I only want to use CancellationTokenSource for timeout and cancellation handling.
How to distinguish if a TaskCanceledException occured due to a timeout or due to manual cancellation?
Here is a simplified example. In the real program I neither know if CancellationTokenSource .CancelAfter() was used nor if someone called CancellationTokenSource.Cancel()
static CancellationTokenSource cts = new CancellationTokenSource();
static void Main(string[] args)
{
Task.Run(async () =>
{
try
{
await SomeClass.DoSomething(cts.Token);
}
catch (TaskCanceledException ex)
{
//How to find out if the exception occured due to timeout or a call to cts.Cancel()
}
});
while (true)
{
Thread.Sleep(100);
if (someCondition)
cts.Cancel();
}
}
public class SomeClass
{
public static async Task DoSomething(CancellationToken ct)
{
using (var innerCts = CancellationTokenSource.CreateLinkedTokenSource(ct))
{
innerCts.CancelAfter(1000);
//Simulate some operation
await Task.Delay(10000, innerCts.Token);
}
}
}
thanks
Tom
AFAIK this is the most commonly used pattern:
Task.Run(async () =>
{
try
{
await SomeClass.DoSomething(cts.Token);
}
catch (OperationCanceledException) when (cts.IsCancellationRequested)
{
// cts cancellation occurred
}
catch (OperationCanceledException)
{
// Timeout occurred
}
});
Another idea is to change the implementation of the SomeClass.DoSomething method, assuming that you are allowed to do it, so that in case of timeout it throws a TimeoutException instead of an OperationCanceledException.
public static async Task DoSomething(CancellationToken cancellationToken)
{
using var innerCts = new CancellationTokenSource(millisecondsDelay: 1000);
using var linkedCts = CancellationTokenSource
.CreateLinkedTokenSource(cancellationToken, innerCts.Token);
try
{
// Simulate some operation
await Task.Delay(10000, linkedCts.Token);
}
catch (OperationCanceledException) when (innerCts.IsCancellationRequested)
{
throw new TimeoutException();
}
}
Related
When our connection drops, ReceiveAsync is throwing WebSocketException (ex.WebSocketErrorCode == WebSocketError.ConnectionClosedPrematurely).
The issue is that it's not handled by Polly for some reason. I believe it doesn't handle it, because it's in a separate Task, although I'm doing Task.WhenAny.
The expected behavior is to trigger the reconnect if WebSocketException is thrown.
public sealed class ChannelWebSocketClient : IDisposable
{
private readonly Uri _uri;
private readonly ILogger<ChannelWebSocketClient> _logger;
private readonly Channel<string> _output;
private CancellationTokenSource? _cancellationTokenSource;
public ChannelWebSocketClient(Uri uri, ILoggerFactory loggerFactory)
{
_uri = uri ?? throw new ArgumentNullException(nameof(uri));
_logger = loggerFactory.CreateLogger<ChannelWebSocketClient>();
_output = Channel.CreateUnbounded<string>(new UnboundedChannelOptions
{
SingleReader = true,
SingleWriter = false
});
}
public void Dispose()
{
_output.Writer.TryComplete();
}
public Task StartAsync()
{
return Policy.Handle<Exception>(ex => ex is not (TaskCanceledException or OperationCanceledException))
.WaitAndRetryForeverAsync(
(_, _) => TimeSpan.FromSeconds(5),
(ex, retryCount, calculatedWaitDuration, _) => { _logger.LogError(ex, "Unable to connect to the web socket server. Retry count: {RetryCount} | Retry in {Seconds} seconds", retryCount, calculatedWaitDuration.TotalSeconds); })
.ExecuteAsync(ConnectAsync);
}
public void Stop()
{
_cancellationTokenSource?.Cancel();
}
private async Task ConnectAsync()
{
_logger.LogDebug("Connecting");
using var ws = new ClientWebSocket();
// WebSocketException, TaskCanceledException
await ws.ConnectAsync(_uri, CancellationToken.None).ConfigureAwait(false);
_logger.LogDebug("Connected to {Host}", _uri.AbsoluteUri);
_cancellationTokenSource = new CancellationTokenSource();
var receiving = ReceiveLoopAsync(ws, _cancellationTokenSource.Token);
var sending = SendLoopAsync(ws, _cancellationTokenSource.Token);
