Hibernate version:

3.2.1

I have two tables:

PARENT
--------------------------
PARENT_ID (PK)
.
. other fields
.

CHILD
-------------------------
PARENT_ID (FK)(PK)
INDEX (PK)

So child has a composite ID, a FK to its parent and and an integer for ordering and uniquely identifying the children.

Mapping documents:

Now the mapping for those two tables:


The Problem

We have a object graph of a parent + children.

We are using merge (on the parent, which cascades to the children) to associate the objects in the graph with the session.

The trouble is this: If the parent and children are both transient (new), then the parent's ID
has not yet been generated. When we attempt to merge the parent and its children into the
session, Hibernate generates a null pointer exception when merging the children because their
parent's ID is null and hasn't yet been generated by hibernate. We can work around this by
using explicitly assigning the parent's ID but we expect that this case should work with hibernate
generating the ID for the parent.


We have opened a bug with an example of the problem

http://opensource.atlassian.com/project ... e/HHH-2326

Anybody have a legitimate workaround that does not require the explicit assignment of the parent ID before merge?

Thanks.

Tyler Van Gorder
tkv@landacorp.com
Landacorp.

How about using save() instead of merge() when you know the object is transient.
If you're not sure whether it's transient use saveOrUpdate() instead.

Then cascade should include "save,update"
This works consistently for me.

Thanks for the response.

The problem we have with your approach is that our objects have been detached from the session and may or may not have been modified. We do not want to update the record or child record if nothing has been changed. It is my understanding that SaveOrUpdate will update the record (increment the version) every time it is called. Of course, I am new to hibernate, so I may be off base.

My understanding of versioning is that the version is only changed if the object is dirty during the flush operation. I have not tested this myself so you could be correct, but I'm sure it's woth a try.