I have a <list> with a many-to-many association. Now, under Hibernate 2.1.8 when there's an element missing (meaning there's a gap in the index column) it would not insert an entry for the missing element. When selecting, it would insert a null element into the List.

Now with Hibernate 3, it wants to insert an entry where the back reference is null (which is not allowed).

Does someone have an idea why this behaviour is different in Hibernate 3 and how to go about making it behave as under Hibernate 2? Any help is appreciated.

If there is truly a change in behavior here, it would be a bug, and you should submit an ultra-simple, runnable test case to JIRA that reproduces this.

I'm not quite sure if there is a change. I think you should know better than I ;)

Okay, I've done some tests: Whenever the _last_ element in the list is null it wants to insert a NULL child reference. It does work with other list indices being null, however. Only the last index is problematic.

It looks like PersistentList.entryExists() returns true when the specified entry is non-null or the specified index is the last of the list (as has been mentioned at http://forum.hibernate.org/viewtopic.php?t=934884&highlight=entryexists, too).

Unfortunately, there is no mention as to why it would return true on the last index, but not on others. Exactly that is giving us problems here.

This is the source: http://opensource.atlassian.com/project ... se/HB-1249

This change makes Hibernate 3 behave differently from Hibernate 2. The problem is that the schema exporter is still creating a non-null column which is breaking our application.

I've filed a bug report at http://opensource.atlassian.com/projects/hibernate/browse/HHH-510

Looks like Gavin rejected my bug report with a (decidedly quite arrogant) response:



I'm not quite sure what to think of that. On the one hand, Hibernate allows to insert null elements into lists, and that is good. But on the other hand, you can't have a null element at the very last position because the schema generator doesn't allow for NULL values which Hibernate wants to insert.

What kind of weird sense does that make?

And aside from that, I'm more than open for a suggestion if our data model is that broken and is stinking to high heaven. Apparently Gavin does have some knowledge we don't have.

For future reference: insulting people who give you free things is simply out of order and not tolerated here.