I am calling Task.Run(() => DoSomething()).Result which causes the UI to freeze and it happens because am using ".Result". I need Result because i want to return the value.
I don't want the Method StartSomething to be async because I don't want to await the method StartSomething. I want the await to happen at DoSomething().
So basically I need a asynchronous method to be called by a synchronous method, without freezing the UI. Plus I want to return the value from the async method to the top level that is on Button Click.
Can this code be improved or is there any other solution?
private TaskCompletionSource<bool> TaskCompletion = null;
private void Button_Click(object sender, RoutedEventArgs e)
{
bool k = StartSomething();
}
private bool StartSomething()
{
return Task.Run(() => DoSomething()).Result;
}
private async Task<bool> DoSomething()
{
TaskCompletion = new TaskCompletionSource<bool>();
await Task.WhenAny(TaskCompletion.Task, Task.Delay(3000));
MessageBox.Show("DoSomething");
return true;
}
Method StartSomething() doesn't make sense to me. It starts a new Task and then just synchronously waits for the result (.Result) of this task, which is effectively useless - it is nearly [*] the same as calling DoSomething() directly. Also DoSomething() is already asynchronous so you don't need to start a new Task for it.
It looks like you don't need StartSomething() method at all. If you make Button_Click handler async, you can then simply await DoSomething() directly:
private TaskCompletionSource<bool> TaskCompletion = null;
private async void Button_Click(object sender, RoutedEventArgs e)
{
bool k = await DoSomething();
}
private async Task<bool> DoSomething()
{
TaskCompletion = new TaskCompletionSource<bool>();
await Task.WhenAny(TaskCompletion.Task, Task.Delay(3000));
MessageBox.Show("DoSomething");
return true;
}
Edit:
While using async all the way down solution (as shown above) is IMO the preferred way, if you really can't change calling code to async, I can think of two ways to call async method from synchronous method without blocking UI. First is to manually set up a continuation tasks like this:
private void Button_Click(object sender, RoutedEventArgs e)
{
DoSomething().ContinueWith((task) =>
{
bool k = task.Result;
// use the result
},
// TaskScheduler argument is needed only if the continuation task
// must run on the UI thread (eg. because it access UI elements).
// Otherwise this argument can be omitted.
TaskScheduler.FromCurrentSynchronizationContext());
// Method can exit before DoSomething().Result becomes
// available, which keep UI responsive
}
So you basicly split synchronous method (one split instead of each await) into several parts (continuation lambda methods) linked by .ContinueWith. This is similar to what await does under a hood. Problem is that unlike await (which produces nice and clean code), your code will be full of these continuation lambdas. And it will get much worse when you add exception handling blocks, using blocks, etc.
The second approach is using nested loops, eg. Stephen Toub's WaitWithNestedMessageLoop extension method:
static T WaitWithNestedMessageLoop<T>(this Task<T> task)
{
var nested = new DispatcherFrame();
task.ContinueWith(_ => nested.Continue = false, TaskScheduler.Default);
Dispatcher.PushFrame(nested);
return task.Result;
}
Nested loops are quite advanced technique (I actually never used it) and I don't recommend using it unless you have to.
[*] There are differences in exception handling, executing thread, etc., but these are not relevant to this question.
Related
I'm buildind a synchronization tool that's working pretty well, but a noise problem's that I've an heavy method that freeze the UI.
Now, I know that I can solve this situation with Thread or Task, but in the Sync(); method I've a lot of access to the UI control.
Actually for access to MainWindow control I've created something like this:
public static MainWindow AppWindow;
public MainWindow()
{
InitializeComponent();
AppWindow = this;
}
so from any class I can simply do this: MainWindow.AppWindow.UIControlName. As you can imagine, if I use Task or Thread, this solution will become useless. 'cause I'll need to write this for each control:
MainWindow.AppWindow.UIControlName.Dispatcher.BeginInvoke(
(Action)(() => {
MainWindow.AppWindow.UIControlName.Content = "Synchronization in progress";
}));
Imagine that I have 20 or more of these lines, the code would become really ugly looking and long, for easy access to the UI control.
Now I wonder what is the best way to call a method asynchronously and return immediately to the caller without blocking the UI.
