Adding Swagger to an ASP .NET Core 2 Web API

If you develop REST APIs, then you have probably heard of Swagger before.

Swagger is a tool that automatically documents your Web API, and provides the means to easily interact with it via an auto-generated UI.

In this article, we’ll see how to add Swagger to an ASP .NET Core Web API. We’ll be using the .NET Core 2 SDK, so if you’re using an earlier version, your mileage may vary. We’re only covering basic setup, so check out ASP.NET Web API Help Pages using Swagger in the Microsoft documentation if you want to go beyond that.

Project Setup

To make things easy, we’ll use one of the templates straight out of Visual Studio to get started. Go to File -> New -> Project… and select the ASP .NET Core Web Application template:

Next, pick the Web API template. You may want to change the second dropdown to ASP .NET Core 2.0, as it is currently not set to that by default.

Adding Swagger

You should now have a simple Web API project template with a ValuesController, providing an easy way to play with Web API out of the box.

To add Swagger, we first need to add a package:

Install-Package Swashbuckle.AspNetCore

Then, we throw in some configuration in Startup.cs to make Swagger work. Replace “My API” with whatever your API is called.

        // This method gets called by the runtime. Use this method to add services to the container.
        public void ConfigureServices(IServiceCollection services)
        {
            services.AddMvc();

            services.AddSwaggerGen(c =>
            {
                c.SwaggerDoc("v1", new Info { Title = "My API", Version = "v1" });
            });
        }

        // This method gets called by the runtime. Use this method to configure the HTTP request pipeline.
        public void Configure(IApplicationBuilder app, IHostingEnvironment env)
        {
            if (env.IsDevelopment())
            {
                app.UseDeveloperExceptionPage();
            }

            app.UseSwagger();

            app.UseSwaggerUI(c =>
            {
                c.SwaggerEndpoint("/swagger/v1/swagger.json", "My API V1");
            });

            app.UseMvc();
        }

Note: if you’re starting with a more minimal project template, it is possible that you may need to install the Microsoft.AspNetCore.StaticFiles package for this to work. This package is already present in the project template we’re using above.

Accessing Swagger

Let’s now run the web application. We should see a basic JSON response from the ValuesController:

If we now change the URL to http://localhost:<port>/swagger/, then we get to the Swagger UI:

Here, we can see a list of all our controllers and their actions. We can also open them up to interact with them.

Summary

That’s all it takes to add Swagger to your ASP .NET Core Web API.

  1. Add the Swashbuckle.AspNetCore package.
  2. Configure Swagger in the startup class.
  3. Access Swagger from http://localhost:<port>/swagger/.

TaskCompletionSource by Example

In this article, we’ll learn how to use TaskCompletionSource. It’s one of those tools which you will rarely need to use, but when you do, you’ll be glad that you knew about it. Let’s dive right into it.

Basic Usage

The source code for this section is in the TaskCompletionSource1 folder at the Gigi Labs BitBucket Repository.

Let’s create a new console application, and in Main(), we’ll have my usual workaround for running asynchronous code in a console application:

        static void Main(string[] args)
        {
            Run();
            Console.ReadLine();
        }

In the Run() method, we have a simple example showing how TaskCompletionSource works:

        static async void Run()
        {
            var tcs = new TaskCompletionSource<bool>();

            var fireAndForgetTask = Task.Delay(5000)
                                        .ContinueWith(task => tcs.SetResult(true));

            await tcs.Task;
        }

TaskCompletionSource is just a wrapper for a Task, giving you control over its completion. Thus, a TaskCompletionSource<bool> will contain a Task<bool>, and you can set the bool result based on your own logic.

Here, we are using TaskCompletionSource as a synchronization mechanism. Our main thread spawns off an operation and waits for its result, using the Task in the TaskCompletionSource. Even if the operation is not Task-based, it can set the result of the Task in the TaskCompletionSource, allowing the main thread to resume its execution.

Let’s add some diagnostic code so that we can understand what’s going on from the output:

        static async void Run()
        {
            var stopwatch = Stopwatch.StartNew();

            var tcs = new TaskCompletionSource<bool>();

            Console.WriteLine($"Starting... (after {stopwatch.ElapsedMilliseconds}ms)");

            var fireAndForgetTask = Task.Delay(5000)
                                        .ContinueWith(task => tcs.SetResult(true));

            Console.WriteLine($"Waiting...  (after {stopwatch.ElapsedMilliseconds}ms)");

            await tcs.Task;

            Console.WriteLine($"Done.       (after {stopwatch.ElapsedMilliseconds}ms)");

            stopwatch.Stop();
        }

And here is the output:

Starting... (after 0ms)
Waiting...  (after 41ms)
Done.       (after 5072ms)

As you can see, the main thread waited until tcs.SetResult(true) was called; this triggered completion of the TaskCompletionSource’s underlying task (which the main thread was awaiting), and allowed the main thread to resume.

Aside from SetResult(), TaskCompletionSource offers methods to cancel a task or fault it with an exception. There are also safe Try...() equivalents:

SDK Example

The source code for this section is in the TaskCompletionSource2 folder at the Gigi Labs BitBucket Repository.

One scenario where I found TaskCompletionSource to be extremely well-suited is when you are given a third-party SDK which exposes events. Imagine this: you submit an order via an SDK method, and it gives you an ID for that order, but not the result. The SDK goes off and does what it has to do to perhaps talk to an external service and have the order processed. When this eventually happens, the SDK fires an event to notify the calling application on whether the order was placed successfully.

We’ll use this as an example of the SDK code:

    public class MockSdk
    {
        public event EventHandler<OrderOutcome> OnOrderCompleted;

        public Guid SubmitOrder(decimal price)
        {
            var orderId = Guid.NewGuid();

            // do a REST call over the network or something
            Task.Delay(3000).ContinueWith(task => OnOrderCompleted(this,
                new OrderOutcome(orderId, true)));

            return orderId;
        }
    }

The OrderOutcome class is just a simple DTO:

    public class OrderOutcome
    {
        public Guid OrderId { get; set; }
        public bool Success { get; set; }

        public OrderOutcome(Guid orderId, bool success)
        {
            this.OrderId = orderId;
            this.Success = success;
        }
    }

Notice how MockSdk‘s SubmitOrder does not return any form of Task, and we can’t await it. This doesn’t necessarily mean that it’s blocking; it might be using another form of asynchrony such as the Asynchronous Programming Model or a messaging framework with a request-response fashion (such as RPC over RabbitMQ).

At the end of the day, this is still asynchrony, and we can use TaskCompletionSource to build a Task-based Asynchronous Pattern abstraction over it (allowing the application to simply await the call).

First, we start building a simple proxy class that wraps the SDK:

    public class SdkProxy
    {
        private MockSdk sdk;

        public SdkProxy()
        {
            this.sdk = new MockSdk();
            this.sdk.OnOrderCompleted += Sdk_OnOrderCompleted;
        }

        private void Sdk_OnOrderCompleted(object sender, OrderOutcome e)
        {
            // TODO
        }
    }

We then add a dictionary, which allows us to relate each OrderId to its corresponding TaskCompletionSource. Using a ConcurrentDictionary instead of a normal Dictionary helps deal with multithreading scenarios without needing to lock:

        private ConcurrentDictionary<Guid,
            TaskCompletionSource<bool>> pendingOrders;
        private MockSdk sdk;

        public SdkProxy()
        {
            this.pendingOrders = new ConcurrentDictionary<Guid,
                TaskCompletionSource<bool>>();

            this.sdk = new MockSdk();
            this.sdk.OnOrderCompleted += Sdk_OnOrderCompleted;
        }

The proxy class exposes a SubmitOrderAsync() method:

        public Task SubmitOrderAsync(decimal price)
        {
            var orderId = sdk.SubmitOrder(price);

            Console.WriteLine($"OrderId {orderId} submitted with price {price}");

            var tcs = new TaskCompletionSource<bool>();
            this.pendingOrders.TryAdd(orderId, tcs);

            return tcs.Task;
        }

This method calls the underlying SubmitOrder() in the SDK, and uses the returned OrderId to add a new TaskCompletionSource in the dictionary. The TaskCompletionSource’s underlying Task is returned, so that the application can await it.

        private void Sdk_OnOrderCompleted(object sender, OrderOutcome e)
        {
            string successStr = e.Success ? "was successful" : "failed";
            Console.WriteLine($"OrderId {e.OrderId} {successStr}");

            this.pendingOrders.TryRemove(e.OrderId, out var tcs);
            tcs.SetResult(e.Success);
        }

When the SDK fires a completion event, the proxy will remove the TaskCompletionSource from the pending order and set its result. The application awaiting the underlying task will resume and take a decision depending on the logic.

We can test this with the following program code in a console application:

        static void Main(string[] args)
        {
            Run();
            Console.ReadLine();
        }

        static async void Run()
        {
            var sdkProxy = new SdkProxy();

            await sdkProxy.SubmitOrderAsync(10);
            await sdkProxy.SubmitOrderAsync(20);
            await sdkProxy.SubmitOrderAsync(5);
            await sdkProxy.SubmitOrderAsync(15);
            await sdkProxy.SubmitOrderAsync(4);
        }

The output shows that the program did indeed wait for each order to complete before starting the next one:

OrderId 3e2d4577-8bbb-46b7-a5df-2efec23bae6b submitted with price 10
OrderId 3e2d4577-8bbb-46b7-a5df-2efec23bae6b was successful
OrderId e22425b9-3aa3-48db-a40f-8b8cfbdcd3af submitted with price 20
OrderId e22425b9-3aa3-48db-a40f-8b8cfbdcd3af was successful
OrderId 3b5a2602-a5d2-4225-bbdb-10642a63f7bc submitted with price 5
OrderId 3b5a2602-a5d2-4225-bbdb-10642a63f7bc was successful
OrderId ffd61cea-343e-4a9c-a76f-889598a45993 submitted with price 15
OrderId ffd61cea-343e-4a9c-a76f-889598a45993 was successful
OrderId b443462c-f949-49b9-a6f0-08bbbb82fe7e submitted with price 4
OrderId b443462c-f949-49b9-a6f0-08bbbb82fe7e was successful

Summary

Use TaskCompletionSource to adapt an arbitrary form of asynchrony to use Tasks, and enable elegant async/await usage.