var trigger = await Task.WhenAny(receiving, sending).ConfigureAwait(false);
if (trigger == receiving)
{
_cancellationTokenSource?.Cancel();
await sending.ConfigureAwait(false);
}
_logger.LogDebug("END");
}
public async Task SendAsync(string message)
{
await _output.Writer.WriteAsync(message, CancellationToken.None).ConfigureAwait(false);
}
private async Task SendLoopAsync(WebSocket webSocket, CancellationToken cancellationToken)
{
_logger.LogDebug("SendLoopAsync BEGIN");
try
{
while (await _output.Reader.WaitToReadAsync(cancellationToken).ConfigureAwait(false))
{
while (_output.Reader.TryRead(out var message))
{
// WebSocketException, TaskCanceledException, ObjectDisposedException
await webSocket.SendAsync(new ArraySegment<byte>(Encoding.UTF8.GetBytes(message)),
WebSocketMessageType.Text, true, cancellationToken).ConfigureAwait(false);
}
}
}
catch (OperationCanceledException)
{
}
finally
{
_logger.LogDebug("SendLoopAsync END");
}
}
private async Task ReceiveLoopAsync(WebSocket webSocket, CancellationToken cancellationToken)
{
_logger.LogDebug("ReceiveLoopAsync BEGIN");
try
{
while (true)
{
ValueWebSocketReceiveResult receiveResult;
using var buffer = MemoryPool<byte>.Shared.Rent(4096);
await using var ms = new MemoryStream(buffer.Memory.Length);
do
{
// WebSocketException, TaskCanceledException, ObjectDisposedException
receiveResult = await webSocket.ReceiveAsync(buffer.Memory, cancellationToken).ConfigureAwait(false);
if (receiveResult.MessageType == WebSocketMessageType.Close)
{
break;
}
await ms.WriteAsync(buffer.Memory[..receiveResult.Count], cancellationToken).ConfigureAwait(false);
} while (!receiveResult.EndOfMessage);
ms.Seek(0, SeekOrigin.Begin);
if (receiveResult.MessageType == WebSocketMessageType.Text)
{
using var reader = new StreamReader(ms, Encoding.UTF8);
var message = await reader.ReadToEndAsync().ConfigureAwait(false);
_logger.LogInformation("Message received: {Message}", message);
}
else if (receiveResult.MessageType == WebSocketMessageType.Close)
{
break;
}
}
}
catch (WebSocketException ex) when (ex.WebSocketErrorCode == WebSocketError.ConnectionClosedPrematurely)
{
_logger.LogError(ex, "");
throw;
}
finally
{
_logger.LogDebug("ReceiveLoopAsync END");
}
}
}
The Task.WhenAll works differently than Task.WhenAny.
Former throws exception is any of the tasks failed with an exception
Latter does not throw exception even if all of the tasks fail
So either you use call two twice the .GetAwaiter().GetResult() since WhenAny returns a Task<Task>
Task.WhenAny(receiving, sending).ConfigureAwait(false)
.GetAwaiter().GetResult()
.GetAwaiter().GetResult();
Or you can re-throw the exception
var trigger = await Task.WhenAny(receiving, sending).ConfigureAwait(false);
if (trigger.Exception != null)
{
throw trigger.Exception;
}
None of these solutions is perfect, but they will trigger your policy.
UPDATE #1
As Monsieur Merso pointed out you can call twice await
await await Task.WhenAny(receiving, sending).ConfigureAwait(false);
This is much better than the above two approaches.
UPDATE #2
If you want to
trigger the policy if faster task failed
or want to know which one has finished sooner with success
then you can "avoid" the double await
var trigger = await Task.WhenAny(receiving, sending).ConfigureAwait(false);
await trigger; //Throws exception if the faster Task has failed
if (trigger == receiving) //Determines which Task finished sooner
{
}
I don't think this question is a duplicate of "Proper way to deal with exceptions in DisposeAsync".
Let's say my class that implements IAsynsDisposable because it has a long-running background task, and DisposeAsync terminates that task. A familiar pattern might be the Completion property, e.g. ChannelReader<T>.Completion (despite ChannelReader doesn't implement IAsynsDisposable).
Is it considered a good practice to propagate the Completion task's exceptions outside DisposeAsync?
