Extra SQL on LAZY/EAGER  ManyToOne relation
-------------------------------------------

                 Key: OPENJPA-744
                 URL: https://issues.apache.org/jira/browse/OPENJPA-744
             Project: OpenJPA
          Issue Type: Sub-task
          Components: jdbc
    Affects Versions: 1.2.1, 1.3.0
            Reporter: Catalina Wei
            Assignee: Catalina Wei


OPENJPA-241 has eliminated some extra SQLs for  2 entities  having a LAZY 
OneToMany,  and inverse EAGER ManyToOne relations.
There are still  extra SQLs that can be avoided in the cases where  2 entities 
are related by  an EAGER OneToMany,  and inverse LAZY ManyToOne relations.

Consider the following 2 entities:

    Publisher  (1) <--->  (M)  Magazine

    @Entity
public class Publisher implements Serializable {

    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    @Column(name="id")
    private int id;

    @OneToMany(mappedBy="idPublisher",  fetch=FetchType.EAGER)
    private Set<Magazine> magazineCollection;   
...
}

@Entity
public class Magazine implements Serializable {

    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    @Column(name="id")
    private int id;

    @ManyToOne(fetch=FetchType.LAZY)
    @JoinColumn(name="id_publisher")
    private Publisher idPublisher;

...
}


Query = em.createQuery("SELECT p from Publisher p);

Since Publisher has a OneToMany Eager relation (magazines), We generate 
following  SQLs 

 (1)  SELECT t0.id, t0.name FROM Publisher t0
 (2)  SELECT t0.id, t1.id, t1.date_published, t1.id_publisher, t1.name FROM 
Publisher t0 INNER JOIN Magazine t1 ON t0.id = t1.id_publisher ORDER BY t0.id 
ASC

However, the following extra SQLs are also generated (making additional 
database trips)
 (3) SELECT t0.name, t1.id_publisher, t1.id, t1.date_published, t1.name FROM 
Publisher t0 LEFT OUTER JOIN Magazine t1 ON t0.id = t1.id_publisher WHERE t0.id 
= ? [params=(int) 2]
... more depending on how many publishers there are in the database.

The fact that SQL(2) has  already returned all magazines of all publishers, SQL 
(3) is unnecessary. 
The inverse ManyToOne relation should  be established from the SQL (2) result.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to