Commands

ReactiveCommand is a Reactive Extensions and asynchronous aware implementation of the ICommand interface. ICommand is often used in the MVVM design pattern to allow the View to trigger business logic defined in the ViewModel. This allows for easier maintenance, unit testing, and the ability to reuse ViewModels across different UI frameworks. Examples of where a View might invoke a command include clicking a Save menu item, tapping a phone icon, or stretching an image. In these cases, the ViewModel will then invoke the business logic of saving outstanding changes, performing a phone call, or zooming into an image.

Creating commands

A ReactiveCommand is created using static factory methods which allows you to create command logic that executes either synchronously or asynchronously. The following are the different static factory methods:

  • CreateFromObservable() - Execute the logic using an IObservable.
  • CreateFromTask() - Execute a C# Task Parallel Library (TPL) Task. This allows use also of the C# async/await operators. Read more on canceling commands here.
  • Create() - Execute a synchronous Func or Action.
  • CreateCombined() - Execute one or more commands. Read more on combining commands here.

ReactiveCommand<TInput, TOutput> adds the concept of Input and Output generic types. The Input is often passed in by the View and it's type is captured as TInput, and the Output is the result of executing the command which type is captured as TOutput. ReactiveCommand<TInput, TOutput> is IObservable<TOutput> which can be used like any other IObservable. For example, since the ReactiveCommand is IObservable you can Subscribe() to it like any other observable, and add the output to a List on your view model. The Unit type is a functional programming construct analogous to void and can be used in cases where you don't care about either the input and/or output value.

// A synchronous command taking a parameter and returning nothing.
// The Unit type is often used to denote the successfull completion
// of a void-returning method (C#) or a sub procedure (VB).
var command = ReactiveCommand.Create<int, Unit>(
    integer => Console.WriteLine(integer)
);

// This outputs: 42
command.Execute(42).Subscribe();

All of the static factory methods that the ReactiveCommand class has will parameterize the resulting ReactiveCommand<TInput, TOutput> to be the return result of the method (i.e. if your async method returns Task<string>, your command will be ReactiveCommand<TInput, string>). This means, that subscribing to the command itself returns the results of the async method as an IObservable.

// An asynchronous command created from IObservable<int> that 
// waits 2 seconds and then returns 42 integer.
var command = ReactiveCommand.CreateFromObservable<Unit, int>(
    () => Observable.Return(42).Delay(TimeSpan.FromSeconds(2))
);

// Subscribing to the observable returned by `Execute()` will 
// tick through the value `42` with a 2-second delay.
command.Execute(Unit.Default).Subscribe();

// We can also subscribe to _all_ values that a command
// emits by using the `Subscribe()` method on the
// ReactiveCommand itself.
command.Subscribe(value => Console.WriteLine(value));

Synchronous commands

If your command is not CPU-intensive or I/O-bound then it probably makes sense to provide synchronous execution logic. You can do so by creating a command via ReactiveCommand.Create:

// Creates a command with synchronous execution logic
// which is always available for execution.
var command = ReactiveCommand.Create(
    () => Console.WriteLine("A reactive command is invoked!")
);

Asynchronous commands

One of the most important features of ReactiveCommand is its built-in facilities for orchestrating asynchronous operations, commands will block re-execution while executing. ReactiveCommands are fully integrated into the Reactive Extensions framework, providing an .IsExecuting property (of type IObservable<bool>) which tells you whether the command is currently executing. This is often useful if you want to trigger activity animations or you want to prevent other commands from executing while the command is executing.

It is important to know, that ReactiveCommand itself as an IObservable will never complete or OnError - errors that happen in the async method will instead show up on the ThrownExceptions property. If it is possible that your async method can throw an exception, you should subscribe to ThrownExceptions or the exception will be rethrown on the UI thread.

