My database has two entities; Company and Person. A Company can have many People, but a Person must have only one Company. The table structure looks as follows.
COMPANY
----------
owner PK
comp_id PK
c_name
PERSON
----------------
owner PK, FK1
personid PK
comp_id FK1
p_fname
p_sname
It has occurred to me that I could remove PERSON.OWNER and derive it through the foreign key; however, I can't make this change without affecting legacy code.
I have modeled these as JPA-annotated classes;
@Entity
@Table(name = "PERSON")
@IdClass(PersonPK.class)
public class Person
implements Serializable {
@Id
private String owner;
@Id
private String personid;
@ManyToOne
@JoinColumns(
{@JoinColumn(name = "owner", referencedColumnName = "OWNER",
insertable = false, updatable = false),
@JoinColumn(name = "comp_id", referencedColumnName = "COMP_ID",
insertable = true, updatable = true)})
private Company company;
private String p_fname;
private String p_sname;
...and standard getters/setters...
}
@Entity
@Table(name = "COMPANY")
@IdClass(CompanyPK.class)
public class Company
implements Serializable {
@Id
private String owner;
@Id
private String comp_id;
private String c_name;
@OneToMany(mappedBy = "company", cascade=CascadeType.ALL)
private List people;
...and standard getters/setters...
}
My PersonPK and CompanyPK classes are nothing special, they just serve as a struct holding owner and the ID field, and override hashCode and equals(o).
So far so good. I come across a problem, however, when trying to deal with associations. It seems if I have an existing Company, and create a Person, and associate to the Person to the Company and persist the company, the association is not saved when the Person is inserted. For example, my main code looks like this:
EntityManager em = emf.createEntityManager();
em.getTransaction().begin();
CompanyPK companyPK = new CompanyPK();
companyPK.owner="USA";
companyPK.comp_id="1102F3";
Company company = em.find(Company.class, companyPK);
Person person = new Person();
person.setOwner("USA");
person.setPersonid("5116628123"); //some number that doesn't exist yet
person.setP_fname("Hannah");
person.setP_sname("Montana");
person.setCompany(company);
em.persist(person);
This completes without error; however in the database I find that the Person record was inserted with a null in the COMP_ID field. With EclipseLink debug logging set to FINE, the SQL query is shown as:
INSERT INTO PERSON (PERSONID,OWNER,P_SNAME,P_FNAME) VALUES (?,?,?,?)
bind = [5116628123,USA,Montana,Hannah,]
I would have expected this to be saved, and the query to be equivalent to
INSERT INTO PERSON (PERSONID,OWNER,COMP_ID,P_SNAME,P_FNAME) VALUES (?,?,?,?,?)
bind = [5116628123,USA,1102F3,Montana,Hannah,]
What gives? Is it incorrect to say updatable/insertable=true for one half of a composite key and =false for the other half? If I have updatable/insertable=true for both parts of the foreign key, then Eclipselink fails to startup saying that I can not use the column twice without having one set to readonly by specifying these options.
