Published at 24 March 2018
EventSauce is designed to be pragmatic. It’s pragmatic because it’s easy to setup and easy to extend. The library is focused on getting you up and running quickly without sacrificing the conceptual integrity.
The core of EventSauce revolves around a set of 3 interfaces:
AggregateRootRepository is the main interface. It is responsible for
retrieving and persisting aggregate root objects. It uses the other two
parts for retrieving and storing
Message objects (in which events are transported)
and dispatching messages to
Consumers. Because EventSauce is based around these
interfaces it’s very easy to modify how the library behaves. The interfaces also
make EventSauce highly customizable without the need for inheritance.
MessageDispatcher you use is totally up to you. There
are benefits (and downsides) to each queueing mechanism and message repository. Because
EventSauce places these implementations behind an interface you’re free to choose whatever
fits best. You can even create your own repositories and dispatchers, the interfaces are
AggregateRootRepository- custom aggregate root construction and persistence
MessageRepository- custom message storage (database)
MessageDispatcher- customer message dispatching (queue)
ClassNameInflector- custom inflection for class-names to event names (interoperability)
MessageSerializer- customer message serialization (storage)
EventSerializer- customer event serialization (storage)
The default implementation of the
AggregateRootRepository is the
ConstructingAggregateRootRepository. This repository requires your aggregate root
to implement the
AggregateRoot interface. This implementation supports simple
aggregate root reconstitution and versioning for sequential integrity.
The library ships with an
InMemoryMessageRepository (which can be used for testing)
SynchronousMessageDispatcher. Apart from that it ships with composition helpers
such as the
MessageDispatcherChain. The dispatcher chain allows you to chain dispatchers,
this allows you to combine synchronous and asynchronous dispatching in the same composition.
Below are short descriptions of all parts that make up EventSauce.
The aggregate root is your primary modelling space. It’s tasked with maintaining the integrity of the model, guarding invariants, and recording events.
Aggregate Root Repository
The aggregate root repository is used to retrieve and “persist” aggregate root entities. It’s your main point of interaction when using the library. You retrieve the aggregate root from it. After you’ve interacted with the aggregate root this is also where you persist the newly raised events.
The aggregate root repository has the following dependencies:
- An aggregate root class name (so it knows what to reconstitute and return)
- A message repository from which it retrieves previously recorded events
- A message dispatcher which dispatches the messages (optional)
- A message decorator which decorated the messages (optional)
The message repository is the library’s connection to the persistence layer. It’s responsible
for storing and retrieving events. Events stored in the repository are wrapped in a
object. This object allows you to store additional metadata alongside the event data.
The message dispatcher is responsible for sending messages to
Consumers. The core
library ships with a
SynchronousMessageDispatcher for when you don’t need to process
messages asynchronously. If you do want to process events in the background there are
multiple options available.
A message decorator has the ability to enrich messages before they are persisted and dispatched. This could be a request identifier so you can track an action from the web all the way down to background processes. This mechanism prevents you from polluting the domain events with specific information.
The message serializer is responsible for converting messages from and to a serialized form. When using (or implementing) a message repository, you’ll want to use this rely on this interface.
The core ships with a default (JSON based) serializer. You’re free to implement your own serialization strategy if your use-case requires it.
Message serializers can be composed using decoration to provide more complex features such as upcasting.
Time is a very important concept in EventSauce. In the core of the library a
Clock is defined.
SystemClock provides a production-ready implementation, while the
TestClock is used during testing.
TestClock allows you to fixate time, allowing you to test processes without worrying about the current
time changing under you.