Do not use it simply expose an asynchronous wrapper for a synchronous method. You should either not do that at all, or use Task.FromResult() instead.

If you’re concerned that the the asynchronous call might never resume, consider adding a timeout.

Common Mistakes in Asynchronous Programming with .NET

In the last few articles, we have seen how to work with asynchronous programming in C#. Although it is now easier than ever to write responsive applications that do asynchronous, non-blocking I/O operations, many people still use asynchronous programming incorrectly. A lot of this is due to confusion over usage of the Task class in .NET, which is used in multithreaded and parallel scenarios as well as asynchronous ones. To make matters worse, it is not obvious to everyone that these are actually different things.

So let’s address this concern first.

Update 16th October 2017: Several people have pointed out errors in this article with regards to the effect that blocking has on the CPU. I apologise for this, and have made corrections. Blocking does not hog the CPU, but prevents threads from doing other work while they wait. Thanks for clarifying the confusion, and I welcome any further corrections.

Update 8th March 2018: See also “Avoid await in Foreach” for another common mistake with async/await.

Asynchronous vs Multithreading etc

When we use a computer, many programs are running at the same time. Before the advent of multicore CPUs, this was achieved by having instructions from different processes (and eventually threads) running one at a time on the same CPU. These instructions are interleaved, and the CPU switches rapidly from one to another (in what we call a context switch), giving the illusion that they are running at the same time. This is concurrency.

CPUs with multiple cores have the additional ability of literally executing multiple intructions at the same time, which is parallel execution. Multithreading gives the developers to make full use of the available cores; without it, instructions from a single process would only be able to execute on a single core at a time.

“Tasks which are executing on distinct processors at any point in time are said to be running in parallel. It may also be possible to execute several tasks on a single processor. Over a period of time, the impression is given that they are running in parallel, when in fact, at any point in time, only one task has control of the processor. In this case, we say that the tasks are being performed concurrently, that is, their execution is being shared by the same processor.” — Practical Parallel Rendering, Chalmers et al, A K Peters, 2002.

In .NET, we can do parallel processing by using multithreading, or an abstraction thereof, such as the Task Parallel Library. Parallel processing is CPU-bound.

Parallel processing is often contrasted with distributed processing, where the computing resources are not physically tightly coupled. This is not, however, relevant to asynchronous programming, so we will not delve into it.

Operations that take a long time to execute will typically hold control of the thread in which they are running, in what we call a blocking operation. However, if these operations involve waiting for I/O to occur (e.g. waiting for results from a file, network or database), then the I/O could occur in a non-blocking fashion without holding the thread at all during the waiting time. We say that the I/O operation is asynchronous: the thread that is waiting for it does not actually wait, but may be reassigned to do other work until the I/O operation is complete.

Asynchronous non-blocking I/O is not enabled by multithreading. In fact, Stephen Cleary goes into detail about how this works in his excellent post, “There Is No Thread“. In brief, a mechanism known as I/O Completion Ports (IOCP) is used to notify a thread that its I/O request is ready; but that thread does not need to block (or indeed run at all) during the waiting time. This is what we enable when we do an asynchronous wait by means of the await keyword.

In order to write efficient code, it is fundamental to understand the nature of what the code is doing. Parallel CPU-based execution involves significant overheads in thread synchronization. It makes no sense to use Parallel.ForEach() for I/O-bound tasks, and many are also surprised to find that executing CPU-based tasks sequentially is often faster than doing them in parallel, especially when such tasks are fine-grained and do very trivial work. In such cases, the synchronization overheads dwarf the cost of executing that code directly on the CPU on a single thread.

See also: “Asynchronous and Concurrent Processing in Akka .NET Actors“, which has a section on Asynchronous vs Concurrent Processing using simple tasks.

The Dangers of Blocking with Asynchronous Code

Asynchronous programming has two main benefits: scalability and offloading. If you block, then you are hogging resources (i.e. threads) that could be better used elsewhere. In an ASP .NET context, this means that a thread cannot service other requests (hurts scalability). In a GUI context, it means that the UI thread cannot be used for rendering because it is busy waiting for a long-running operation (so the work should be offloaded).

There are several methods which are part of the Task API which block, such as Wait(), WaitAny() and WaitAll(). The Result property also has the effect of blocking until the task is complete. Stephen Cleary has a table in his Async and Await intro showing these blocking API calls and how to turn them into asynchronous calls. Best practice is to await the asynchronous equivalent of the blocking method

However, simply wasting threading resources is not the only problem with blocking. It is actually very easy to end up with a deadlock and stall your application. Stephen Cleary has an excellent explanation of how this happens, with two concise examples based on GUI applications (e.g. WPF or Windows Forms) and ASP .NET. I am only going to attempt to simplify the scenario and illustrate it with a diagram.

