Getting started

Initializing EventFlow always start with an EventFlowOptions.New as this performs the initial bootstrap and starts the fluent configuration API. The very minimum initialization of EventFlow can be done in a single line, but wouldn’t serve any purpose as no domain has been configured.

var resolver = EventFlowOptions.New.CreateResolver();

The above line does configures several important defaults

  • Custom internal IoC container
  • In-memory event store
  • Console logger
  • A “null” snapshot store, that merely writes a warning if used (no need to do anything before going to production if you aren’t planning to use snapshots)
  • And lastly, default implementations of all the internal parts of EventFlow

Important

Before using EventFlow in a production environment, you should configure an alternative event store, an alternative IoC container and another logger that sends log messages to your production log store.

To start using EventFlow, a domain must be configure which consists of the following parts

In addition to the above, EventFlow provides several optional features. Whether or not these features are utilized, depends on the application in which EventFlow is used.

Example application

To get started, we start with our entire example application which consists of one of each of the required parts: aggregate, event, aggregate identity, command and a command handler. After we will go through the individual parts created.

Note

The example code provided here is located within the EventFlow code base exactly as shown, so if you would like to debug and step through the entire flow, checkout the code and execute the GettingStartedExample test.

https://github.com/eventflow/Documentation/tree/master/Source/EventFlow.Documentation/GettingStarted

All classes create for the example application are prefixed with Example.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
// We wire up EventFlow with all of our classes. Instead of adding events,
// commands, etc. explicitly, we could have used the the simpler
// AddDefaults(Assembly) instead.
using (var resolver = EventFlowOptions.New
    .AddEvents(typeof(ExampleEvent))
    .AddCommands(typeof(ExampleCommand))
    .AddCommandHandlers(typeof(ExampleCommandHandler))
    .UseInMemoryReadStoreFor<ExampleReadModel>()
    .CreateResolver())
{
    // Create a new identity for our aggregate root
    var exampleId = ExampleId.New;

    // Define some important value
    const int magicNumber = 42;

    // Resolve the command bus and use it to publish a command
    var commandBus = resolver.Resolve<ICommandBus>();
    var executionResult = await commandBus.PublishAsync(
        new ExampleCommand(exampleId, magicNumber),
        CancellationToken.None)
        .ConfigureAwait(false);

    // Verify that we didn't trigger our domain validation
    executionResult.IsSuccess.Should().BeTrue();

    // Resolve the query handler and use the built-in query for fetching
    // read models by identity to get our read model representing the
    // state of our aggregate root
    var queryProcessor = resolver.Resolve<IQueryProcessor>();
    var exampleReadModel = await queryProcessor.ProcessAsync(
        new ReadModelByIdQuery<ExampleReadModel>(exampleId),
        CancellationToken.None)
        .ConfigureAwait(false);

    // Verify that the read model has the expected magic number
    exampleReadModel.MagicNumber.Should().Be(42);
}

The above example publishes the ExampleCommand to the aggregate with the exampleId identity with the magical value of 42. After the command has been published, the accompanying read model ExampleReadModel is fetched and we verify that the magical number has reached it.

During the execution of the example application, a single event is emitted and stored in the in-memory event store. The JSON for the event is shown here.

{
  "MagicNumber": 42
}

The event data itself is straightforward as it is merely the JSON serialization of an instance of the type ExampleEvent with the value we defined. A bit more interesting is the metadata that EventFlow stores along the event, which is used by the EventFlow event store.

{
  "timestamp": "2016-11-09T20:56:28.5019198+01:00",
  "aggregate_sequence_number": "1",
  "aggregate_name": "ExampleAggrenate",
  "aggregate_id": "example-c1d4a2b1-c75b-4c53-ae44-e67ee1ddfd79",
  "event_id": "event-d5622eaa-d1d3-5f57-8023-4b97fabace90",
  "timestamp_epoch": "1478721389",
  "batch_id": "52e9d7e9-3a98-44c5-926a-fc416e20556c",
  "source_id": "command-69176516-07b7-4142-beaf-dba82586152c",
  "event_name": "example",
  "event_version": "1"
}

All the built-in meta data is available on each instance of IDomainEvent<,,>, which is accessible from event handlers for e.g. read models or subscribers. It also possible create your own meta data providers or add additional EventFlow built-in providers as needed.

Aggregate identity

The aggregate ID is in EventFlow represented as a value objected that inherits from the IIdentity interface. You can provide your own implementation, but EventFlow provides a convenient implementation that will suit most needs. Be sure to read the read the section about the Identity<> class to get details on how to use it.

For our example application we use the built-in class making the implementation very simple.

1
2
3
4
5
6
/// Represents the aggregate identity (ID)
public class ExampleId :
    Identity<ExampleId>
{
    public ExampleId(string value) : base(value) { }
}

Aggregate

Now we’ll take a look at the ExampleAggregate. Its rather simple as the only thing it can, is apply the magic number once.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
public class ExampleAggregate :
    AggregateRoot<ExampleAggregate, ExampleId>,
    IEmit<ExampleEvent>
{
    private int? _magicNumber;

    public ExampleAggregate(ExampleId id) : base(id) { }

    // Method invoked by our command
    public IExecutionResult SetMagicNumer(int magicNumber)
    {
        if (_magicNumber.HasValue)
            return ExecutionResult.Failed("Magic number already set");

        Emit(new ExampleEvent(magicNumber));
        
        return ExecutionResult.Success();
    }

    // We apply the event as part of the event sourcing system. EventFlow
    // provides several different methods for doing this, e.g. state objects,
    // the Apply method is merely the simplest
    public void Apply(ExampleEvent aggregateEvent)
    {
        _magicNumber = aggregateEvent.MagicNumber;
    }

Be sure to read the section on aggregates to get all the details right, but for now the most important thing to note, is that the state of the aggregate (updating the _magicNumber variable) happens in the Apply(ExampleEvent) method. This is the event sourcing part of EventFlow in effect. As state changes are only saved as events, mutating the aggregate state must happen in such a way that the state changes are replayed the next the aggregate is loaded. EventFlow has a set of different approaches that you can select from, but in this example we use the Apply methods as they are the simplest.