Without an asynchronous programming for this kind of applications the user experience is not really write home about.
You can either wrap all ugly stuff with dispatcher in separate methods and use those, or use async\await like this:
private async void ButtonClicked(object sender, RoutedEventArgs e) {
var result = await LongRunningOperation();
MainWindow.AppWindow.UIControlName.Content = "Synchronization in progress";
await YetAnotherLongRunningOperation();
// update again
}
Where LongRunningOperation is:
public async Task<SomeResult> LongRunningOperation() {
// do some stuff here
return new SomeResult();
}
or
public async Task LongRunningOperation() {
// do some stuff
}
You can also do it like this:
private async void ButtonClicked(object sender, RoutedEventArgs e) {
var result = await Task.Run(() => LongRunningOperation());
MainWindow.AppWindow.UIControlName.Content = "Synchronization in progress";
await YetAnotherLongRunningOperation();
// update again
}
That is because by default before await current synchronization context is captured, and after await this context is restored. In UI applications (like WPF or WinForms) that means if before await you were on UI thread - you will be there after await too, and as such can freely access UI controls.
You might ask - when should you declare your methods async and when just use Task.Run. If you do pure CPU computations - you can use Task.Run. If you do any IO (file access, network access, etc) - declare your method async and use async methods provided by IO-related classes (like FileStream.ReadAsync and such).
For example:
private async Task LongRunningOperation() {
string contents;
using (var fs = File.OpenRead("some file")) {
using (var reader = new StreamReader(fs)) {
contents = await reader.ReadToEndAsync();
}
}
var downloadedFile = await new WebClient().DownloadDataTaskAsync("some file url");
}
In CefSharp WinForms, I'm trying to get the html source of the page using JS once the page has loaded, however the application is freezing. I'm using a BackgroundWorker and the concerned functions are as follows:
void bw_DoWork(object sender, DoWorkEventArgs e)
{
browser.Load("http://www.google.com");
browser.FrameLoadEnd += delegate
{
object js = EvaluateScript(browser, "1+1");
MessageBox.Show(js.ToString());
};
}
object EvaluateScript(ChromiumWebBrowser b, string script)
{
var task = b.EvaluateScriptAsync(script);
task.Wait();
return task.Result;
}
As amaitland pointed out, FrameLoadEnd was causing the hang by running in the UI thread. The below code is working:
void bw_DoWork(object sender, DoWorkEventArgs e)
{
first.Load("http://www.google.com");
browser.FrameLoadEnd += delegate
{
Task task = new Task(() => {
object js = EvaluateScript(browser, "document.getElementsByTagName('html')[0].innerHTML;");
MessageBox.Show(js.ToString());
});
task.Start();
};
}
static object EvaluateScript(ChromiumWebBrowser b, string script)
{
var task = b.EvaluateScriptAsync(script);
task.Wait();
JavascriptResponse response = task.Result;
return response.Success ? (response.Result ?? "") : response.Message;
}
Whilst you assign FrameLoadEnd in the BackgroundWorker thread, it's actually executed on the underlying CEF UI thread, for which you cannot block without issues.
I'd typically suggest you spawn a Task from within the event handler to complete your work.
As a general rule, it's a bad idea to use Task.Wait on async code; rather, you should use async "all the way down". See also Don't Block on Async Code by Stephen Cleary. The short answer is that if you do this in an application with a synchronization context, you can cause a circular wait for the synchronization context (and hence a deadlock). The article I linked to has several examples of this, but I'd strongly suggest replacing Task.Wait with await here if possible.
I'm confused why the output of these 2 programs differs:
private async void button1_Click(object sender, EventArgs e)
{
for (int i = 0; i < 33; i++)
{
await LongProcess();
}
}
private async Task LongProcess()
{
await Task.Delay(1000);
progressBar1.Value += 3;
}
and
private async void button1_Click(object sender, EventArgs e)
{
for (int i = 0; i < 33; i++)
{
await Task.Run(() => LongProcess());
}
}
private async void LongProcess()
{
await Task.Delay(1000);
progressBar1.Value += 3;
}
I realize the first example returning Task is more correct, but I don't understand why wrapping the void function in a Task.Run doesn't produce the same output? The first function does what I expect, updates the progress bar every 1 second. The second code attempts to update the progress bar all at once, causing problems attempting to update the same UI element from multiple threads.
