So I was told recently that how I was using my .ContinueWith for Tasks was not the proper way to use them. I have yet to find evidence of this on the internet so I will ask you guys and see what the answer is. Here is an example of how I use .ContinueWith:
public Task DoSomething()
{
return Task.Factory.StartNew(() =>
{
Console.WriteLine("Step 1");
})
.ContinueWith((prevTask) =>
{
Console.WriteLine("Step 2");
})
.ContinueWith((prevTask) =>
{
Console.WriteLine("Step 3");
});
}
Now I know this is a simple example and it will run very fast, but just assume each task does some longer operation. So, what I was told is that in the .ContinueWith, you need to say prevTask.Wait(); otherwise you could do work before the previous task finishes. Is that even possible? I assumed my second & third task would only run once their previous task finishes.
What I was told how to write the code:
public Task DoSomething()
{
return Task.Factory.StartNew(() =>
{
Console.WriteLine("Step 1");
})
.ContinueWith((prevTask) =>
{
prevTask.Wait();
Console.WriteLine("Step 2");
})
.ContinueWith((prevTask) =>
{
prevTask.Wait();
Console.WriteLine("Step 3");
});
}
Ehhh.... I think some of the current answers are missing something: what happens with exceptions?
The only reason you would call Wait
in a continuation would be to observe a potential exception from the antecedent in the continuation itself. The same observation would happen if you accessed Result
in the case of a Task<T>
and also if you manually accessed the Exception
property.
Frankly, I wouldn't call Wait
or access Result
because if there is an exception you'll pay the price of re-raising it which is unnecessary overhead. Instead you can just check the IsFaulted
property off the antecedent Task
. Alternatively you can create forked workflows by chaining on multiple sibling continuations that only fire based on either success or failure with TaskContinuationOptions.OnlyOnRanToCompletion
and TaskContinuationOptions.OnlyOnFaulted
.
Now, it's not necessary to observe the exception of the antecedent in the continuation, but you may not want your workflow to move forward if, say, "Step 1" failed. In that case: specifying TaskContinuationOptions.NotOnFaulted
to your ContinueWith
calls would prevent the continuation logic from ever even firing.
Keep in mind that, if your own continuations don't observe the exception, the person who is waiting on this overall workflow to complete is going to be the one to observe it. Either they're Wait
ing on the Task
upstream or have tacked on their own continuation to know when it is complete. If it is the latter, their continuation would need to use the aforementioned observation logic.
You are using it correctly.
Creates a continuation that executes asynchronously when the target
Task completes.
Source: Task.ContinueWith Method (Action as MSDN)
Having to call prevTask.Wait()
in every Task.ContinueWith
invocation seems like a weird way to repeat unnecessary logic - i.e. doing something to be "super duper sure" because you actually don't understand what a certain bit of code does. Like checking for a null just to throw an ArgumentNullException
where it would've been thrown anyway.
So, no, whoever told you that is wrong and probably doesn't understand why Task.ContinueWith
exists.
Who told you that?
Quoting MSDN:
Creates a continuation that executes asynchronously when the target
Task completes.
Also, what would be the purpose of Continue With if it wasn't waiting for the previous task to be completed?
You can even test it by yourself:
Task.Factory.StartNew(() =>
{
Console.WriteLine("Step 1");
Thread.Sleep(2000);
})
.ContinueWith((prevTask) =>
{
Console.WriteLine("I waited step 1 to be completed!");
})
.ContinueWith((prevTask) =>
{
Console.WriteLine("Step 3");
});
From the MSDN on Task.Continuewith
The returned Task will not be scheduled for execution until the
current task has completed. If the criteria specified through the
continuationOptions parameter are not met, the continuation task will
be canceled instead of scheduled.
I think that the way you expect it to work in the first example is the correct way.
You might also want to consider using Task.Run instead of Task.Factory.StartNew.
Stephen Cleary's blog post and the Stephen Toub's post that he references explain the differences. There is also a discussion in this answer.
By Accessing Task.Result
you are actually doing similar logic to task.wait
I will reiterate what many have spoken already, prevTask.Wait()
is unnecessary.
For more examples one can go to Chaining Tasks using Continuation Tasks, yet another link by Microsoft with good examples.