[Suggestion] Acces to Playhead property of EventSourcedAggregateRoot sub class

[broadway]

[Suggestion] Acces to Playhead property of EventSourcedAggregateRoot sub class

[asm89]

You're aggregate root has a certain state (derived from the history of events if you're doing event sourcing). Based on the state the method call on the AR will either fail or succeed?

If you're using the DBAL event store, it will throw an exception if the same playhead gets inserted twice (for one id), that probably covers your "problem"? :)

[macintoshplus]

Let me explain:

In the Model Reading I keep the version of the aggregat linked to the last event.

User 1 : creates a aggregat (version 0).

User 1 and 2 : load the view for the aggregat (version 0).

User 2 : change the aggregat by a command indicating that the aggregat must be version 0. (change version from 0 to version 1)

User 1 : changes aggregat by a command indicating that the aggregat must be version 0. Changes can not be made because it is based on data been deprecated.

I do not know if I express myself well. It is to avoid this "problem" I ask this question;-)

So what's the best solution?

[asm89]

I'd have to think about a "nice" way of doing this, but I'm wondering if you really need this? Do you have an example of your actual domain? What are the chances of this happening? What would the cost of such an error be?

As said before, if two processes are committing at the same time, only one of them will be able to save the events as of this moment.

[macintoshplus]

I have an application where the risk of competition between two users is very high.

Another solution would be to check the event happened between the two versions (current and those notified in the order) to block similar commands.

Otherwise, the last command to the last word!

The concurrency is very difficult to manage.

If a command overwrites the previous amendment, I must provide a means to reapply the changes overwritten.

[aaronmu]

👎

I think you need to deal with this yourself. Perhaps some sort of process manager that detects concurrent events and initiates merging? Perhaps you need a different model? I'm just thinking out loud here.

[macintoshplus]

It's true. Thank for your help !