My assumption was since the buttonClick method awaits the long process to complete, both sets of code should not allow the progressBar1 update to happen until the previous process has completed. Why does the second set of code allow it to happen all at once?
This isn't doing what you think it is:
await Task.Run(() => LongProcess());
The code is awaiting Task.Run(), but nothing within that task is awaiting LongProcess(). So Task.Run() returns immediately in this case.
This is actually an interesting illustration of a failure to be "async all the way down", because the inline function is essentially hiding the fact that it's async. And in fact the compiler should be warning you that nothing is awaiting LongProcess() and that it would return immediately.
Contrast it with this:
await Task.Run(async () => await LongProcess());
Edit: I just noticed why the compiler probably isn't warning you. Because of this:
async void
Never, ever, ever do this :) (Well, ok, there's one valid reason to do this. And I'm sure it haunts the C# team to this day that they had to support it just for that one reason. But unless you encounter that one reason, don't do it.)
Always return a Task for async methods so that the method can be awaited.
In the first program, LongProcess returns a Task, and Task.Run is wrapping it -- basically just launching it in the default scheduler rather than whatever context you're currently on.
You'll notice that Task.Run has overloads specifically to do this wrapping, and is not returning a Task<Task>.
In the second program, LongProcess is an async void method, which means Task.Run has nothing to wrap and will complete more or less immediately, and before the work is guaranteed to be done.
I have a WebBrowser control and it has InvokeScript method, which you should call only after WebBrowser is loaded.
So I've tried something like this:
private readonly ManualResetEventSlim browserLoaded = new ManualResetEventSlim(false);
private void BrowserLoaded(object sender, NavigationEventArgs navigationEventArgs)
{
browserLoaded.Set();
}
private async Task<object> InvokeScript(string invoke, object[] parameters = null)
{
return await Task.Factory
.StartNew(() =>
{
if (!browserLoaded.Wait(TimeSpan.FromSeconds(10)))
{
throw new Exception("Timeout for waiting browser to load.");
}
})
.ContinueWith(task => parameters == null
? browser.InvokeScript(invoke)
: browser.InvokeScript(invoke, parameters), TaskScheduler.FromCurrentSynchronizationContext());
}
It does not look very nice to me, but works ok when called asynchronously.
Problem appears, when I try to read result value synchronously - app just hangs:
private string GetEnteredText()
{
return (string)InvokeScript("getEnteredText").Result;
}
I know, that I should go all the way async, but I'm wondering what to do with properties:
public override string UserText
{
get
{
return GetEnteredText();
}
set
{
SetEnteredText(value);
}
}
Or async is wrong way to go in this case at all?
Update
Property is a 'glue' between input field value in browser's page and view model in WPF, so I don't see a good way to make it as separate methods, especially because it is a part of the bigger framework (notice override keyword on it).
Once browser control is loaded, execute logic should not take long, I guess less than 10 milliseconds, that is why I would be ok with sync execution in this case. And usually browser control loads fast enough, the only reason here to delay is to make sure InvokeScript is not called before load, not because it taking long time or smth.
app just hangs
We'll, you said it yourself. You know why that happens. You're blocking on the call using Task.Result, that is what causes the deadlock.
Or async is wrong way to go in this case at all?
We don't have async properties. The reason we don't have them are because properties aren't asynchronous by nature. Stephan Cleary describes it nicely:
If your “property” needs to be asynchronously evaluated every time
it’s accessed, then you’re really talking about an asynchronous
operation.The best solution is to change the property to an async
method. Semantically, it shouldn’t be a property.
Instead of a property, make it an asynchronous method which you can await properly.
