Review AkkaTaggerAdapter's API

We have recently introduced the AkkaTaggerAdapter, but we may need to reconsider if we should pass the EntityContext or simply a String.

The method that is defined as:

def fromLagom[Event <: AggregateEvent[Event]](
  entityCtx: EntityContext,
  lagomTagger: AggregateEventTagger[Event]
): Event => Set[String]

The main reason we chose to pass the EntityContext is that we can then make sure that we will use EntityContext.entityId when generating the virtual partition numbers.

However, that puts a constraint on how one can initialize the EventSourcedBehavior and that has an impact on testing as it requires EntityContext on tests.

We should reconsider this. Maybe the method signature should be:

def fromLagom[Event <: AggregateEvent[Event]](
  entityId: String,
  lagomTagger: AggregateEventTagger[Event]
): Event => Set[String]

Anyway, I just want to raise the flag, so we tackle that before we cut the first RC.

@ignasi35, since you have been experimenting with testing and Akka Persistence Typed, please dump here all your thoughts.

[dwijnand]

We’ll need to address this later, quite probably on Akka’s side. For now we’ll keep EntityContext, which favours some more type-safety at the cost of some lack of ergonomics (having to use null for the sharing actor, in tests).

[octonato]

And, in the java implementation in particular, we won’t even need a private EntityContext entityContext field.

We won’t need a private EntityContext, but we will need a private entityId. We will need to fill it at construction time using the passed EntityContext. We can’t use the PersistenceId because we will need to parse it.