I had an interesting conversation at Jazoon with a fellow who has a
simple problem that he can't seem to get to work in OpenJPA without
using implementation specific APIs.
He has a Department and Employees in a OneToMany and ManyToOne
relationship.
He has the Employee.department field lazily loaded because it's just
not needed in most application usages.
The application code wants to get the Department.id after querying
Employees.
The simple employee.getDepartment().getId() doesn't need to go to the
database, since the department id was loaded from the database when
the employee was fetched. And yet, OpenJPA goes to the database in
order to instantiate all the values in the department object, even
though all the data that the application needs (id) is already in
memory.
This case was one of the big reasons we implemented the hollow state
of the object, so the id would always be available. The
employee.getDepartment() should simply create a hollow Department
based on the id and await further action by the user. If all the user
needs is the id, it's available. If any other field is referenced,
then OpenJPA should go to the database and fetch the load fetch group
of the field that's fetched. If OpenJPA eagerly loads the department
based on the getId method, it defeats the purpose of the load fetch
group for the field that's accessed (after the id field).
I haven't checked to see whether this behavior is simply a fantasy of
mine or whether the behavior changed over time.
Craig
Craig L Russell
Architect, Oracle
http://db.apache.org/jdo
408 276-5638 mailto:[email protected]
P.S. A good JDO? O, Gasp!
