C# provides two ways of creating asynchronous methods:
Task()
:
static Task<string> MyAsyncTPL() {
Task<string> result = PerformWork();
return result.ContinueWith(t => MyContinuation());
}
async Task()
:
static async Task<string> MyAsync() {
string result = await PerformWork();
return MyContinuation();
}
Both of the above methods are async
and achieve the same thing. So, when should I choose one method over the other? Are there any guidelines or advantages of using one over the other?
I recommend you use await
rather than ContinueWith
. While - at a high level - they are very similar, they also have different default behavior.
When you use ContinueWith
, you are choosing a lower-level abstraction. In particular, here are some "danger points", and this is why I don't recommend using ContinueWith
unless the method is really simple (or your name is Stephen Toub):
async Task
methods are placed on the returned task; exceptions raised from non-async
methods are propagated directly.await
will by default will resume the async
method in the same "context". This "context" is SynchronizationContext.Current
unless it is null
, in which case it is TaskScheduler.Current
. This means that if you call MyAsync
on a UI thread (or within an ASP.NET request context), then MyContinuation
will also execute on the UI thread (or in that same ASP.NET request context). I explain this more on my blog.ContinueWith
; otherwise, it will pick up TaskScheduler.Current
, which can cause surprising behavior. I describe this problem in detail on my blog. That post is about StartNew
; but ContinueWith
has the same "non-default default scheduler" problem described in that post.await
uses appropriate behavior and optimization flags that are not set by default in ContinueWith
. For example, it uses DenyChildAttach
(to ensure asynchronous tasks are not mistakenly used as parallel tasks) and ExecuteSynchronously
(an optimization).In short, the only reason to use ContinueWith
for asynchronous tasks is to save an extremely small amount of time and memory (by avoiding the async
state machine overhead), and in exchange your code is less readable and maintainable.
With an extremely simple example, you might get away with it; but as Jon Skeet pointed out, as soon as you have loops the ContinueWith
code simply explodes in complexity.