Add DomainCollection concept

[sairyss]

Add DomainCollection concept

[Sairyss]

I'm not sure what exactly you are trying to solve here. I don't see any value of saving or loading child entities separately from their parent aggregate. They are a part of the same root and have to be stored/loaded together.
Imagine this aggregate root:

class UserAggregate {
  address: UserAddressEntity; // <-- child entity
  details: UserDetailsEntity; // <-- child entity
}

Will have something similar to this OrmEntity:

class UserOrmEntity {
  details: UserDetailsOrmEntity; // <-- child ORM entity
  address: UserAddressOrmEntity; // <-- child ORM entity
}

Mapper:

class UserMapper {
  toDomain(ormEntity: UserOrmEntity): UserAggregate {
    const user = new UserAggregate();
    user.details = ormEntity.details;
    user.address = ormEntity.address;
    return user;
  }

  toOrmEntity(aggregate: UserAggregate): UserOrmEntity {
    const userOrm = new UserOrmEntity();
    userOrm.details = aggregate.details;
    userOrm.address = aggregate.address;
    return userOrm;
  }

  }

If you want to save/load this aggregate typeorm will take care of it automatically:

class UserRepository implements UserRepositoryPort {
  constructor(
      @InjectRepository(UserOrmEntity)
      private readonly userRepository: Repository<UserOrmEntity>
  )

    async save(userAggregate: UserAggregate): Promise<void> {
      const userOrm = userMapper.toOrmEntity(userAggregate);
      await this.userRepository.save(userOrm);
    }

    async load(id: string): Promise<UserAggregat> {
      const userOrmEntity = await this.userRepository.readOne({ id });
      return userMapper.toDomain(userOrmEntity);
    }

Your domain is using UserRepositoryPort so it's agnostic to ORM or a database technology. If you want to stop using ORM or change a database, you just create a new repository that satisfies a UserRepositoryPort interface and that's it.

Or maybe I'm misunderstanding what you are trying to achieve here.

[JoelVenable]

So the driving force for us going down this road was a frustration with all of the node ORMs including Mikro*, so I wrote a script that hits Postgres introspection and generates EntityRepositories and OrmEntities for all of the database tables backed by the pg driver directly. Currently each EntityRepository only hits a single table and doesn't have any joining capability.

Thanks for your feedback. It sounds like I need to do more work on the repository generator rather than shifting the complexity to the Domain.

*Mikro: I used to love it due to being able to hydrate Value Objects from raw primitives. It basically meant that we didn't need to have separate Domain Entities from Orm entities. However, we started seeing weird errors where it would try to update records that were fetched but not even touched during a unit of work, so ultimately we're having to bail out of it. It's probably something I did wrong, but we had no idea where to look to debug 🤷.