How do I use annotations to define different types of relationships in Hibernate 4 and Spring?

Solution 1:

Creating Relationships with Annotations

Assume all classes annotated with @Entity and @Table

Uni-directional One to One Relationship

public class Foo{
    private UUID fooId;

    @OneToOne
    private Bar bar;
}

public class Bar{
    private UUID barId;
    //No corresponding mapping to Foo.class
}

Bi-Directional One to One Relationship managed by Foo.class

public class Foo{
    private UUID fooId;

    @OneToOne(cascade = CascadeType.ALL)
    @JoinColumn(name = "barId")
    private Bar bar;
}

public class Bar{
    private UUID barId;

    @OneToOne(mappedBy = "bar")
    private Foo foo;
}

Uni-Directional One to Many Relationship using user managed join table

public class Foo{
    private UUID fooId;

    @OneToMany
    @JoinTable(name="FOO_BAR",
        joinColumns = @JoinColumn(name="fooId"),
        inverseJoinColumns = @JoinColumn(name="barId"))
    private List<Bar> bars;
}

public class Bar{
    private UUID barId;

    //No Mapping specified here.
}

@Entity
@Table(name="FOO_BAR")
public class FooBar{
    private UUID fooBarId;

    @ManyToOne
    @JoinColumn(name = "fooId")
    private Foo foo;

    @ManyToOne
    @JoinColumn(name = "barId")
    private Bar bar;

    //You can store other objects/fields on this table here.
}

Very commonly used with Spring Security when setting up a User object who has a list of Role's that they can perform. You can add and remove roles to a user without having to worry about cascades deleting Role's.

Bi-directional One to Many Relationship using foreign key mapping

public class Foo{
    private UUID fooId;

    @OneToMany(mappedBy = "bar")
    private List<Bar> bars;
}

public class Bar{
    private UUID barId;

    @ManyToOne
    @JoinColumn(name = "fooId")
    private Foo foo;
}

Bi-Directional Many to Many using Hibernate managed join table

public class Foo{
    private UUID fooId;

    @OneToMany
    @JoinTable(name="FOO_BAR",
        joinColumns = @JoinColumn(name="fooId"),
        inverseJoinColumns = @JoinColumn(name="barId"))
    private List<Bar> bars;
}

public class Bar{
    private UUID barId;

    @OneToMany
    @JoinTable(name="FOO_BAR",
        joinColumns = @JoinColumn(name="barId"),
        inverseJoinColumns = @JoinColumn(name="fooId"))
    private List<Foo> foos;
}

Bi-Directional Many to Many using user managed join table object

Commonly used when you want to store extra information on the join object such as the date the relationship was created.

public class Foo{
    private UUID fooId;

    @OneToMany(mappedBy = "bar")
    private List<FooBar> bars;
}

public class Bar{
    private UUID barId;

    @OneToMany(mappedBy = "foo")
    private List<FooBar> foos;
}

@Entity
@Table(name="FOO_BAR")
public class FooBar{
    private UUID fooBarId;

    @ManyToOne
    @JoinColumn(name = "fooId")
    private Foo foo;

    @ManyToOne
    @JoinColumn(name = "barId")
    private Bar bar;

    //You can store other objects/fields on this table here.
}

Determining which side of the bi-directional relationship 'owns' the relationship:

This is one of the trickier aspects of working out Hibernate relationships because Hibernate will operate correctly no matter which way to set up the relationship. The only thing that will change is which table the foreign key is stored on. Generally the object that you have a collection of will own the relationship.

Example: A User object has a list of Roles declared on it. In most applications, the system will be manipulating instances of the User object more often than instances of the Roles object. Hence I would make the Role object the owning side of the relationship and manipulate the Role objects through the list of Role's on a User by cascade. For a practical example see the bi-directional One to Many example. Typically you will cascade all changes in this scenario unless you have a specific requirement to do otherwise.

Determining your fetchType

Lazily fetched collections have resulted in more issues on SO than I care to look at because by default Hibernate will load related objects lazily. It doesn't matter if the relationship is a one-to-one or many-to-many as per the Hibernate docs:

By default, Hibernate uses lazy select fetching for collections and lazy proxy fetching for single-valued associations. These defaults make sense for most associations in the majority of applications.

Consider this my two cents on when to use fetchType.LAZY vs fetchType.EAGER on your objects. If you know that 50% of the time you won't need to access the collection on your parent object, I'd be using fetchType.LAZY.

The performance benefits are of this are huge and only grow as you add more objects to your collection. This is because for an eagerly loaded collection, Hibernate does a ton of behind the scenes checking to ensure that none of your data is out of date. While I do advocate using Hibernate for collections, be aware that there is a performance penalty** for using fetchType.EAGER. However, take our Person object example. Its fairly likely that when we load a Person we will want to know what Roles they perform. I will usually mark this collection as fetchType.EAGER. DON'T REFLEXIVELY MARK YOUR COLLECTION AS fetchType.EAGER SIMPLY TO GET AROUND A LazyInitializationException. Not only is it bad for performance reasons, it generally indicates that you have a design issue. Ask yourself, should this collection actually be an eagerly loaded collection, or am I doing this just to access the collection in this one method. Hibernate has ways around this, that doesn't impact the performance of your operations quite as much. You can use the following code in your Service layer if you want to initialize a lazily loaded collection just for this one call.