Important

The Apply(ExampleEvent) is invoked by the Emit(...) method, so after the event has been emitted, the aggregate state has changed.

The ExampleAggregate exposes the SetMagicNumer(int) method, which is used to expose the business rules for changing the magic number. If the magic number hasn’t been set before, the event ExampleEvent is emitted and the aggregate state is mutated.

If the magic numer was changed, we return a failed IExecutionResult with an error message. Returning a failed execution result will make EventFlow disregard any events the aggregate has emitted.

If you need to return something more useful than a bool in a execution result, merely create a new class that implements the IExecutionResult interface and specific the type as generic arguments for the command and command handler.

Note

While possible, do not use the execution results as a method of reading values from the aggregate, that’s what the IQueryProcessor and read models are for.

Event

Next up is the event which represents some that has happend in our domain. In this example, its merely that some magic number has been set. Normally these events should have a really, really good name and represent something in the ubiquitous language for the domain.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
/// A basic event containing some information
[EventVersion("example", 1)]
public class ExampleEvent :
    AggregateEvent<ExampleAggregate, ExampleId>
{
    public ExampleEvent(int magicNumber)
    {
        MagicNumber = magicNumber;
    }

    public int MagicNumber { get; }
}

We have applied the [EventVersion("example", 1)] to our event, marking it as the example event version 1, which directly corresponds to the event_name and event_version from the meta data store along side the event mentioned. The information is used by EventFlow to tie name and version to a specific .NET type.

Important

Even though the using the EventVersion attribute is optional, its highly recommended. EventFlow will infer the information if it isn’t provided and thus making it vulnerable to e.g. type renames.

Important

Once have aggregates in your production environment that have emitted an event, you should never change the .NET implementation. You can deprecate it, but you should never change the type or the data stored in the event store.

Command

Commands are the entry point to the domain and if you remember from the example application, they are published using the ICommandBus as shown here.

1
2
3
4
5
var commandBus = resolver.Resolve<ICommandBus>();
var executionResult = await commandBus.PublishAsync(
    new ExampleCommand(exampleId, magicNumber),
    CancellationToken.None)
    .ConfigureAwait(false);

In EventFlow commands are simple value objects that merely how the arguments for the command execution. All commands implement the ICommand<,> interface, but EventFlow provides an easy-to-use base class that you can use.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
/// Command for update magic number
public class ExampleCommand :
    Command<ExampleAggregate, ExampleId, IExecutionResult>
{
    public ExampleCommand(
        ExampleId aggregateId,
        int magicNumber)
        : base(aggregateId)
    {
        MagicNumber = magicNumber;
    }

    public int MagicNumber { get; }
}

A command doesn’t do anything without a command handler. In fact, EventFlow will throw an exception if a command doesn’t have exactly one command handler registered.

Command handler

The command handler provides the glue between the command, the aggregate and the IoC container as it defines how a command is executed. Typically they are rather simple, but they could contain more complex logic. How much is up to you.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
/// Command handler for our command
public class ExampleCommandHandler :
    CommandHandler<ExampleAggregate, ExampleId, IExecutionResult, ExampleCommand>
{
    public override Task<IExecutionResult> ExecuteCommandAsync(
        ExampleAggregate aggregate,
        ExampleCommand command,
        CancellationToken cancellationToken)
    {
        var executionResult = aggregate.SetMagicNumer(command.MagicNumber);
        return Task.FromResult(executionResult);
    }
}

The ExampleCommandHandler in our case here merely invokes the SetMagicNumer on the aggregate and returns the execution result. Remember, if a command handler returns a failed execution result, EventFlow will disregard any events the aggregate has emitted.

Important

Everything inside the ExecuteAsync(...) method of a command handler may be executed more than once if there’s an optimistic concurrency exception, i.e., something else has happened to the aggregate since it as loaded from the event store and its therefor automatically reloaded by EventFlow. It is therefor essential that the command handler doesn’t mutate anything other than the aggregate.

Read model

If you ever need to access the data in your aggregates efficiently, its important that read models are used. Loading aggregates from the event store takes time and its impossible to query for e.g. aggregates that have a specific value in its state.

In our example we merely use the built-in in-memory read model store. Its useful in many cases, e.g. executing automated domain tests in an CI build.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
/// Read model for our aggregate
public class ExampleReadModel :
    IReadModel,
    IAmReadModelFor<ExampleAggregate, ExampleId, ExampleEvent>
{
    public int MagicNumber { get; private set; }

    public void Apply(
        IReadModelContext context,
        IDomainEvent<ExampleAggregate, ExampleId, ExampleEvent> domainEvent)
    {
        MagicNumber = domainEvent.AggregateEvent.MagicNumber;
    }
}

Notice the IDomainEvent<ExampleAggrenate, ExampleId, ExampleEvent> domainEvent argument, its merely a wrapper around the specific event we implemented earlier. The IDomainEvent<,,> provides additional information, e.g. any meta data store along side the event.

The main difference between the event instance emitted in the aggregate and the instance wrapped here, is that the event has been committed to the event store.

Next steps

Although the implementation in this guide enables you to create a complete application, there are several topics that are recommended as next steps.