// Here we declare a ReactiveCommand, an OAPH and a property.
private readonly ObservableAsPropertyHelper<List<User>> _users;
public ReactiveCommand<Unit, List<User>> LoadUsers { get; }
public List<User> Users => _users.Value;

// Create a command with asynchronous execution logic. The 
// command is always available for execution, it triggers
// LoadUsersAsync method which returns Task<List<User>>.
LoadUsers = ReactiveCommand.CreateFromTask(LoadUsersAsync);

// Update the UI with a new value when users are loaded.
// ToProperty extension method allows us to subscribe to 
// LoadUsers Observable, update our OAPH and notify the UI 
// that the value of Users property has changed.
_users = LoadUsers.ToProperty(
    this, x => x.Users, RxApp.MainThreadScheduler
);

// Here we subscribe to all exceptions thrown by our 
// command and log them using ReactiveUI logging system.
// If we forget to do this, our application will crush
// if anything goes wrong in LoadUsers command.
LoadUsers.ThrownExceptions.Subscribe(exception => 
{
    this.Log().WarnException("Error!", exception);
});

Note For performance based solutions you can also use the nameof() operator override of ToProperty() which won't use the Expression. Read more on ObservableAsPropertyHelper here.

ReactiveCommand guarantees the result of events are delivered to the provided outputScheduler. The executing logic thread safety is the user's responsibility but any result from the logic is guaranteed to arrive on the specified outputScheduler. Read more on scheduling here.

Controlling executability

A ReactiveCommand may or may not be executable in a given situation. For example, the command backing the Save menu item might be unavailable if there are no unsaved changes. We pass into the ReactiveCommand an IObservable<bool> of when the ReactiveCommand should be allowed to be executed. The ReactiveCommand uses an IObservable eventing system to determine if execution should be allowed which differs from other frameworks where you might have the command continuous poll if execution is allowed. The ReactiveCommand approach has some performance advantages in that the value is cached between the can execute observable being fired. You commonly will create your can execute observable using the WhenAnyValue functions provided by the ReactiveUI framework:

// Each time values of UserName and Password properties change,
// the canExecute observable is signalled and the logic that 
// determines if command execution should be allowed is executed.
var canExecute = this.WhenAnyValue(
    x => x.UserName, x => x.Password,
    (userName, password) => 
        !string.IsNullOrEmpty(userName) && 
        !string.IsNullOrEmpty(password));

// The command will be unavailable during execution of LogOnAsync
// method, or while UserName and Password ViewModel properties are 
// nulls or empty strings. In other words, canExecute supplements 
// the default executability behavior, it doesn't replace it.
var command = ReactiveCommand.CreateFromTask(LogOnAsync, canExecute);

Parameters, unlike in other frameworks, are typically not used in the canExecute conditions, instead, binding View properties to ViewModel properties and then using the WhenAnyValue() is far more common.

Warning For performance reasons, ReactiveCommand does not marshal your canExecute observable to the main scheduler. You almost certainly want your canExecute observable to be ticking on the main thread, so be sure to add a call to ObserveOn if necessary.

Handling exceptions

If the logic you provide to a ReactiveCommand can fail in expected ways, you need a means of dealing with those failures. For command execution, the pipeline you get back from Execute will tick any errors that occur in your execution logic. However, the subscription to this observable is often instigated by the binding infrastructure. As such, it's likely that you cannot even get a hold of the observable to observe any errors.

To address this dilemma, ReactiveCommand includes a ThrownExceptions observable (of type IObservable<Exception>). Any errors that occur in your execution logic will also tick through this observable. If you haven't subscribed to it, ReactiveUI will bring down your application by default. This forces you towards a pit of error-handling success.

// Here we prevent LoadCommand from bringing our app down.
LoadCommand.ThrownExceptions.Subscribe(error => { });

Use RxApp.DefaultExceptionHandler if you'd like to override the default ThrownExceptions behaviour. This may be useful if you have crash analytics plugins installed and would like to handle all exceptions.

It can be tempting to always add a subscription to ThrownExceptions, even if the only recourse is to just log the problem. However, it is advisable to treat this like any other exception handling and only handle problems you can redress. If, for example, your command merely updates a property in your view model and it should never fail, any subscription to ThrownExceptions will serve only to obscure implementation problems. That said, be aware of the potential for intermittent problems, such as network and I/O errors. As always, a strong suite of tests will help you identify where a subscription to ThrownExceptions makes sense.

Note Your canExecute pipeline also has the potential to produce an error. Such cases are almost certainly a programmer error because you never want your canExecute pipeline to end in error. Even so, these errors will also tick through ThrownExceptions.