Consider the following code in a WPF application’s codebehind (MainWindow.xaml.cs):

        private Uri uri = new Uri("http://ip.jsontest.com/");

        public async Task WaitABit()
        {
            await Task.Delay(3000);
        }

        private void Button_Click(object sender, RoutedEventArgs e)
        {
            var task = WaitABit();
            task.Wait();
        }

This results in deadlock. Let’s see why:

  1. Button_Click() calls WaitABit() without awaiting it. This starts a task in fire-and-forget mode (so far).
  2. WaitABit() calls Task.Delay() and awaits its result asynchronously. The thread does not block, since the execution has gone into “I/O mode” (technically a delay isn’t really I/O, but the idea is the same).
  3. In the meantime, Button_Click() resumes execution, and calls Wait(), effectively blocking until the result of WaitABit() is ready.
  4. The delay completes.
  5. WaitABit() should resume, but the UI thread that it’s supposed to run on is already blocked.
  6. The deadlock occurs because the continuation of WaitABit() needs to run on the UI thread, but the UI thread is blocked waiting for the result of WaitABit().

Note: I intentionally haven’t simplified WaitABit() such that it just returns the delay rather than doing a single-line async/await, as it would not deadlock. Can you guess why?

This example has shown blocking of the UI thread, but the concept stretches beyond GUI applications. In order to fully understand what is happening, we need to understand what a SynchronizationContext is. In short, it’s an abstraction of a threading model within an application. A GUI application needs a single UI thread for rendering and updating GUI components (although you can use other threads, they cannot touch the GUI directly). An ASP .NET application, on the other hand, handles requests using the thread pool. The SynchronizationContext is the abstraction that allows us to use the same multithreaded and asynchronous programming models across applications with fundamentally different internal threading models.

As a result, GUI applications (e.g. Windows Forms and WPF) and ASP .NET applications (those targeting the full .NET Framework) can deadlock. ASP .NET Core applications don’t have a SynchronizationContext, so they will not deadlock. Console applications won’t normally deadlock because the task continuation can just execute on another thread.

The deadlock occurs because the SynchronizationContext (e.g. the UI thread in the above example) is captured and used for the task continuation. However, we can prevent this from happening by using ConfigureAwait(false):

        public async Task WaitABit()
        {
            await Task.Delay(3000).ConfigureAwait(false);
        }

The GUI application does not deadlock now, because a thread pool can be picked to execute the continuation. Once WaitABit() completes, then the blocking Wait() in Button_Click() can resume on the same UI thread where it started. Any modifications to UI elements would work fine.

While ConfigureAwait(false) is no replacement for doing async/await all the way, it does have its benefits. Capturing the SynchronizationContext incurs a performance penalty, so if it is not actually necessary, library code should avoid it by using ConfigureAwait(false). Also, if application code must block on an asynchronous call for legacy reasons, then ConfigureAwait(false) would avoid the resulting deadlocks.

Of course, the real fix for our deadlock example here is really as easy as this (without ConfigureAwait(false)):

        private async void Button_Click(object sender, RoutedEventArgs e)
        {
            var task = WaitABit();
            await task;
        }

I’ve just replaced the blocking Wait() call with an await, and marked the method as async as a result.

Asynchronous Wrappers for Synchronous Methods

A lot of library APIs with asynchronous methods have pairs of synchronous and asynchronous methods, e.g. Read() and ReadAsync(), Write() and WriteAsync(), etc. If your library API is purely synchronous, then you should not expose asynchronous wrappers for the synchronous methods (e.g. by using Task.Run()).

Stephen Toub goes into detail on why this is a bad idea in the article linked above, but I think the following paragraph summarises it best:

“If a developer needs to achieve better scalability, they can use any async APIs exposed, and they don’t have to pay additional overhead for invoking a faux async API. If a developer needs to achieve responsiveness or parallelism with synchronous APIs, they can simply wrap the invocation with a method like Task.Run.” — Stephen Toub, “Should I expose asynchronous wrappers for synchronous methods?

Asynchronous Properties

You can’t use async/await in properties. You can use them indirectly via an asynchronous method, but it’s a rather weird thing to do.

“[You can’t use await i]nside of a property getter or setter. Properties are meant to return to the caller quickly, and thus are not expected to need asynchrony, which is geared for potentially long-running operations. If you must use asynchrony in your properties, you can do so by implementing an async method which is then used from your property.” — Stephen Toub, Async/Await FAQ

async void methods

async void methods should only be used for top-level event handlers. In “The Dangers of async void Event Handlers“, I explain the general dangers of async void methods (mainly related to not having a task that you can await), but I also demonstrate and solve the additional problem of async void event handlers interleaving while they run (which is problematic if you’re expecting to handle events in sequence, such as with a message queue).

