Hello, I would not ask, this seems to be a common issue, but I have been struggling with this for a week and feel that I am following the examples reasonably well. I am using Pointbase 4.8 and Hibernate 2.1.

I have a parent Person.hbm.xml that specifies

<set name="contacts" inverse="true" lazy="true" cascade="all">
<key column="PERSON_ID"/>
<one-to-many class="Contact"/>
</set>

The parent's child Contact.hbm.xml specifies

<many-to-one name="person" class="Person" column="PERSON_ID" not-null="true"/>

The test code is

session = sessionFactory.openSession();
transaction = session.beginTransaction();
person1 = new Person("100", ... new HashSet()); PK is "100"
session.save(person1);
contact1 = new Contact("I", ... person1); // PK is "I"
contact2 = new Contact("B", person1); // PK is "B"
person1.getContacts().add(contact1);
person1.getContacts().add(contact2);
session.flush();
transaction.commit(); // Commit the inserts.
closeSession(session);

I have tried many variations of this code (e.g. session.save(person1) instead of session.flush()) and basically at session.flush() I get

16:49:18,282 ERROR SessionImpl:2368 - Could not synchronize database state with session
net.sf.hibernate.HibernateException: SQL insert, update or delete failed (row not found)
at net.sf.hibernate.impl.NonBatchingBatcher.addToBatch(NonBatchingBatcher.java:25)
at net.sf.hibernate.persister.EntityPersister.update(EntityPersister.java:687)...

Is there an answer to this? Thank you, David

Could you please post your full mappings for your Person and Contact classes - I think this is a problem with not setting the unsaved-value on the ID property on one of them.

As a seconday point, is there any reason you do the save before adding the contacts to the person? You also don't need to flush the session, as committing a transaction does that automatically.

Dear sdknott,

Your simple and clear explanation solved my problem. Contrary to the Hibernate/general DB design recommendations (and as I hinted in my post) I am using keys with business value. Now, in each of my two hbm.xml files I set unsaved-value="any" (the attribute defaulted to null, but I was instantiating the children with id values.) I realize that the "unsaved-value" attribute determines Hibernate's handling of objects between sessions, but I'll have to keep testing to see what the ramifications are.

Per your suggestion I eliminated session.flush() and moved the save of the parent object to just before transaction.commit(). The children now save automatically.

Thank you once again for taking the time to answer my post. You have started my week on a good note!

I have the same problem. Though using unsaved-value="any" you will be able to insert it successfully. But it fails when you do updates.

Please let me know if update works fine.

Thanks,
Jeelani

I am struggling with the very thing as we speak. As you said, updates on the child are not working. From the log output I see that Hibernate is trying to update the child with a statement like

update CONTACT set CONTACT_NAME=?, ...PERSON_ID=? where CONTACT_TYPE=?

The "where" clause specifies only the child key, but does not include the parent key. This I think is a problem with using the documented one-to-many technique, with keys having business meaning, since I do not believe that sequential keys would cause a problem for such an update statement. In my case, CONTACT_TYPE is not unique among children with different parents.

I also created a composite-id version of the scenario I am trying to model, and briefly went back to that, to see if Hibernate will manage the associations. So far I have not found a formula that works.