Here is a complete example that can be copied/pasted into a dotnet new console project. Note await this.Completion inside DisposeAsync:
try
{
await using var service = new BackgroundService(TimeSpan.FromSeconds(2));
await Task.Delay(TimeSpan.FromSeconds(3));
}
catch (Exception ex)
{
Console.WriteLine(ex);
Console.ReadLine();
}
class BackgroundService: IAsyncDisposable
{
public Task Completion { get; }
private CancellationTokenSource _diposalCts = new();
public BackgroundService(TimeSpan timeSpan)
{
this.Completion = Run(timeSpan);
}
public async ValueTask DisposeAsync()
{
_diposalCts.Cancel();
try
{
await this.Completion;
}
finally
{
_diposalCts.Dispose();
}
}
private async Task Run(TimeSpan timeSpan)
{
try
{
await Task.Delay(timeSpan, _diposalCts.Token);
throw new InvalidOperationException("Boo!");
}
catch (OperationCanceledException)
{
}
}
}
Alternatively, I can observe service.Completion explicitly in the client code (and ignore its exceptions inside DiposeAsync to avoid them being potentially thrown twice), like below:
try
{
await using var service = new BackgroundService(TimeSpan.FromSeconds(2));
await Task.Delay(TimeSpan.FromSeconds(3));
await service.Completion;
}
catch (Exception ex)
{
Console.WriteLine(ex);
Console.ReadLine();
}
class BackgroundService: IAsyncDisposable
{
public Task Completion { get; }
private CancellationTokenSource _diposalCts = new();
public BackgroundService(TimeSpan timeSpan)
{
this.Completion = Run(timeSpan);
}
public async ValueTask DisposeAsync()
{
_diposalCts.Cancel();
try
{
await this.Completion;
}
catch
{
// the client should observe this.Completion
}
finally
{
_diposalCts.Dispose();
}
}
private async Task Run(TimeSpan timeSpan)
{
try
{
await Task.Delay(timeSpan, _diposalCts.Token);
throw new InvalidOperationException("Boo!");
}
catch (OperationCanceledException)
{
}
}
}
Is there a concensus about which option is better?
For now, I've settled on a reusable helper class LongRunningAsyncDisposable (here's a gist, warning: barely tested yet), which allows:
to start a background task;
stop this task (via a cancellation token) by calling IAsyncDisposable.DisposeAsync at any time, in a thread-safe, concurrency-friendly way;
configure whether DisposeAsync should re-throw the task's exceptions (DisposeAsync will await the task's completion either way, before doing a cleanup);
observe the task's status, result and exceptions at any time via LongRunningAsyncDisposable.Completion property.
I want to restart a Task in case it is already running using a CancellationToken. Here is my code:
CancellationTokenSource cancellationTokenSource;
CancellationToken cancellationToken;
private void OnButtonCLick(object sender, RoutedEventArgs e)
{
cancellationTokenSource?.Cancel();
cancellationTokenSource = new CancellationTokenSource();
cancellationToken = cancellationTokenSource.Token;
Task.Run(() =>
{
try
{
while (true)
{
cancellationToken.ThrowIfCancellationRequested();
Console.WriteLine("test");
// ... calculation lasting few milliseconds ...
}
}
catch (OperationCanceledException ex) { }
}, cancellationToken);
}
For some reason the OperationCanceledException does not get thrown when calling ThrowIfCancellationRequested() gets called. I compared it to other examples but don't see a reason why this does not work. What am I doing wrong here?
Your task is reading the cancellationToken field, which will point to the new (not yet cancelled) token. Instead, you must make sure to keep using the one that was originally given to your task:
var token = cancellationToken;
Task.Run(() =>
{
try
{
while (true)
{
token.ThrowIfCancellationRequested();
Console.WriteLine("test");
// ... calculation lasting few milliseconds ...
}
}
catch (OperationCanceledException ex) { }
}, token);
We start using CancellationToken in out app a lot, so we have to change exception handling correspondingly:
class Program
{
static async Task Main(string[] args)
{
CancellationTokenSource cts = new CancellationTokenSource(100);
await DoJob(cts.Token);
Console.WriteLine("Successfully finished");
}
private static async Task DoJob(CancellationToken ct)
{
try
{
await Task.Delay(1000, ct);
}
catch (Exception e) when(!(e is OperationCanceledException))
{
Console.WriteLine("Do cleanup in case of error.");
}
}
}
The idea behind this code is that if someone use catch(Exception e) (please do not blame me for this) and forgot to exclude CancellationToken, an error handling is executed, for example, there is a log that operation failed. But it is not true, id doesn't fail, it just has been canceled. And cancellation should be handled differently then failure.
It seems to me like a big boilerplate to write practically in every general catch
catch (Exception e) when(!(e is OperationCanceledException))
Is there some more robust solution with less boilerplate?