In regards to using a ManualResetEvent, I would use a TaskCompletionSource<bool> instead. It works "nicer" with TPL:
private TaskCompletionSource<bool> tcs = new TaskCompletionSource<bool>();
private void BrowserLoaded(object sender, NavigationEventArgs navigationEventArgs)
{
tcs.TrySetResult(true);
}
private async Task<object> InvokeScript(string invoke, object[] parameters = null)
{
var timeoutTask = Task.Delay(TimeSpan.FromSeconds(10));
if (timeoutTask == await Task.WhenAny(tcs.Task, timeoutTask))
{
// You've timed out;
}
return Task.Run(() =>
{
parameters == null ? browser.InvokeScript(invoke)
: browser.InvokeScript(invoke, parameters)
});
}
I also see that you used TaskScheduler.FromCurrentSynchronizationContext() in your continuation. If you need this to execute on the UI thread, there is no point in using a threadpool thread at all.
#Noseratio adds that WebBrowser is an STA object. His answer here might help.
I have a WPF GUI, where I want to press a button to start a long task without freezing the window for the duration of the task. While the task is running I would like to get reports on progress, and I would like to incorporate another button that will stop the task at any time I choose.
I cannot figure the correct way to use async/await/task. I can't include everything I've tried, but this is what I have at the moment.
A WPF window class :
public partial class MainWindow : Window
{
readonly otherClass _burnBabyBurn = new OtherClass();
internal bool StopWorking = false;
//A button method to start the long running method
private async void Button_Click_3(object sender, RoutedEventArgs e)
{
Task slowBurn = _burnBabyBurn.ExecuteLongProcedureAsync(this, intParam1, intParam2, intParam3);
await slowBurn;
}
//A button Method to interrupt and stop the long running method
private void StopButton_Click(object sender, RoutedEventArgs e)
{
StopWorking = true;
}
//A method to allow the worker method to call back and update the gui
internal void UpdateWindow(string message)
{
TextBox1.Text = message;
}
}
And a class for the worker method:
class OtherClass
{
internal Task ExecuteLongProcedureAsync(MainWindow gui, int param1, int param2, int param3)
{
var tcs = new TaskCompletionSource<int>();
//Start doing work
gui.UpdateWindow("Work Started");
While(stillWorking)
{
//Mid procedure progress report
gui.UpdateWindow("Bath water n% thrown out");
if (gui.StopTraining) return tcs.Task;
}
//Exit message
gui.UpdateWindow("Done and Done");
return tcs.Task;
}
}
This runs, but the WPF function window is still blocked once the worker method starts.
I need to know how to arrange the async/await/task declarations to allow
A) the worker method to not block the gui window
B) let the worker method update the gui window
C) allow the gui window to stop interrupt and stop the worker method
Any help or pointers are much appreciated.
Long story short:
private async void ButtonClickAsync(object sender, RoutedEventArgs e)
{
// modify UI object in UI thread
txt.Text = "started";
// run a method in another thread
await HeavyMethodAsync(txt);
// <<method execution is finished here>>
// modify UI object in UI thread
txt.Text = "done";
}
// This is a thread-safe method. You can run it in any thread
internal async Task HeavyMethodAsync(TextBox textBox)
{
while (stillWorking)
{
textBox.Dispatcher.Invoke(() =>
{
// UI operation goes inside of Invoke
textBox.Text += ".";
// Note that:
// Dispatcher.Invoke() blocks the UI thread anyway
// but without it you can't modify UI objects from another thread
});
// CPU-bound or I/O-bound operation goes outside of Invoke
// await won't block UI thread, unless it's run in a synchronous context
await Task.Delay(51);
}
}
Result:
started....................done
You need to know about (1) how to write async code (2) how to run UI operations in another thread and (3) how to cancel a task.
I'm not getting into (3) cancellation mechanism in this post. Just know that you can create a CancellationTokenSource, which gives you a CancellationToken which you can pass into any method. You cancel the source, all tokens will know.
async and await:
Basics of async and await
You can only await in an async method.
You can only await an awaitable object (i.e. Task, ValueTask, Task<T>, IAsyncEnumerable<T>, etc.) These objects wrap around the return type of an async method and await keyword unwraps them. (see Wrapping and Unwrapping section)
Asynchronous method names should always end with Async to increase readability and to prevent mistakes.
// Synchronous method:
TResult MethodName(params) { }
// Asynchronous method:
async Task<TResult> MethodNameAsync(params) { }
The magic of async and await
The async-await syntactic feature, uses a state-machine to let the compiler give up and take back the control over the awaited Task in an async method.
