list of objects returned by query partially correct

• Key: OPENJPA-313
• Type: Bug
• Status: Resolved
• Resolution: Fixed
• Priority: Major
• Assignee: Unassigned
• Reporter: Catalina Wei
• Component/s: jdbc
• Fix Version/s: 1.0.0

Description

We have an object inheritance hierarchy as follows.

    FixedAnnuity extends Annuity
    EquityAnnuity extends Annuity

Then we have following code:

    List<IAnnuity> annuities = getServerAdapter().findHolderAnnuities(holder);

from which we expect to have a list of "Annuity" (could be Annuity, FixedAnnuity or
EquityAnnuity) from following code
EntityManager em = null;
try{
em = factory.createEntityManager();
Query query = em.createNamedQuery("GetHolderAnnuities");
query.setParameter("holderId", holder.getId());
return (List<IAnnuity>) query.getResultList();
}

Here is the query
<named-query name="GetHolderAnnuities">
<query>SELECT a FROM Annuity AS a WHERE a.annuityHolderId = :holderId</query>
</named-query>

In the end, the list returned only the first object with the correct Object, the
rest of the objects all casted into the basic type "Annuity"
e.g. we have following code:

EntityManager em = null;
try{
em = factory.createEntityManager();
Query query = em.createNamedQuery("GetHolderAnnuities");
query.setParameter("holderId", holder.getId());
return (List<IAnnuity>) query.getResultList();
}

FixedAnnuity fixed1 = new FixedAnnuity();
fixed1.setHolder(holder);
EquityAnnuity equity1 = new EquityAnnuity();
equity1.setHolder(holder);
FixedAnnuity fixed1 = new FixedAnnuity();
fixed2.setHolder(holder);
EquityAnnuity equity2 = new EquityAnnuity();
equity2.setHolder(holder);

List<IAnnuity> annuities = getServerAdapter().findHolderAnnuities(holder);

Only annuities.get(0) returns the correct object as FixedAnnuity, the other 3
objects returned all returned as Annuity

Comments

[Catalina Wei]

The attached patch seemed to have fixed the problem.
The testcase of the Annuity has FetchType EAGER set on both side of a OneToMany relation. An Annuity may have many Payouts.
I'd like to get some feedback from commiters before checking in this patch.
TCK verification is in process.

[Catalina Wei]

same patch as the last patch, but using the new naming convention as recommended by the community.

[Kevin Sutter]

Catalina,
Thanks for the patch. A couple of questions and/or comments though...

It looks like this updated code is only if we're working with an OpenJPAID oid and subs is true. I can understand the oid part, but what exactly is the subs boolean used for? I know this wasn't your original code, but a quick glance at the code shows a need for a javadoc update at least. You must have spent some time figuring out how this parameter is used. It would be helpful to put add to the javadoc for this method to explain what the subs parameter is used for (at least). Thanks.

Also, I really dislike empty catch clauses. I can't tell if this was on purpose or an oversight. It also forces me to think through the logic whether this exception clause was necessary or not. I think the empty clause is okay, but we should at least put a comment in the clause indicating why we can safely ignore this exception.

I also personally don't like the idea of comparing strings for the strategy type, but it looks like we're just following suit from the existing code. We can only change so much at a time, I suppose... :-)

Thanks again,
Kevin

[Catalina Wei]

add block comments to the code.

[Catalina Wei]

attached revised patch.
mvn clean install ran successful.

[Kevin Sutter]

Since Catalina doesn't quite have the necessary karma yet, I will resolve this Issue for her... :-) Thanks, Catalina, for the patch!