    1. “There is no way for the caller to await completion of the method.
    2. “As a result of this, async void calls are fire-and-forget.
    3. “Thus it follows that async void methods (including event handlers) will execute in parallel if called in a loop.
    4. “Exceptions can cause the application to crash (see the aforementioned article by Stephen Cleary for more on this).”

— Daniel D’Agostino, The Dangers of async void Event Handlers

Summary

  1. Parallel is CPU-based. Asynchronous is I/O-based. Don’t mix the two. (Running asynchronous I/O tasks “in parallel” is OK, as long as you’re doing it following the proper patterns rather than using something like Parallel.Foreach().)
  2. Asynchronous I/O does not use any threads.
  3. Blocking affects scalability and can hold higher-priority resources (such as a UI thread).
  4. Blocking can also result in deadlocks. Prevent them by using async/await all the way if you can. ConfigureAwait(false) is useful in library code both for performance reasons and to prevent deadlocks resulting from application code that must block for legacy reasons.
  5. For asynchronous libraries, don’t expose synchronous wrappers. There is an overhead associated with it, and the client can decide whether it’s worth doing from their end.
  6. You can’t have asynchronous properties, except indirectly via asynchronous methods. Avoid this.
  7. async void is for event handlers. Even so, if ordering is important, beware of interleaving.

Patterns for Asynchronous Composite Tasks in C#

In the previous two articles, I’ve explained why and how to use async/await for asynchronous programming in C#.

Now, we will turn our attention to more interesting things that we can do when combining multiple tasks within the same method.

Update 22nd September 2018: Another pattern not covered here is fire-and-forget. There are many ways to achieve this, including simply not awaiting (causes warnings – see ways to ignore them), using Task.Run(), using Task.Factory.StartNew(), or async void (not recommended, see “Common Mistakes in Asynchronous Programming with .NET“. This is suitable when you want to trigger some kind of processing but don’t care whether/when it completes. It doesn’t really fit the fast food scenario used in this article — placing an order without ever being notified of its completion/failure is sure to annoy customers. Which I suppose is also why applying for jobs is such a pain in the ass for many people.

Fast Food Example

In order to see each pattern at work, we need a simple example involving multiple tasks. Imagine you walk into your favourite fast food restaurant, and order a meal involving a burger, fries and a drink. Each of these takes a different amount of time to prepare, and the total time of the order may vary depending on how the execution of these three tasks takes place.

Sequential Tasks

The simplest approach is to just execute tasks one after another, waiting for one to finish before starting the next.

        static void Main(string[] args)
        {
            OrderAsync();
            Console.ReadLine();
        }

        static async void OrderAsync()
        {
            var stopwatch = Stopwatch.StartNew();

            await Task.Delay(3000)
                .ContinueWith(task => ShowCompletion("Fries", stopwatch.Elapsed));
            await Task.Delay(1000)
                .ContinueWith(task => ShowCompletion("Drink", stopwatch.Elapsed));
            await Task.Delay(5000)
                .ContinueWith(task => ShowCompletion("Burger", stopwatch.Elapsed));

            ShowCompletion("Order", stopwatch.Elapsed);

            stopwatch.Stop();
        }

        static void ShowCompletion(string name, TimeSpan time)
        {
            Console.WriteLine($"{name} completed after {time}");
        }

In this example code, we are representing the fries, drink and burger tasks as delays of different length. The rest of the code is purely diagnostic in order to allow us to get some output and understand the results. There is also a workaround allowing us to use asynchronous code in Main(), that was described in the previous article.

Here is the output from the above:

Fries completed after 00:00:03.0359621
Drink completed after 00:00:04.0408785
Burger completed after 00:00:09.0426927
Order completed after 00:00:09.0434057

Because we performed each task sequentially, the total order took 9 seconds. In a fast food restaurant, it probably does not make sense to wait for the fries to be ready before preparing the drink, and to wait for both to be ready before starting to prepare the burger. These could be done in parallel, as we will see in the next sections.

However, there are many legitimate cases where sequential task execution makes sense. We’ve seen one in “Motivation for async/await in C#“:

private async void Button_Click(object sender, RoutedEventArgs e)
{
    var baseAddress = new Uri("http://mta.com.mt");
 
    using (var httpClient = new HttpClient() { BaseAddress = baseAddress })
    {
        var response = await httpClient.GetAsync("/");
        var content = await response.Content.ReadAsStringAsync();
 
        MessageBox.Show("Response arrived!", "Slow website");
    }
}

In this case, the tasks are dependent on each other. In order to get the content of the response, the response itself must first finish executing. Because there is this dependency, the tasks must be executed one after the other.

