TaskCompletionSource - Trying to understand threadless async work

Question

I'm trying to understand the purpose of TaskCompletionSource and its relation to async/threadless work. I think I have the general idea but I want to make sure my understanding is correct.

I first started looking into the Task Parallel Library (TPL) to figure out if there was a good way to create your own threadless/async work (say you're trying to improve scalability of your ASP.NET site) plus understanding of the TPL looks like it will be very important in the future (async/await). Which led me to the TaskCompletionSource.

From my understanding it looks like adding TaskCompletionSource to a one of your classes doesn't really do much in as making your coding async; if you're still executing sync code then the call to your code will block. I think this is even true of microsoft APIs. For example, say in DownloadStringTaskAsync off of WebClient class, any setup / sync code they are doing initially will block. The code you're executing has to run on some thread, either the current thread or you will have to spin off a new one.

So you use TaskCompletionSource in your own code when you're calling other async calls from Microsoft so the client of your classes doesn't have to create a new thread for your class to not block.

Not sure how Microsoft does their async APIs internally. For example, there is a new async method off of the SqlDataReader for .Net 4.5. I know there is IO Completion Ports. I think it's a lower level abstraction (C++?) that probably most C# developers won't use. Not sure if IO completion Ports will work for Database or network calls (HTTP) or if its just used for file IO.

So the question is, am I correct in my understanding correct? Are there certain things I've represented incorrectly?

Answer 1

TaskCompletionSource is used to create Task objects that don't execute code.

They're used quite a bit by Microsoft's new async APIs - any time there's I/O-based asynchronous operations (or other non-CPU-based asynchronous operations, like a timeout). Also, any async Task method you write will use TCS to complete its returned Task.

I have a blog post Creating Tasks that discusses different ways to create Task instances. It's written from an async/await perspective (not a TPL perspective), but it still applies here.

Also see Stephen Toub's excellent posts:

• The Nature of TaskCompletionSource
• Mechanisms for Creating Tasks
• await anything; (using TaskCompletionSource to await anything).
• Using Tasks to implement the APM Pattern (creating Begin/End using TaskCompletionSource).

Answer 2

I like the explanation which was provided in http://tutorials.csharp-online.net/TaskCompletionSource

(sorry, the link may be dead at the moment)

First two paragraphs are below

We've seen how Task.Run creates a task that runs a delegate on a pooled (or non-pooled) thread. Another way to create a task is with TaskCompletionSource.

TaskCompletionSource lets you create a task out of any operation that starts and finishes some time later. It works by giving you a "slave" task that you manually drive—by indicating when the operation finishes or faults. This is ideal for I/O- bound work: you get all the benefits of tasks (with their ability to propagate return values, exceptions, and continuations) without blocking a thread for the duration of the operation.

To use TaskCompletionSource, you simply instantiate the class. It exposes a Task property that returns a task upon which you can wait and attach continuations—just as with any other task. The task, however, is controlled entirely by the TaskCompletionSource object via the following methods:

public class TaskCompletionSource<TResult> 
{ 
 public void SetResult(TResult result); 
 public void SetException (Exception exception); 

 public void SetCanceled();   
 public bool TrySetResult (TResult result); 
 public bool TrySetException (Exception exception); 
 public bool TrySetCanceled();
 ... 
}

Calling any of these methods signals the task, putting it into a completed, faulted, or canceled state (we'l cover the latter in the section "Cancellation"). You'e supposed to call one of these methods exactly once: if called again, SetResult, SetException, or SetCanceled will throw an exception, whereas the Try* methods return false.

The following example prints 42 after waiting for five seconds:

var tcs = new TaskCompletionSource<int>();
new Thread (() =>     {
                       Thread.Sleep (5000); 
                       tcs.SetResult (42); 
                      })    
           .Start();   
Task<int> task = tcs.Task;    // Our "slave" task. 
Console.WriteLine(task.Result);  // 42

Other interesting quotes

The real power of TaskCompletionSource is in creating tasks that don't tie up threads.

.. and later on

Our use of TaskCompletionSource without a thread means that a thread is engaged only when the continuation starts, five seconds later. We can demonstrate this by starting 10,000 of these operations at once without error or excessive resource consumption: