我可以按数据源创建存储库还是应该按聚合创建存储库？

Question

Right now I'm implementing repository pattern on my project, and I have doubts about how exactly should I implement repository pattern. My idea was to create repository by data source, eg RedisRepository, DbRepository. And I have not found any strict rules on what repository should be based.

So my question is it a good practice to create repository based on data source, or in any case the best practice is to create repository by aggregate eg CarRepository? Maybe you can also share some nice articles related to this topic, if you know ones.

Answer 1

If you're adopting the domain-driven design approach, you should have a repository for each aggregate root: https://learn.microsoft.com/en-us/dotnet/architecture/microservices/microservice-ddd-cqrs-patterns/infrastructure-persistence-layer-design .

Answer 2

Within Hexagonal architecture, the standard way to do this would be like so:

Application layer defines the repository interface (eg what it needs) and here you intentionally omit any persistence technology details, because it doesn't care how the functionality is provided. It might look like this:

interface CarPersistence {
    update(car);
    findOne(model, brand, ...);
}

In your Infrastructure layer , you will create a repository implementation for a specific persistence technology that will implement this interface, thus providing the required functionality to the application layer. This implementation defines how the needed functionality is provided.

class MongoDbCarPersistence implements CarPersistence {
    // ...
}

The key aspect is that the interface definition in the Application layer is technology agnostic. So whether you implement the actual repository for MongoDB or MySQL would have no impact on the interface, and therefore on any class using that interface. This is one of the main purposes of Hexagonal architecture.