Parallel Tasks, All Must Finish

If we fire off the tasks without awaiting them right away, there are more interesting things we can do with them. Essentially, by removing await, we are running the tasks in parallel.

        static async void OrderAsync()
        {
            var stopwatch = Stopwatch.StartNew();

            var friesTask = Task.Delay(3000)
                .ContinueWith(task => ShowCompletion("Fries", stopwatch.Elapsed));
            var drinkTask = Task.Delay(1000)
                .ContinueWith(task => ShowCompletion("Drink", stopwatch.Elapsed));
            var burgerTask = Task.Delay(5000)
                .ContinueWith(task => ShowCompletion("Burger", stopwatch.Elapsed));

            await Task.WhenAll(friesTask, drinkTask, burgerTask);

            ShowCompletion("Order", stopwatch.Elapsed);

            stopwatch.Stop();
        }

Aside from removing await before each task, we are assigning them to variables so that we can keep track of them. We then rely on Task.WhenAll() to wait until all tasks have completed (as an analogy, think of it as a memory barrier). Task.WhenAll() is awaitable, unlike its blocking cousin Task.WaitAll(). This gives us a way to easily run asynchronous tasks in parallel where it makes sense to do so.

And in a fast food restaurant, preparing fries and drink while the burger is cooking makes a lot of sense. In fact, the order is ready after just 5 seconds, which is the time of the longest task (the burger). Because the fries and drink were prepared concurrently with the burger, they did not add anything to the total time of the order.

Drink completed after 00:00:01.1696855
Fries completed after 00:00:03.0363008
Burger completed after 00:00:05.0443482
Order completed after 00:00:05.0445130

Note that Task.WhenAll() takes an IEnumerable<Task>, and as such, you can easily pass it a list of tasks (e.g. when the number of tasks is dynamic based on input or data).

Parallel Tasks, First To Finish

If you’re hungry and thirsty after an unexpected trip in the desert, it’s unlikely that you’re going to want to wait for all items to finish before starting to eat and drink. Instead, you’ll consume each item as soon as it arrives.

        static async void OrderAsync()
        {
            var stopwatch = Stopwatch.StartNew();

            var friesTask = Task.Delay(3000)
                .ContinueWith(task => ShowCompletion("Fries", stopwatch.Elapsed));
            var drinkTask = Task.Delay(1000)
                .ContinueWith(task => ShowCompletion("Drink", stopwatch.Elapsed));
            var burgerTask = Task.Delay(5000)
                .ContinueWith(task => ShowCompletion("Burger", stopwatch.Elapsed));

            await Task.WhenAny(friesTask, drinkTask, burgerTask);

            ShowCompletion("Order", stopwatch.Elapsed);

            stopwatch.Stop();
        }

Task.WhenAny() will wait until the first task has completed, and then resume execution of the method. It also returns the task that completed (though we’re not using that here).

Drink completed after 00:00:01.0390588
Order completed after 00:00:01.0412190
Fries completed after 00:00:01.0413729
Burger completed after 00:00:01.0413729

Our results are a little messed up. Since Task.WhenAny() only waits for the first task to complete, the entire order was considered complete as soon as the drink was ready. The stopwatch was subsequently stopped, and the output shows 1 second for everything even though the fries and burger actually took longer.

This scenario is useful when you want to retrieve data from different sources and just use the result that arrived fastest. It is not very intuitive for when you’re dying of hunger and want to gobble up everything as it arrives. We’ll address this in the next section.

Parallel Tasks, All Must Finish, Process As They Arrive

So here’s the scenario: we’re famished, and we want to consume our drink, fries and burger as they are ready. We want to consume all of them, but Task.WhenAny() only gives us the first task that completed.

It’s easy to reuse Task.WhenAny() to wait for all tasks to complete, by using a simple loop.

        static async void OrderAsync()
        {
            var stopwatch = Stopwatch.StartNew();

            var friesTask = Task.Delay(3000)
                .ContinueWith(task => ShowCompletion("Fries", stopwatch.Elapsed));
            var drinkTask = Task.Delay(1000)
                .ContinueWith(task => ShowCompletion("Drink", stopwatch.Elapsed));
            var burgerTask = Task.Delay(5000)
                .ContinueWith(task => ShowCompletion("Burger", stopwatch.Elapsed));

            var tasks = new List<Task>() { friesTask, drinkTask, burgerTask };
            
            while (tasks.Count > 0)
            {
                var task = await Task.WhenAny(tasks);
                tasks.Remove(task);

                Console.WriteLine($"Yum! {tasks.Count} left!");
            }

            ShowCompletion("Order", stopwatch.Elapsed);

            stopwatch.Stop();
        }

We’re putting all tasks in a list, and as each task completes, we remove it from the list. We know we’re done when there’s nothing left in the list.

Drink completed after 00:00:01.0506610
Yum! 2 left!
Fries completed after 00:00:03.0328112
Yum! 1 left!
Burger completed after 00:00:05.0317576
Yum! 0 left!
Order completed after 00:00:05.0331167

From this example, it might appear that there’s no benefit from using this approach when compared to just using continuations on tasks and using Task.WhenAll(). However, in real scenarios that don’t involve french fries, it is often reasonable to check the result of each task for failure. If one of the tasks fails, then the operation is aborted without having to wait for all the other tasks to complete.

Task With Timeout

As it turns out, we’re so hungry that we’re only willing to wait up to 4 seconds for each item, since the start of the order. If they take longer than 4 seconds, we’ll cancel that part of the order.

