php - In DDD how to pass aggregate into methods of other aggregates, entirely or id? - Stack Overflow

admin管理员组

文章数量:1346662

I know that If one aggregate has reference to another, the first one has to have ID property of the second. For example:

FooId extends AbstractIdValueObject { }

Foo extends AggregateRoot { }

Bar extends AggregateRoot {
   private FooId $fooId
}

But what should be done for the methods? Which is correct:

User {
    changeGroup(Group $group) {
        $this->groupId = $group->id;
    }
}

or

User {
    changeGroup(GroupId $groupId) {
        $this->groupId = $groupId;
    }
}

NOTE: it is very simplified example. In the real project, number of parameters can be more, including 2-3 aggregates.

The first example shows/describes real world: we change group, so even customer who doesn't know the program language can understand it...so we pass group into method. But here is a problem. I get groupId from request, then get the whole object by id from a repository (database), pass it to the method, inside method I get that id...It's weird and have unnecessary request to a database. It's more bad if there will be more aggregates

The second example is simpler and doesn't have unnecessary requests to repository(ies). But now it's not kind of "domain/ubiquitous language style"? Or I'm wrong?

Please, tell me what the right practice of doing it? How to do it and why.

I know that If one aggregate has reference to another, the first one has to have ID property of the second. For example:

FooId extends AbstractIdValueObject { }

Foo extends AggregateRoot { }

Bar extends AggregateRoot {
   private FooId $fooId
}

But what should be done for the methods? Which is correct:

User {
    changeGroup(Group $group) {
        $this->groupId = $group->id;
    }
}

or

User {
    changeGroup(GroupId $groupId) {
        $this->groupId = $groupId;
    }
}

NOTE: it is very simplified example. In the real project, number of parameters can be more, including 2-3 aggregates.

The first example shows/describes real world: we change group, so even customer who doesn't know the program language can understand it...so we pass group into method. But here is a problem. I get groupId from request, then get the whole object by id from a repository (database), pass it to the method, inside method I get that id...It's weird and have unnecessary request to a database. It's more bad if there will be more aggregates

The second example is simpler and doesn't have unnecessary requests to repository(ies). But now it's not kind of "domain/ubiquitous language style"? Or I'm wrong?

Please, tell me what the right practice of doing it? How to do it and why.

• php
• domain-driven-design

Share Improve this question Follow asked 2 days ago Alexey ShimanskyAlexey Shimansky 63222 gold badges1212 silver badges3838 bronze badges 2

• 1 I think you hit the nail on the head: If only ever the id is needed by the method, supply the id, but if there's a chance that anything else from the "aggregate" might be needed, then ask for the whole aggregate. Remember, you don't have to choose one or the other for all methods, you are allowed to vary it. It is my experience that many people try to conform too strict to certain design principles instead of thinking logically for themselves. Design principles are there to guide you, not to be used as diktats. – KIKO Software Commented 2 days ago
• yep, you're right. I came to this conclusion about 10 minutes after I posted it)) – Alexey Shimansky Commented 2 days ago

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 0

In DDD how to pass aggregate into methods of other aggregates, entirely or id?

The heuristic that I like: prioritize making it easier for the next programmer to understand what is going on.

An AGGREGATE ("a cluster of associated objects that we treat as a unit for the purpose of data changes") is really two different things - a bunch of information behind a facade of affordances for manipulating that information.

Robert Martin's Interface Segregation Principle asserts that clients should not be forced to depend upon interfaces that they don't use.

If User is "just" copying information from Group, and not invoking any of the methods that change the group data, then you will probably communicate that more clearly with an interface that accepts values rather than an interface that accepts entities.

(See also: CQRS.)

Restricting yourself to values means that you don't need Group, and therefore you don't need GroupRepository, and therefore anyplace that you do find a GroupRepository is advertising the fact there is an intent to change Group.

That said, it's a trade off: yes, you make changes to groups easier to reason about, but also

• As you noticed, it's not particularly well aligned with the language of the business (Pat is a Platinum Honors customer, not Pat holds a reference to the Platinum Honors group identifier)
• You lose some of the maintenance benefits of hiding the decision to not change the group in this method.

---

Here's the really bad news: "toy" problems provide very little insight about the design trade offs when you are working with complex code manages critical competitive advantages in your domain.

本文标签： phpIn DDD how to pass aggregate into methods of other aggregatesentirely or idStack Overflow