Unfortunately you can't filter exceptions from ThrownExceptions. It's an "all-or-nothing affair" as Kent Boogaart says in his book. You have to do it manually by handling the exceptions you can and forwarding all the others somewhere else like the RxApp.DefaultExceptionHandler. Use RxApp.DefaultExceptionHandler.OnNext(exceptionICantHandle).

Assume your command calls another command that might throw an exception. Likely, you would like to handle that exception only once, but ReactiveUI will propagate it to ThrownExceptions observables for both commands. See an example:

// Create a command with implementation that might throw an exception.
var commandA = ReactiveCommand.CreateFromTask(() => throw new Exception());
commandA.ThrownExceptions.Subscribe(ex => ErrorInteraction.Handle("Error in A!"));

// Create a command that calls command A.
var commandB = ReactiveCommand.CreateFromTask(async () =>
{
    // If command A throws an exception, the .Execute() method call 
    // will also throw. That's why we get two error notifications - 
    // one from commandA.ThrownExceptions and another from 
    // commandB.ThrownExceptions.
    await commandA.Execute(); // <= Could throw here!
    DoSomethingElse();
});

// If anything goes wrong in command A, we get ErrorInteraction handled twice.
commandB.ThrownExceptions.Subscribe(ex => ErrorInteraction.Handle("Error in B!"));

The easiest way of resolving this issue is using Throttle() operator over merged ThrownExceptions from both commands. See StackOverflow. Read more on handling Interactions here.

// Now our ErrorInteraction will be handled only once if command A throws!
commandA.ThrownExceptions.Merge(commandB.ThrownExceptions)
    .Throttle(TimeSpan.FromMilliseconds(250), RxApp.MainThreadScheduler)
    .Subscribe(error => ErrorInteraction.Handle("Error in B!"));

Invoking commands

The best way to execute ReactiveCommands is via the Execute() method:

// Create a command with asynchronous execution logic.
LoadUsers = ReactiveCommand.CreateFromTask(LoadUsersAsync);

// Invoke LoadUsers command using async/await syntax.
// We could also use .Subscribe() here.
var users = await LoadUsers.Execute();
Console.WriteLine("You've got {0} users!", users.Count());

Regardless of whether your command is synchronous or asynchronous in nature, you execute it via the Execute method. You get back an observable that will tick the command's result value when execution completes. Synchronous commands will execute immediately, so the observable you get back will already have completed. The returned observable is behavioral though, so subscribing after the fact will still tick through the result value.

Warning As is often the case with idiomatic Rx, the observable returned by Execute is cold. That is, nothing will happen unless something subscribes to it or awaits it. In those cases where you're calling Execute directly, it's very important to remember that it's lazy.

ReactiveCommand implements the ICommand for UI framework compatibility and backwards compatibility only. It is recommended you don't use the ICommand interface directly in your code. ReactiveCommand is explicitly derived from the ICommand interface to avoid users accidentally calling the non-reactive style methods. The ICommand methods do not lend well to long-running and also asynchronous commands, such as those that perform I/O operations. The ICommand also focuses on an imperative style of execution over the reactive style.ReactiveCommand provides methods and observable properties that are the equivalent of the ICommand interface. Execute() provides an Observable which you can Subscribe() to execute the logic of the ReactiveCommand and CanExecute is also exposed through a read-only property. Additionally ReactiveCommand provides the IsExecuting observable which is functionally not provided by the ICommand interface.

Hint Try not to execute commands in the ViewModel constructor. If commands are invoked in the constructor, your ViewModel classes become more difficult to test, because you always have to mock out the effects of calling that commands, even if the thing you are testing is unrelated. Instead, use WhenActivated.

Invoking commands in an Observable pipeline

At times it can be convenient to execute a command in response to some Observable<T> that isn't perhaps tied to a user interaction. For example, a feature that automatically saves the current document by executing a ReactiveCommand every 5 minutes. The InvokeCommand extension makes it easy to achieve this:

// Creates a hot Observable<T> that emits a new value every 5 
// minutes and invokes the SaveCommand<Unit, Unit>. Don't forget
// to dispose the subscription produced by InvokeCommand().
var interval = TimeSpan.FromMinutes(5);
Observable.Timer(interval, interval)
    .Select(time => Unit.Default)
    .InvokeCommand(this, x => x.SaveCommand);

Hint InvokeCommand respects the command's executability. That is, if the command's CanExecute method returns false, InvokeCommand will not execute the command when the source observable ticks.