Fortunately, there’s an excellent blog post on the Parallel Programming MSDN blog from 2011 that shows how to write a TimeoutAfter() method that does exactly this. I’ll go ahead and steal it:

    public static class TaskExtensions
    {
        public static async Task TimeoutAfter(this Task task, int millisecondsTimeout)
        {
            if (task == await Task.WhenAny(task, Task.Delay(millisecondsTimeout)))
                await task;
            else
                throw new TimeoutException();
        }
    }

It’s an extension method, so we can easily use it with the tasks we already have:

        static async void OrderAsync()
        {
            var stopwatch = Stopwatch.StartNew();

            var friesTask = Task.Delay(3000).TimeoutAfter(4000)
                .ContinueWith(task => ShowCompletion("Fries", stopwatch.Elapsed));
            var drinkTask = Task.Delay(1000).TimeoutAfter(4000)
                .ContinueWith(task => ShowCompletion("Drink", stopwatch.Elapsed));
            var burgerTask = Task.Delay(5000).TimeoutAfter(4000)
                .ContinueWith(task => ShowCompletion("Burger", stopwatch.Elapsed));

            var tasks = new List<Task>() { friesTask, drinkTask, burgerTask };
            
            while (tasks.Count > 0)
            {
                var task = await Task.WhenAny(tasks);
                tasks.Remove(task);

                Console.WriteLine($"Yum! {tasks.Count} left!");
            }

            ShowCompletion("Order", stopwatch.Elapsed);

            stopwatch.Stop();
        }

Running this, the burger task will timeout and an exception will be thrown. Since we’re not actually checking for this, all we see in the output is that the burger task finished after 4 seconds instead of 5.

Drink completed after 00:00:01.0819761
Yum! 2 left!
Fries completed after 00:00:03.0493526
Yum! 1 left!
Burger completed after 00:00:04.0952924
Yum! 0 left!
Order completed after 00:00:04.0974441

By putting a breakpoint or turning on first chance exceptions, though, we see that the TimeoutException was indeed thrown:

Summary

  1. awaiting tasks one after another will result in sequential execution.
  2. Use Task.WhenAll() to wait for all tasks to complete before proceeding.
  3. Use Task.WhenAny() to get the first task that finished, and proceed before waiting for the others.
  4. Use Task.WhenAny() in a loop to process all tasks as they arrive, and potentially break out early in case of failure.
  5. Apply a timeout to a task using the TimeoutAfter() extension method from the Parallel Programming blog on MSDN.

Working with Asynchronous Methods in C#

In yesterday’s article, “Motivation for async/await in C#“, we have seen why asynchronous programming is important. We have also seen basic usage of the await keyword, which requires its containing method to be marked as async.

When learning to write asynchronous methods, it is not trivial to get the interactions between various methods (which may or may not be asynchronous) right. In fact, the examples in yesterday’s article which use an async void method should normally only be used in event handlers, and even so, there are caveats to consider.

In this article, we’ll go through various different scenarios in which async/await can be used.

async Task methods

Let’s take another look at the asynchronous (event handler) method from yesterday’s article:

        private async void Button_Click(object sender, RoutedEventArgs e)
        {
            var baseAddress = new Uri("http://mta.com.mt");

            using (var httpClient = new HttpClient() { BaseAddress = baseAddress })
            {
                var response = await httpClient.GetAsync("/");
                var content = await response.Content.ReadAsStringAsync();

                MessageBox.Show("Response arrived!", "Slow website");
            }
        }

Try moving out the code into a separate method, and awaiting it from the event handler. You’ll find that you can’t await an async void method:

In order to be able to await a method, it must return Task (if it doesn’t need to return anything, such as void methods) or Task<T> (if it needs to return a value of type T). We also append an –Async suffix to the method name by convention to make it obvious for people who use such methods that they’re meant to be awaited.

Thus, this example becomes:

        private async void Button_Click(object sender, RoutedEventArgs e)
        {
            await GetHtmlAsync();
        }

        private async Task GetHtmlAsync()
        {
            var baseAddress = new Uri("http://mta.com.mt");

            using (var httpClient = new HttpClient() { BaseAddress = baseAddress })
            {
                var response = await httpClient.GetAsync("/");
                var content = await response.Content.ReadAsStringAsync();

                MessageBox.Show("Response arrived!", "Slow website");
            }
        }

This is an example of an async Task method, which does not return anything. Let’s change it such that it returns the HTML from the response:

        private async void Button_Click(object sender, RoutedEventArgs e)
        {
            string html = await GetHtmlAsync();
        }

        private async Task<string> GetHtmlAsync()
        {
            var baseAddress = new Uri("http://mta.com.mt");

            using (var httpClient = new HttpClient() { BaseAddress = baseAddress })
            {
                var response = await httpClient.GetAsync("/");
                var content = await response.Content.ReadAsStringAsync();

                return content;
            }
        }

We’ve changed the signature of GetHtmlAsync() to return Task<string> intead of just Task. Correspondingly, we are now returning content (a string) from the method. At the event handler, we are now assigning the result of the await into the html variable. Apart from waiting asynchronously until the method completes, await has the additional function of unwrapping the result from the Task that contains it; thus html is of type string.

