Question

Can aggregate roots have childs which also is aggregate roots? If yes, does it only reference it or does the aggregate root that holds a property of the other aggregate root have methods that changes it through that method?

Let say you have a class named “Worker” and another named “Company”. Both are aggregate roots. Company have a property of Worker. Sorry for making a bad example

public class Company {
    private Worker worker;

    ...

    public Worker getWorker() {
        ...
    }
}


public class Worker {
    ...
}

or does the company class “hide” the worker?

public class Company {
    private Worker worker;

    ...

    public String getWorkerName() {
        ...
    }
}


public class Worker {
    ...
}

And does worker talk only to the company class because it is not part of the company context? Why so? Are aggregate roots contained inside other aggregate roots always accessed like my first example? (I would think so, but I have no reason why)

Answer 1

In short, no.

This is kinda tough to explain succinctly in a SO answer, but here’s a few bullet points that may help your understanding:

• An aggregate root is a consistency boundary. Basically, all the data that needs to be kept consistent belongs in the aggregate root. You’ll know it’s structured correctly when it has high cohesion – most methods touch a lot of private properties.
• All interaction with an aggregate root or its children should be done via public methods on the aggregate root. In your example, if a Worker belongs to a Company then adding/removing a Worker or performing any operations on a Worker should be done via public methods on Company.
• You can have entities below aggregate roots, but in my experience this is often a design smell. Most often in my experience, you have value objects as children of aggregates.
• Correlation between aggregate roots is done via an ID, so Company may know about a GUID called WorkerId, but it definitely should not be coupled to the Worker class. (This will allow you to refactor more easily.)
• You only need a domain model (and therefore aggregate roots) if an operation (i.e. a method) depends on the previous state of the data. If there’s no state changes involved, just do CRUD.