversion, but you need to be aware of the deadlock problem as well as the error-handling problem. I’ll explain the error-handling problem now and show how to avoid the deadlock problem later in this article.

Every Task will store a list of exceptions. When you await a Task, the first exception is re-thrown, so you can catch the specific exception type (such as InvalidOperationException). However, when you synchronously block on a Task using Task.Wait or Task.Result, all of the exceptions are wrapped in an AggregateException and thrown. Refer again to Figure 4. The try/catch in MainAsync will catch a specific exception type, but if you put the try/catch in Main, then it will always catch an AggregateException. Error handling is much easier to deal with when you don’t have an AggregateException, so I put the “global” try/catch in MainAsync.

So far, I’ve shown two problems with blocking on async code: possible deadlocks and more-complicated error handling. There’s also a problem with using blocking code within an async method. Consider this simple example:

public static class NotFullyAsynchronousDemo
{
  // This method synchronously blocks a thread.
  public static async Task TestNotFullyAsync()
  {
    await Task.Yield();
    Thread.Sleep(5000);
  }
}

This method isn’t fully asynchronous. It will immediately yield, returning an incomplete task, but when it resumes it will synchronously block whatever thread is running. If this method is called from a GUI context, it will block the GUI thread; if it’s called from an ASP.NET request context, it will block the current ASP.NET request thread. Asynchronous code works best if it doesn’t synchronously block. Figure 5 is a cheat sheet of async replacements for synchronous operations.

Figure 5 The “Async Way” of Doing Things

To Do This …	Instead of This …	Use This
Retrieve the result of a background task	Task.Wait or Task.Result	await
Wait for any task to complete	Task.WaitAny	await Task.WhenAny
Retrieve the results of multiple tasks	Task.WaitAll	await Task.WhenAll
Wait a period of time	Thread.Sleep	await Task.Delay

To summarize this second guideline, you should avoid mixing async and blocking code. Mixed async and blocking code can cause deadlocks, more-complex error handling and unexpected blocking of context threads. The exception to this guideline is the Main method for console applications, or—if you’re an advanced user—managing a partially asynchronous codebase.