You could create a method that accepts a Func<Task> and catches the exception(s), e.g.:
class Program
{
static async Task Main(string[] args)
{
CancellationTokenSource cts = new CancellationTokenSource(100);
await GeneralDoJobAndCatchException(() => DoJob(cts.Token));
Console.WriteLine("Successfully finished");
}
private static async Task GeneralDoJobAndCatchException(Func<Task> func)
{
try
{
await func();
}
catch (OperationCanceledException) { }
catch (Exception e)
{
Console.WriteLine("Do error handling");
}
}
private static async Task DoJob(CancellationToken ct)
{
await Task.Delay(1000, ct);
}
}
We're having the exact same problem. Mainly there is a while-loop that checks for the CancellationToken but you've to catch this exception.
We created the following extension method:
public static async Task<TaskStatus> HideCancellationException(this Task task)
{
try
{
await task;
return task.Status;
}
catch (OperationCanceledException)
{
return TaskStatus.Canceled;
}
}
Having this extension method allows to change this code:
while (!cancellationToken.IsCancellationRequested)
{
// do stuff here...
try
{
await Task.Delay(..., cancellationToken);
}
catch (OperationCanceledException)
{
// expected
}
}
to something like that:
while (!cancellationToken.IsCancellationRequested)
{
// Do stuff here.
await Task.Delay(..., cancellationToken).HideCancellationException();
}
Keep in mind that there is explicitly no overload for Task<T> because the return value in case of cancellation is default. You can't distinguish between default as normal task result and default as result of cancellation. In that case it's better to catch the exception.
You could get rid of the try-catch block altogether by awaiting indirectly with Task.WhenAny, and then querying the status of the completed task:
private static async Task DoJob(CancellationToken ct)
{
var completedTask = await Task.WhenAny(Task.Delay(1000, ct));
if (completedTask.IsFaulted)
{
Console.WriteLine("Error: " + completedTask.Exception.InnerException);
}
else if (completedTask.IsCanceled)
{
// Do nothing
}
else // Success
{
// Do nothing
}
}
This question has already been asked but I still don't get it... I'm trying to cancel a task but when I include ThrowIfCancellationRequested(), it is not catched and if I don't include it, the GUI freezes...
Here is my code that run without the ThrowIfCancellationRequested() but freeze the GUI:
public void StartProcess(double myvariable)
{
tokenSource = new CancellationTokenSource();
CancellationToken token = tokenSource.Token;
processThread = Task.Factory.StartNew(() =>
{
while (true)
{
//Do some work with myvariable
if (token.IsCancellationRequested)
{
break;
}
}
}, token);
}
And the code with the ThrowIfCancellationRequested() that is not catched, the debugger stops on the line token.ThrowIfCancellationRequested():
public void StartProcess(double myvariable)
{
tokenSource = new CancellationTokenSource();
CancellationToken tokenDispatcher = tokenSource.Token;
processThread = Task.Factory.StartNew(() =>
{
try
{
while (true)
{
//Do some work with myvariable
if (token.IsCancellationRequested)
{
token.ThrowIfCancellationRequested();
break;
}
}
}
catch (AggregateException ae)
{
if (ae.InnerException is OperationCanceledException)
{
Console.WriteLine("Process Thread Cancelled");
}
else
{
Console.WriteLine("ERROR");
}
}
}, token);
}
What am I doing wrong?
The ThrowIfCancellationRequested is the proper way to cancel the task. What you need to understand is that there will be no exception when you cancel a Task. Only the Status of the task will be set to Canceled. The try catch block you have inside the task will do nothing.
To get an exception you need to await the Task in an async method or use the Wait method to wait for the finish of the task. Using the Wait method is not recommended because you are blocking the current thread until the task finishes.
Here is a suggestion how you could implement this from your code.
public static Task StartProcess(double myvariable)
{
tokenSource = new CancellationTokenSource();
CancellationToken token = tokenSource.Token;
return Task.Factory.StartNew(() =>
{
while (true)
{
//Do some work with myvariable
token.ThrowIfCancellationRequested();
}
}, token);
}
public async Task ExecuteProcess(double myvariable)
{
try
{
await StartProcess(myvariable);
}
catch (OperationCanceledException ex)
{
Console.WriteLine("Process Thread Canceled");
}
catch (Exception ex)
{
Console.WriteLine("ERROR");
}
}
The ExecuteProcess method demonstrates how you need to call your function to get the exception. This way you also prevent the GUI from blocking if you call the method from the UI thread.
I recommend you read the documentation on asynchronous programming to get more understanding on how it works.