Combining commands

At times it can be useful to have several commands aggregated into one. As an example, consider a browser that allows the user to clear individual caches (browsing history, download history, cookies), or clear all caches. There would be a command for clearing each individual cache, each of which might have its own logic to dictate the executability of the command. It would be onerous and error-prone to have to repeat or combine all this logic for the command that clears all caches. Combined commands provide an elegant means of addressing this situation:

var clearBrowsingHistory = ReactiveCommand.CreateFromObservable(
    this.ClearBrowsingHistoryAsync, canClearBrowsingHistory);

var clearDownloadHistory = ReactiveCommand.CreateFromObservable(
    this.ClearDownloadHistoryAsync, canClearDownloadHistory);

var clearCookies = ReactiveCommand.CreateFromObservable(
    this.ClearCookiesAsync, canClearCookies);

// Combine all these commands into one "parent" command.
// This "parent" command will respect the executability 
// of all child commands defined above.
var clearAll = ReactiveCommand.CreateCombined(
    new [] { clearBrowsingHistory, 
             clearDownloadHistory, 
             clearCookies });

The combined command will execute the child commands asynchronously when executed. The combined command respects the executability of all child commands. That is, if any child command cannot currently execute, neither can the combined command. In addition, it is also possible for you to pass in extra executability logic when creating your combined command:

// In this case, `clearAll` command will only be 
// executable if all child commands are executable 
// _and_ the latest value from `canClearAll` is `true`.
IObservable<bool> canClearAll = ...;
var clearAll = ReactiveCommand.CreateCombined(
    new [] { clearBrowsingHistory, 
             clearDownloadHistory, 
             clearCookies },
    canClearAll);

All child commands provided to the CreateCombined method must be of the same type. You cannot combine, say, a ReactiveCommand<Unit, Unit> with a ReactiveCommand<int, Unit>. Nor can you combine, say, a ReactiveCommand<Unit, Unit> with a ReactiveCommand<Unit, int>. This is because all child commands will receive the parameter provided to the combined command, and the result of executing the combined command is a list of all child results.

Controlling scheduling

By default, ReactiveCommand uses RxApp.MainThreadScheduler to surface events. That is, values from CanExecute, IsExecuting, ThrownExceptions, and result values from the command itself. Typically UI components are subscribed to these observables, so it's a sensible default. However, when writing unit tests for your view models, you may want more control over scheduling. All Create* methods take an optional outputScheduler parameter, so you can pass in a custom scheduler if you need to:

var command = ReactiveCommand.Create(() => { }, outputScheduler: someScheduler);

It's important to understand that the execution logic for a reactive command is not scheduled to execute on the provided scheduler (just as is the case for any canExecute observable you provide). Instead, it is left to the caller to implement any required scheduling inside their execution pipeline. This means it is entirely possible for your execution logic to execute on a thread other than that owned by the provided scheduler:

var command = ReactiveCommand.Create(
    () => Console.WriteLine(Environment.CurrentManagedThreadId), 
    outputScheduler: RxApp.MainThreadScheduler
);

// This will output the ID of the thread from which you make 
// this call, not necessarily the ID of the main thread!
command.Execute().Subscribe();

Note If you're using ReactiveUI's With extension method in your tests, you can create commands using the default scheduling behavior. That's because the With extension method will switch out RxApp.MainThreadScheduler with the scheduler you provide it.

Bindings

ReactiveCommand can be connected to the View by either using XAML binding on supported platforms, or using the inbuilt ReactiveUI binding method BindCommand. Use of BindCommand is preferred but not required where XAML bindings are supported. Read more on this here.

Unit Testing

Read: http://kent-boogaart.com/blog/using-the-visual-studio-test-runner-for-mobile-development

Don't mock ReactiveCommands. ReactiveCommand itself is already designed around testability. Also, the likelihood that you will correctly mock ReactiveCommand semantics via Moq is pretty low, it's a pretty complicated class (and if you did, you would end up doing a ton of unnecessary work).