The execution waits at await for the task to finish and returns back its results, without blocking the main thread.
Task.Run queues a Task in the thread pool. (Unless the it's a pure operation.)
i.e. The async method does not run in another thread. async and await by themselves don't have anything to do with thread creation.
So
When you run a Task (e.g. Task.Run(action)) you (re)use a thread for that action. And you can put that task in an async method to control its flow. By putting async in the method signature you tell the compiler to use state-machine to control the flow of that method (this does not mean threading at all). And by awaiting the task you prevent the execution flow within that method from moving past the awaited statement without blocking UI thread. If you want to pass the flow onto the caller then the async method itself can become a Task so you'll be able to cascade the same pattern out into the caller and so forth:
async Task Caller() { await Method(); }
async Task Method() { await Inner(); }
async Task Inner() { await Task.Run(action); }
The event handler looks like the code below.
Two possible cases for presense of async in the signature of ExecuteLongProcedure (case 1 and 2) and MyButton_ClickAsync (case A and B) are explained:
private async void MyButton_ClickAsync(object sender, RoutedEventArgs e)
{
//queue a task to run on threadpool
// 1. if ExecuteLongProcedure is a normal method and returns void
Task task = Task.Run(()=>
ExecuteLongProcedure(this, intParam1, intParam2, intParam3)
);
// or
// 2. if ExecuteLongProcedure is an async method and returns Task
Task task = ExecuteLongProcedureAsync(this, intParam1, intParam2, intParam3);
// either way ExecuteLongProcedure is running asynchronously here
// the method will exit if you don't wait for the Task to finish
// A. wait without blocking the main thread
// -> requires MyButton_ClickAsync to be async
await task;
// or
// B. wait and block the thread (NOT RECOMMENDED AT ALL)
// -> does not require MyButton_ClickAsync to be async
task.Wait();
}
Async method return types:
Suppose you have the following declaration:
private async ReturnType MethodAsync() { ... }
If ReturnType is Task then await MethodAsync(); returns void
If ReturnType is Task<T> then await MethodAsync(); returns a value of type T
This is called Unwrapping, see the next section (Wrapping and Unrwapping).
If ReturnType is void you can't await it
If you try writing await MethodAsync();, you will get a compile error saying:
cannot await void
You can only fire and forget i.e. just call the method normally: MethodAsync(); and then go on with your life.
The MethodAsync execution will be synchronous, however since it has async it will allow you to take advantage of the magic, i.e. you can write await task within the method to control the flow of execution.
This is how WPF handles your button click event handler, obviously because your event handler returns void.
The return type of an async method must be void, Task, Task<T>, a task-like type, IAsyncEnumerable<T>, or IAsyncEnumerator<T>
Wrapping and Unrwapping:
Wrapping:
async methods wrap their return values in a Task.
E.g., this method wraps a Task around an int and returns it:
// async Task<int>
private async Task<int> GetOneAsync()
{
int val = await CalculateStuffAsync();
return val;
// returns an integer
}
Unwrapping:
To retrieve or unwrap the value which is wrapped inside a Task<>:
asynchronous option: await
synchronous option: task.Result or task.GetAwaiter().GetResult() or task.WaitAndUnwrapException() or read How to call asynchronous method from synchronous method in C#?
e.g. await unwraps the int out of the Task:
Task<int> task = GetOneAsync();
int number = await task;
//int <- Task<int>
Different ways to wrap and unwrap:
private Task<int> GetNumber()
{
Task<int> task;
task = Task.FromResult(1); // the correct way to wrap a quasi-atomic operation, the method GetNumber is not async
task = Task.Run(() => 1); // not the best way to wrap a number
return task;
}
private async Task<int> GetNumberAsync()
{
int number = await Task.Run(GetNumber); // unwrap int from Task<int>
// bad practices:
// int number = Task.Run(GetNumber).GetAwaiter().GetResult(); // sync over async
// int number = Task.Run(GetNumber).Result; // sync over async
// int number = Task.Run(GetNumber).Wait(); // sync over async
return number; // wrap int in Task<int>
}
Still confused? Read async return types on MSDN.