//Service Class
@Override
@Transactional
public Person getPersonWithRoles(UUID personId){
    Person person = personDAO.find(personId);

    Hibernate.initialize(person.getRoles());

    return person;
}

The call to Hibernate.initialize forces the creation and loading of the collection object. However, be careful, if you only pass it the Person instance, you will get a proxy of your Person back. See the documentation for more information. The only downside to this method is that you have no control over how Hibernate will actually fetch your collection of objects. If you want to control this, then you can do so in your DAO.

//DAO
@Override
public Person findPersonWithRoles(UUID personId){
    Criteria criteria = sessionFactory.getCurrentSession().createCritiera(Person.class);

    criteria.add(Restrictions.idEq(personId);
    criteria.setFetchMode("roles", FetchMode.SUBSELECT);
}

The performance here depends on what FetchMode you specify. I've read answers that say to use FetchMode.SUBSELECT for performance reasons. The linked answer goes into more detail if you are really interested.

If you want to read me as I repeat myself, feel free to check out my other answer here

Determining Cascade Direction

Hibernate can cascade operations either or both ways in a bi-directional relationship. So if you have a List of Role's on a User you can cascade changes to Role's in both directions. If you change the name of a particular Role on a User Hibernate can automatically update the associated Role on the Role Table.

However this is not always desired behaviour. If you think about it, in this case, making changes to Role's based on changes to User doesn't make any sense. However it makes sense going in the opposite direction. Change a Role's name on the Role object itself, and that change can be cascaded to all User objects that have that Role on it.

In terms of efficiency, it makes sense to create/update Role objects by saving the User object that they belong to. This means you would mark your @OneToMany annotation as the cascading one. I'll give an example:

public User saveOrUpdate(User user){
    getCurrentSession.saveOrUpdate(user);
    return user;
}

In the above example, Hibernate will generate a INSERT query for the User object, and then cascade the creation of the Role's once the User has been inserted into the database. These insert statements will then be able to use the PK of the User as their foreign key, so you would end up with N + 1 insert statements, where N is the number of Role objects in the list of users.

Conversely if you wanted to save the individual Role objects cascading back to the User object, could be done:

//Assume that user has no roles in the list, but has been saved to the
//database at a cost of 1 insert.
public void saveOrUpdateRoles(User user, List<Roles> listOfRoles){
    for(Role role : listOfRoles){
        role.setUser(user);
        getCurrentSession.saveOrUpdate(role);
    }
}

This results in N + 1 inserts where N is the number of Role's in the listOfRoles, but also N update statements being generated as Hibernate cascades the addition of each Role to the User table. This DAO method has a higher time complexity than our previous method, O(n) as opposed to O(1) because you have to iterate through the list of roles. Avoid this if at all possible.

In practice however, usually the owning side of the relationship will be where you mark your cascades, and you will usually cascade everything.

Orphan Removal

Hibernate can work out for you if you remove all associations to an object. Suppose you have a User who has a list of Role's and in this list are links to 5 different roles. Lets say you remove a Role called ROLE_EXAMPLE and it happens that the ROLE_EXAMPLE doesn't exist on any other User object. If you have orphanRemoval = true set on the @OneToMany annotation, Hibernate will delete the now 'orphaned' Role object from the database by cascade.

Orphan removal should not be enabled in every case. In fact, having orphanRemoval in our example above makes no sense. Just because no User can perform whatever action the ROLE_EXAMPLE object is representing, that doesn't mean that any future User will never be able to perform the action.

This Q&A is intended to complement the official Hibernate documentation, which has a large amount of XML configuration for these relationships.

These examples are not meant to be copy-pasted into production code. They are generic examples of how to create and manage various objects and their relationships using JPA annotations to configure Hibernate 4 within the Spring Framework. The examples assume that all classes have ID fields declared in the following format: fooId. The type of this ID field is not relevant.

** We recently had to abandon using Hibernate for an insert job where we were inserting <80,000+ objects into the database through a collection. Hibernate ate up all the heap memory doing checking on the collection and crashed the system.

DISCLAIMER: I DO NOT KNOW IF THESE EXAMPLES WILL WORK WITH STANDALONE HIBERNATE

I am not in any way affiliated with Hibernate or the Hibernate dev team. I'm providing these examples so I have a reference to point to when I'm answering questions on the Hibernate tag. These examples and discussions are my based on my own opinion as well as how I develop my applications using Hibernate. These examples are in no way comprehensive. I'm basing them on the common situations I've used Hibernate for in the past.

If you encounter issues trying to implement these examples, do not comment and expect me to fix your problem. Part of learning Hibernate is learning the in's and out 's of its API. If there is a mistake with the examples, please feel free to edit them.