If you try removing the async keyword from GetHtmlAsync(), you’ll learn a little more about the actual function of the async keyword:

Without async, you are expected to return what the method advertises: a Task<string>. On the other hand, if you mark the method as async, the meaning of the method is changed such that you can return a string directly. The underlying Task-related plumbing is handled by the compiler.

Chaining Asynchronous Methods

In the previous section, we have seen how methods need to return a Task in order to be awaited. Typically, one async Task method will call another and await its result.

The chain of calls ends at a last async Task, typically provided by the .NET Framework or other library, which interfaces directly with I/O (e.g. network or filesystem). It must be an asynchronous method; attempting to disguise a blocking call as async here will lead to deadlocks.

async Task may (and should) be used all the way from an incoming request to the final I/O library method in application types that support top-level asynchronous methods, such as Web API or WCF.

The situation is a little different for other applications (e.g. Windows Forms, WPF) that are event-driven. Asynchronous event handlers are a special case where we need to use async void methods, as we have already seen in the WPF example from yesterday’s article:

async void methods

Event handlers are void methods that are called by the runtime in a dispatcher loop. Thus, async void methods are necessary to allow usage of await within event handlers without requiring them to return a Task

However, as we have seen before, there is no way to await an async void method. This makes async void methods very dangerous to use outside of their intended context, as I have detailed in “The Dangers of async void Event Handlers“. This is one of the more common mistakes when programming with async/await, and it is good to become familiar with the problems in order to avoid repeating the same mistakes in future.

Fake Asynchronous Methods

Sometimes, you’ll have an interface that requires asynchronous methods, yet your implementation does not need anything asynchronous in it. Let’s look at a practical example:

    public interface ISimpleStorage
    {
        Task WriteAsync(string str);
        Task<string> ReadAsync();
    }

You could implement this interface using a simple file as a backing store, in which case your methods will be suitably asynchronous:

    public class FileStorage : ISimpleStorage
    {
        public async Task<string> ReadAsync()
        {
            using (var fs = File.OpenRead("file.txt"))
            using (var sr = new StreamReader(fs))
            {
                var str = await sr.ReadToEndAsync();
                return str;
            }
        }

        public async Task WriteAsync(string str)
        {
            using (var fs = File.OpenWrite("file.txt"))
            using (var sw = new StreamWriter(fs))
            {
                await sw.WriteAsync(str);
                await sw.FlushAsync();
            }
        }
    }

However, you could have another implementation which just uses memory as storage, and in this case there’s nothing asynchronous:

    public class MemoryStorage : ISimpleStorage
    {
        private string str;

        public Task<string> ReadAsync()
        {
            return Task.FromResult(str);
        }

        public Task WriteAsync(string str)
        {
            this.str = str;
            return Task.CompletedTask;
        }
    }

In that case, your methods need not be marked async. However, this means that you will actually need to return Task instances from each method. If you have nothing to return, then just return a Task.CompletedTask (available from .NET Framework 4.6 onwards). You can also use Task.FromResult() to construct a task from a variable that you want to return.

Simplifying Single Line Asynchronous Methods

Consider the following asynchronous method:

        static async Task WaitALittleAsync()
        {
            await Task.Delay(10000);
        }

Here, we are waiting for the delay to finish, and then returning the result.

Instead, we can just return the Task itself, and let the caller do the awaiting:

        static Task WaitALittleAsync()
        {
            return Task.Delay(10000);
        }

Once again, a method does not need to be marked async if (a) it does not await anything, and (b) it can return a Task, rather than some other type that needs to be wrapped in a Task.

Using async/await in Main()

Until recently, you couldn’t use async/await in a console application’s Main() method. You can have an async Task Main() method as from C# 7.1, but you need to make sure you’re using C# 7.1 first from your project properties -> Build -> Advanced…:

You can choose either “C# 7.1”, or “C# latest minor version (latest)”. With that, you can await directly from within Main():

        static async Task Main(string[] args)
        {
            await WaitALittleAsync();
        }

Before C# 7.1, you had to use some other workaround to await from Main(). One option is to use a special AsyncContext such as the one written by Stephen Cleary. Another is to just move asynchrony out of Main() and use a Console.ReadLine() to keep the window open:

        static void Main(string[] args)
        {
            Run();
            Console.ReadLine();
        }

        static async void Run()
        {
            await WaitALittleAsync();
        }

Summary

  1. Use async Task methods wherever you can.
  2. Use async void methods for event handlers, but beware the dangers even there.
  3. Use the –Async suffix in asynchronous method names to make them easily recognisable as awaitable.
  4. Use Task.CompletedTask and Task.FromResult() to satisfy asynchronous contracts with synchronous code.
  5. Asynchronous code in Main() has not been possible until recently. Use C# 7.1+ or a workaround to get this working.