To unwrap a task result, Always try to use await instead of .Result otherwise there will be no asynchronous benefit but only asynchronous disadvantages. The latter is called "sync over async".
Note:
await is a asynchronous and is different from task.Wait() which is synchronous. But they both do the same thing which is waiting for the task to finish.
await is a asynchronous and is different from task.Result which is synchronous. But they both do the same thing which is waiting for the task to finish and unwrapping and returning back the results.
To have a wrapped value, you can always use Task.FromResult(1) instead of creating a new thread by using Task.Run(() => 1).
Task.Run is newer (.NetFX4.5) and simpler version of Task.Factory.StartNew
WPF GUI:
This is where I explain how to run UI operations in another thread.
Blocking:
First thing you need to know about WPF async event handlers is that the Dispatcher will provide a synchronization context. Explained here
CPU-bound or IO-bound operations such as Sleep and task.Wait() will block and consume the thread even if they are called in a method with async keyword. but await Task.Delay() tells the state-machine to stop the flow of execution on the thread so it does not consume it; meaning that the thread resources can be used elsewhere:
private async void Button_Click(object sender, RoutedEventArgs e)
{
Thread.Sleep(1000);//stops, blocks and consumes threadpool resources
await Task.Delay(1000);//stops without consuming threadpool resources
Task.Run(() => Thread.Sleep(1000));//does not stop but consumes threadpool resources
await Task.Run(() => Thread.Sleep(1000));//literally the WORST thing to do
}
Thread Safety:
If you have to access GUI asynchronously (inside ExecuteLongProcedure method), invoke any operation which involves modification to any non-thread-safe object. For instance, any WPF GUI object must be invoked using a Dispatcher object which is associated with the GUI thread:
void UpdateWindow(string text)
{
//safe call
Dispatcher.Invoke(() =>
{
txt.Text += text;
});
}
However, If a task is started as a result of a property changed callback from the ViewModel, there is no need to use Dispatcher.Invoke because the callback is actually executed from the UI thread.
Accessing collections on non-UI Threads
WPF enables you to access and modify data collections on threads other than the one that created the collection. This enables you to use a background thread to receive data from an external source, such as a database, and display the data on the UI thread. By using another thread to modify the collection, your user interface remains responsive to user interaction.
Value changes fired by INotifyPropertyChanged are automatically marshalled back onto the dispatcher.
How to enable cross-thread access
Remember, async method itself runs on the main thread. So this is valid:
private async void MyButton_ClickAsync(object sender, RoutedEventArgs e)
{
txt.Text = "starting"; // UI Thread
await Task.Run(()=> ExecuteLongProcedure1());
txt.Text = "waiting"; // UI Thread
await Task.Run(()=> ExecuteLongProcedure2());
txt.Text = "finished"; // UI Thread
}
Another way to invoke UI operations from UI thread is to use SynchronizationContext as described here. SynchronizationContext is a stronger abstraction than Dispatcher and it's cross-platform.
var uiContext = SynchronizationContext.Current;
while (stillWorking)
{
uiContext.Post(o =>
{
textBox.Text += ".";
}, null);
await Task.Delay(51);
}
Patterns:
Fire and forget pattern:
For obvious reasons this is how your WPF GUI event handlers such as Button_ClickAsync are called.
void Do()
{
// CPU-Bound or IO-Bound operations
}
async void DoAsync() // returns void
{
await Task.Run(Do);
}
void FireAndForget() // not blocks, not waits
{
DoAsync();
}
Fire and observe:
Task-returning methods are better since unhandled exceptions trigger the TaskScheduler.UnobservedTaskException.
void Do()
{
// CPU-Bound or IO-Bound operations
}
async Task DoAsync() // returns Task
{
await Task.Run(Do);
}
void FireAndWait() // not blocks, not waits
{
Task.Run(DoAsync);
}
Fire and wait synchronously while wasting thread resources:
This is known as Sync over async, it is a synchronous operation but it uses more than one thread which may cause starvation. This happens when you call Wait() or try to read results directly from task.Result before the task is finished.
(AVOID THIS PATTERN)
void Do()
{
// CPU-Bound or IO-Bound operations
}
async Task DoAsync() // returns Task
{
await Task.Run(Do);
}
void FireAndWait() // blocks, waits and uses 2 more threads. Yikes!
