Defining boundaries and communicating between aggregate roots_问答_开发者

I could use some help understanding my domain model a bit and making sure I am approaching the design correctly.

I have an aggregate root called Department. The Department object has several child value types that help define the business notion of a 'department.' In my UI users can list, create, edit and delete Department objects.

I have another aggregate root called Project. A Project has several child value types but also has a relationship to a Department in that each Project is 'owned' by a department. Projects can be created, edited, deleted, etc. and doing so has no impact on the department whereas removing a Department also removes any projects it owns.

My UI will display a list of Projects based on the departments the current user is authorized to access. They may be able to access more than one department. When displayed 开发者_JS百科as both a list item as well as in detail, I need to show the Department logo with the Project.

My first thought was the Project was an aggregate root with a simple DepartmentID property that could be used to 'lookup' the Department. But now I'm starting to think that I really only have one aggregate root: Department.

What do you think?

UPDATE

I don't know if it is key to the discussion or changes anything but the following thought occurred to me after reading the first couple of answers.

Department appears to have two contexts:

As a stand-alone entity that supports modification.
As a child of a Project in which case is contains read-only data and no behavior.

This makes me think that I should have two 'objects' in my model, an aggregate root for case #1 and a value type for case #2. Am I on the right track?

Since the Project can't exist without a Department it's probably not an Aggregate Root. From your description it sounds like you only have one AR - the Department, which you use to manage the projects inside it.

If your behavior is mostly CRUD, i would not recommend building a full blown domain model for it since there are probably simpler approaches you can take.

UPDATE As you mention, you might have 2 bounded contexts here. One where the Department is an AR and the Projects are entities of this AR. In this context you would perform operations on your Departments. In a second BC your Project could be the AR and the Departments could be entities or VOs. This would allow you to work directly with projects.

I would also recommend going over this with your domain expert and see if these concepts fit well in your UL, or maybe search for some missing concept that can clarify your model. I would especially look for a concept that might link projects to departments.

I think it's perfectly defensible to have both Project and Department be aggregate roots, since they are both managed independently.

That is, every Project and every Department have some kind of unique identifier, and while you can add Projects to Departments, Projects themselves are probably complicated enough (with their own lifecycles, their own child objects etc.) to warrant aggregate root status.

You just have to keep a reference to the Department in each Project.

Few simple questions to be answered:

1) can the department domain object live by its own without the Project domain object. - If yes, then the department is an aggregate.

2) Is the Project domain object can live by its own - If yes, then the Project is also an aggregate

3) Is Project has relationship with one Department - Then it should be part of the Project aggregate exposed as property Department

4) Is Department has relation ship with one or more Project objects - The the Project aggregate should be part of the Department aggregate object.

So, Using Department aggregate object you might need to access the list of Project(s) object and once you have the Project object you might need to access the Department object. It is a circular reference which is obsoletely fine.

It is a typical example of Employee has a manager and manager has a list of employees