Composite Primary Key Selection- What works with Hibernate doesnt work with EclipseLink. Why?

问题

I have two tables.

CREATE TABLE profile_table
( 
    profile_id NUMBER(10,0) PRIMARY KEY, 
    creator_manager_id NUMBER(10,0), 
    lastupdate DATE, 
    description VARCHAR2(255 BYTE), 
    profile_name VARCHAR2(255 BYTE), 
    creatorname VARCHAR2(255 BYTE)
);

CREATE TABLE profile_basket_table
( 
    profile_id NUMBER(10,0), 
    from_id  NUMBER(10,0), 
    action NUMBER(10,0),
    constraint pbt_pk_constraint PRIMARY KEY(profile_id, from_id),
    constraint pbt_fk_constraint FOREIGN KEY(profile_id) REFERENCES profile_table(profile_id)
);

I have 1 Profile and * ProfileItem.

1 Profile --- * ProfileItems ( One profile can have Many ProfileItems).

Note : 1. Its a one-to-many relationship from profile to profileItem. 2. Its a unidirectional relationship.

PK for Profile is profile_id.

PK for ProfileItem is a composite primary key ( profile_id, from_id )

Firstly, I would I like to complain that EclipseLink doesnt work like Hibernate :

1st Problem with EclipseLink: Just becasue I have a composite primary key in child entity called ProfileItem, its forcing me to use JoinColumns. I cant use @JoinColumns as other column from_id is not a foreign key to any other table. I initially used same annotation which worked perfectly with hibernate as follows :

@Entity
@Table(name="profile_table")
class Profile {

    @Id
    @Column(name="profile_id")
    private Integer profileId ;
     ...

    /**
     *  WORKS WITH HIBERNATE AS PERSISTENCE PROVIDER, BUT FAILS WITH ECLIPSELINK
    */
    @OneToMany( cascade = CascadeType.ALL )
    @JoinColumn( name= "profile_id", referencedColumnName="profile_id" )
    private Set<XVMUpdateProfileItem> profileItems = null;

    ...
}

Above code throws following error with eclipselink :

The @JoinColumns on the annotated element [field profileItems] from the entity class [class arun.ucerelay.datastructures.XVMUpdateProfile] is incomplete. When the source entity class uses a composite primary key, a @JoinColumn must be specified for each join column using the @JoinColumns. Both the name and the referencedColumnName elements must be specified in each such @JoinColumn.

Some how with several permutations and combinations of @JoinColumn annotation, I managed to get this working for loading profiles and related profile items.

/**
 *  WORKS WITH ECLIPSELINK AS PERSISTENCE PROVIDER
 */

@OneToMany( cascade = CascadeType.ALL)
@JoinColumns({
    @JoinColumn( name= "profile_id", referencedColumnName="profile_id" , insertable = true, updatable = true),
    @JoinColumn( name= "profile_id", referencedColumnName="profile_id" , nullable = false)    
})
private Set<XVMUpdateProfileItem> profileItems = null;

This doesnt make sense to declare @JoinColumn again...But weirdly it works with eclipselink. This is a bug. Isnt it?

Can anyone tell me if I am missing anything ? Is there any other right way of implementing unidirectional relationship, please tell me.

This is a bug about @JoinColumns isnt it ? Should I log it ?

Is this called "portability" from one persistence provider to another ??????? What is the use of using JPA specification if it doesnt solve portability.

There is a second problem with saving of parent and cascading child items which works with hibernate and doesnt work with eclipselink. Ill post it as a separate question to keep this short and precise.

Thanks. Arun Kandregula

回答1:

This seems to be a bug in the annotation processing. Please log this bug on EclipseLink if it still occurs on the latest version.

Note that @JoinColumn is normally used for @ManyToOne or @OneToOne, not @OneToMany. Normally for a @OneToMany you should have an inverse @ManyToOne and use the mappedBy. Using a @JoinColumn on a @OneToMany is a special type of @OneToMany that maintains the target table's foreign key itself. Since the profile_id is part of your primary key, to be correct you really should have the @ManyToOne back, you could then remove the profileId and just add @Id to the @ManyToOne.