{
var task = Task.Run(DoAsync);
task.Wait();
}
Is that all to it?
No. There is a lot more to learn about async, its context and its continuation. This blogpost is especially recommended.
Task uses Thread? Are you sure?
Not necessarily. Read this answer to know more about the true face of async.
Stephen Cleary has explained async-await perfectly. He also explains in his other blog post when there is no thread involved.
Read more
ValueTask and Task
MSDN explains Task
MSDN explains async
how-to-call-asynchronous-method-from-synchronous-method
async await - Behind the scenes
async await - FAQ
Make sure you know the difference between Asynchronous, Parallel and Concurrent.
You may also read a simple asynchronous file writer to know where you should concurrent.
Investigate concurrent namespace
Ultimately, read this e-book: Patterns_of_Parallel_Programming_CSharp
Your use of TaskCompletionSource<T> is incorrect. TaskCompletionSource<T> is a way to create TAP-compatible wrappers for asynchronous operations. In your ExecuteLongProcedureAsync method, the sample code is all CPU-bound (i.e., inherently synchronous, not asynchronous).
So, it's much more natural to write ExecuteLongProcedure as a synchronous method. It's also a good idea to use standard types for standard behaviors, in particular using IProgress<T> for progress updates and CancellationToken for cancellation:
internal void ExecuteLongProcedure(int param1, int param2, int param3,
CancellationToken cancellationToken, IProgress<string> progress)
{
//Start doing work
if (progress != null)
progress.Report("Work Started");
while (true)
{
//Mid procedure progress report
if (progress != null)
progress.Report("Bath water n% thrown out");
cancellationToken.ThrowIfCancellationRequested();
}
//Exit message
if (progress != null)
progress.Report("Done and Done");
}
Now you have a more reusable type (no GUI dependencies) that uses the appropriate conventions. It can be used as such:
public partial class MainWindow : Window
{
readonly otherClass _burnBabyBurn = new OtherClass();
CancellationTokenSource _stopWorkingCts = new CancellationTokenSource();
//A button method to start the long running method
private async void Button_Click_3(object sender, RoutedEventArgs e)
{
var progress = new Progress<string>(data => UpdateWindow(data));
try
{
await Task.Run(() => _burnBabyBurn.ExecuteLongProcedure(intParam1, intParam2, intParam3,
_stopWorkingCts.Token, progress));
}
catch (OperationCanceledException)
{
// TODO: update the GUI to indicate the method was canceled.
}
}
//A button Method to interrupt and stop the long running method
private void StopButton_Click(object sender, RoutedEventArgs e)
{
_stopWorkingCts.Cancel();
}
//A method to allow the worker method to call back and update the gui
void UpdateWindow(string message)
{
TextBox1.Text = message;
}
}
Here is an example using async/await, IProgress<T> and CancellationTokenSource. These are the modern C# and .Net Framework language features that you should be using. The other solutions are making my eyes bleed a bit.
Code Features
Count to 100 over a period of 10 seconds
Display progress on a progress bar
Long running work (a 'wait' period) performed without blocking the UI
User triggered cancellation
Incremental progress updates
Post operation status report
The view
<Window x:Class="ProgressExample.MainWindow"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
mc:Ignorable="d"
Title="MainWindow" SizeToContent="WidthAndHeight" Height="93.258" Width="316.945">
<StackPanel>
<Button x:Name="Button_Start" Click="Button_Click">Start</Button>
<ProgressBar x:Name="ProgressBar_Progress" Height="20" Maximum="100"/>
<Button x:Name="Button_Cancel" IsEnabled="False" Click="Button_Cancel_Click">Cancel</Button>
</StackPanel>
</Window>
The code
/// <summary>
/// Interaction logic for MainWindow.xaml
/// </summary>
public partial class MainWindow : Window
{
private CancellationTokenSource currentCancellationSource;
public MainWindow()
{
InitializeComponent();
}
private async void Button_Click(object sender, RoutedEventArgs e)
{
// Enable/disabled buttons so that only one counting task runs at a time.
this.Button_Start.IsEnabled = false;
this.Button_Cancel.IsEnabled = true;
try
{
// Set up the progress event handler - this instance automatically invokes to the UI for UI updates
// this.ProgressBar_Progress is the progress bar control
IProgress<int> progress = new Progress<int>(count => this.ProgressBar_Progress.Value = count);
currentCancellationSource = new CancellationTokenSource();
await CountToOneHundredAsync(progress, this.currentCancellationSource.Token);
// Operation was successful. Let the user know!
