Shuttle.Core.Mediator #

PM> Install-Package Shuttle.Core.Mediator

The Shuttle.Core.Mediator package provides a mediator pattern implementation.

Configuration #

In order to get all the relevant bits working you would need to register the IMediator dependency along with all the relevant IParticipant dependencies.

You can register the mediator using IServiceCollection:

services.AddMediator(builder =>
{
    builder.AddParticipants(assembly);
    builder.AddParticipant<Participant>();
    builder.AddParticipant(participantType)
    builder.AddParticipant<Message>(participant)
});

IMediator #

The core interface is the IMediator interface and the default implementation provided is the Mediator class.

This interface provides a synchronous calling mechanism and all IParticipant implementations need to be thread-safe singleton implementations that are added to the mediator at startup. Any operations that require transient mechanisms should be handled by the relevant participant.

void Send(object message, CancellationToken cancellationToken = default);

The Send method will find all participants that implements the IParticipant<T> with the type T of the message type that you are sending. Participants that are marked with the BeforeParticipantAttribute filter will be executed first followed by all participants with no filter attributes and then finally all participants marked with the AfterParticipantAttribute filter will be called.

Extensions #

Task SendAsync(this IMediator mediator, object message, CancellationToken cancellationToken = default)

Sends a message asynchronously.

public static T Send<T>(this IMediator mediator, T message, CancellationToken cancellationToken = default);

The same as Send except that it returns the given message.

IParticipant #

public interface IParticipant<in T>
{
    void ProcessMessage(IParticipantContext<T> context);
}

A participant would handle the message that is sent using the mediator. There may be any number of participants that process the message.

Design philosophy #

There are no request/response semantics and the design philosophy here is that the message encapsulates the state that is passed along in a pipes & filters approach.

However, you may wish to make use of one of the existing utility classes:-

RequestMessage<TRequest> #

The only expectation from a RequestMessage<TRequest> instance is either a success or failure (along with the failure message).

RequestResponseMessage<TRequest, TResponse> #

The RequestResponseMessage<TRequest, TResponse> takes an initial TRequest object and after the mediator processing would expect that there be a TResponse provided using the .WithResponse(TResponse) method. The same success/failure mechanism used in the RequestMessage<TRequest> class is also available on this class.

Considerations #

If you have a rather predictable sequential workflow and you require something faster execution then you may wish to consider the Shuttle.Core.Pipelines package. A performance testing application for your use-case would be able to indicate the more suitable option.