What is the difference between DAO and Repository patterns?

What is the difference between Data Access Objects (DAO) and Repository patterns? I am developing an application using Enterprise Java Beans (EJB3), Hibernate ORM as infrastructure, and Domain-Driven Design (DDD) and Test-Driven Development (TDD) as design techniques.


DAO is an abstraction of data persistence.
Repository is an abstraction of a collection of objects.

DAO would be considered closer to the database, often table-centric.
Repository would be considered closer to the Domain, dealing only in Aggregate Roots.

Repository could be implemented using DAO's, but you wouldn't do the opposite.

Also, a Repository is generally a narrower interface. It should be simply a collection of objects, with a Get(id), Find(ISpecification), Add(Entity).

A method like Update is appropriate on a DAO, but not a Repository - when using a Repository, changes to entities would usually be tracked by separate UnitOfWork.

It does seem common to see implementations called a Repository that is really more of a DAO, and hence I think there is some confusion about the difference between them.


OK, think I can explain better what I've put in comments :). So, basically, you can see both those as the same, though DAO is a more flexible pattern than Repository. If you want to use both, you would use the Repository in your DAO-s. I'll explain each of them below:

REPOSITORY:

It's a repository of a specific type of objects - it allows you to search for a specific type of objects as well as store them. Usually it will ONLY handle one type of objects. E.g. AppleRepository would allow you to do AppleRepository.findAll(criteria) or AppleRepository.save(juicyApple). Note that the Repository is using Domain Model terms (not DB terms - nothing related to how data is persisted anywhere).

A repository will most likely store all data in the same table, whereas the pattern doesn't require that. The fact that it only handles one type of data though, makes it logically connected to one main table (if used for DB persistence).

DAO - data access object (in other words - object used to access data)

A DAO is a class that locates data for you (it is mostly a finder, but it's commonly used to also store the data). The pattern doesn't restrict you to store data of the same type, thus you can easily have a DAO that locates/stores related objects.

E.g. you can easily have UserDao that exposes methods like

Collection<Permission> findPermissionsForUser(String userId)
User findUser(String userId)
Collection<User> findUsersForPermission(Permission permission)

All those are related to User (and security) and can be specified under then same DAO. This is not the case for Repository.

Finally

Note that both patterns really mean the same (they store data and they abstract the access to it and they are both expressed closer to the domain model and hardly contain any DB reference), but the way they are used can be slightly different, DAO being a bit more flexible/generic, while Repository is a bit more specific and restrictive to a type only.