MessageBox.Show("Done counting!");
}
catch (OperationCanceledException)
{
// Operation was cancelled. Let the user know!
MessageBox.Show("Operation cancelled.");
}
finally
{
// Reset controls in a finally block so that they ALWAYS go
// back to the correct state once the counting ends,
// regardless of any exceptions
this.Button_Start.IsEnabled = true;
this.Button_Cancel.IsEnabled = false;
this.ProgressBar_Progress.Value = 0;
// Dispose of the cancellation source as it is no longer needed
this.currentCancellationSource.Dispose();
this.currentCancellationSource = null;
}
}
private async Task CountToOneHundredAsync(IProgress<int> progress, CancellationToken cancellationToken)
{
for (int i = 1; i <= 100; i++)
{
// This is where the 'work' is performed.
// Feel free to swap out Task.Delay for your own Task-returning code!
// You can even await many tasks here
// ConfigureAwait(false) tells the task that we dont need to come back to the UI after awaiting
// This is a good read on the subject - https://blog.stephencleary.com/2012/07/dont-block-on-async-code.html
await Task.Delay(100, cancellationToken).ConfigureAwait(false);
// If cancelled, an exception will be thrown by the call the task.Delay
// and will bubble up to the calling method because we used await!
// Report progress with the current number
progress.Report(i);
}
}
private void Button_Cancel_Click(object sender, RoutedEventArgs e)
{
// Cancel the cancellation token
this.currentCancellationSource.Cancel();
}
}
This is a simplified version of the most popular answer here by Bijan. I simplified Bijan's answer to help me think through the problem using the nice formatting provided by Stack Overflow.
By carefully reading and editing Bijan's post I finally understood: How to wait for async method to complete?
In my case the chosen answer for that other post is what ultimately led me to solve my problem:
"Avoid async void. Have your methods return Task instead of void. Then you can await them."
My simplified version of Bijan's (excellent) answer follows:
1) This starts a task using async and await:
private async void Button_Click_3(object sender, RoutedEventArgs e)
{
// if ExecuteLongProcedureAsync has a return value
var returnValue = await Task.Run(()=>
ExecuteLongProcedureAsync(this, intParam1, intParam2, intParam3));
}
2) This is the method to execute asynchronously:
bool stillWorking = true;
internal void ExecuteLongProcedureAsync(MainWindow gui, int param1, int param2, int param3)
{
//Start doing work
gui.UpdateWindow("Work Started");
while (stillWorking)
{
//put a dot in the window showing the progress
gui.UpdateWindow(".");
//the following line blocks main thread unless
//ExecuteLongProcedureAsync is called with await keyword
System.Threading.Thread.Sleep(50);
}
gui.UpdateWindow("Done and Done");
}
3) Invoke the operation which involves a property from gui:
void UpdateWindow(string text)
{
//safe call
Dispatcher.Invoke(() =>
{
txt.Text += text;
});
}
Or,
void UpdateWindow(string text)
{
//simply
txt.Text += text;
}
Closing comments) In most cases you have two methods.
First method (Button_Click_3) calls the second method and has the async modifier which tells the compiler to enable threading for that method.
Thread.Sleep in an async method blocks the main thread. but awaiting a task does not.
Execution stops on current thread (second thread) on await statements until task is finished.
You can't use await outside an async method
Second method (ExecuteLongProcedureAsync) is wrapped within a task and returns a generic Task<original return type> object which can be instructed to be processed asynchronously by adding await before it.
Everything in this method in executed asynchronously
Important:
Liero brought up an important issue. When you are Binding an element to a ViewModel property, the property changed callback is executed in UI thread. So there is no need to use Dispatcher.Invoke. Value changes fired by INotifyPropertyChanged are automatically marshalled back onto the dispatcher.