Java / Hibernate / JPA：无法使用复合键持久化->临时对象

Question

my problem is that I cannot save my entity because it contains another entity, mapped by a key that is also a part of this table's primary key. The table looks like this:

table C:

+-----+------+
| id_A | id_B |
+-----+------+

..where idA is the primary key of table A with EntityA and idB the primary key of table B with EntityB .

so its basically a n-to-m relation. This is the entity I'm using for table C:

@Entity
public class EntityC {

    private long idA;
    private EntityB b;

    @Id
    @Column(name = "id_A")
    public long getIdA() {
        return idA;
    }

    @Id
    @OneToOne(cascade = CascadeType.ALL)
    @JoinColumn(name = "id_B")
    public EntityB getB() {
        return b;
    }

    ...setters are here...

}

Please note that id_A is mapped as is (the id), while id_B is mapped as its object representation, EntityB . This is what I want to do with it:

EntityC c = new EntityC();
c.setIdA(123);
c.setB(new EntityB());

em.persist(c);
tx.commit();
em.close();

I want to persist EntityB ONLY IF I can persist EntityC .

on tx.commit() I get this exception: org.hibernate.TransientObjectException: object references an unsaved transient instance

I suppose this happens because part of the primary key, id_B , is not saved. But i set cascading to all so there should be no problem!

Why is this not working?

---

EDIT:

When I do this:

em.persist(c.getB());
em.persist(c);

it works. But can't Hibernate/JPA do that automatically? I thought that's what cascading is good for.

---

EDIT2:

added an embeddedId instead of id_A and id_B:

@Embeddable
public class EntityCID implements Serializable {

public long idA;

@OneToOne(cascade = CascadeType.ALL)
@JoinColumn(name = "id_B", referencedColumnName = "id")
public EntryB b;

}

EntityC now looks like:

@Entity
public class EntityC implements Serializable {

    private EntityCID id;
    ...

    @EmbeddedId
    public void getId() {
        return id;
    }

}

but I still get the transient object exception if I don't em.persist(c.getId().b); before em.persist(c) . Sticking to that, although it is ugly.

@Trein: it is not bidirectional. EntityB code:

@Entity
public class EntityB implements Serializable {
    public long id;
    public String text;
}

Answer 1

Is it a bidirectional relationship? I would suggest you to remove @Id getB() and perform the modifications:

@OneToOne(cascade = CascadeType.ALL, mappedBy = "id_B")
@PrimaryKeyJoinColumn(name = "id_B")
public EntityB getB() {
    return b;
}

Your entity class must have only one attribute annotated with @Id . Usually when you need this, you create a class that will store both properties and this will act as a Id Class.

Answer 2

You can not pass new Entity() for reference. Because it won't have any values in it(even primary key). So how can hibernate will insert it as foreign key to the table. And cascade will save your parent object if its not saved,no need to call save method for all. But when you passing new object it won't do.

Answer 3

If you think about it what you are seeing makes perfect sense.

EntityC is is the 'owning side' of the relationship C<>B: it defines the JoinColumn and EntityB has the 'mappedBy' attribute.

So on saving C, order of events would normally be:

• insert into C/update C
• insert into B/update B

Now in your case this causes issues as obviously C can only be saved if B has been persisted first.

In terms of your statement above: I want to persist "EntityB ONLY IF I can persist EntityC." How can this ever be the case?

JPA has a concept of 'Derived Identifiers', which I am not overly familiar with however is defined in the book Pro JPA as occurring when:

When an identifier in one entity includes a foreign key to another entity, we call it a derived identifier. Because the entity containing the derived identifier depends upon another entity for its identity, we call the first the dependent entity. The entity that it depends upon is the target of a many-to-one or one-toone relationship from the dependent entity, and is called the parent entity

Now, despite the original advice that you had two @Id attributes defined and this was wrong it would however appear that having an additional @Id on a 1-2-m is in fact valid in JPA 2 for precisely this case.

The book gives a number of ways of dealing with Derived Identifiers however one example given below looks fairly similar to your case. So you may want to investigate further the @MapsId attribute.

@Entity
public class Project {

@EmbeddedId private ProjectId id;
@MapsId("dept")
@ManyToOne
@JoinColumns({
@JoinColumn(name="DEPT_NUM", referencedColumnName="NUM"),
@JoinColumn(name="DEPT_CTRY", referencedColumnName="CTRY")})
private Department department;
// ...
}

@Embeddable
public class ProjectId implements Serializable {

@Column(name="P_NAME")
private String name;
@Embedded
private DeptId dept;
// ...
}

See further:

How do I properly cascade save a one-to-one, bidirectional relationship on primary key in Hibernate 3.6