I'm using the SchemaExportTask to generate my schema from the mapping docs, and it isn't putting the correct "ON DELETE CASCADE" into the constraints. It's not putting anything for the cascade information. The cascade attribute is set to delete in appropriate locations in the mapping document.

Is this a limitation with the SchemaExport utility? Is there a workaround?

I'm running Hibernate 2.0.2 and PostgreSQL 7.3.4.

Thanks for any responses.

cascade is done in-memory, by the Hibernate session.

Thanks for the feedback. Wouldn't it be better to use the database's cascading constraints rather than managing it in memory?

nope. we need to know exactly what got deleted, in order to keep the session state synchronized with the database state.

I can appreciate that. The only reason I ask is that I'm using dbunit to load/update sample data into the database and it's complaining about FK violations, so it's not a Hibernate problem.

Maybe a tool like dbunit but using Hibernate to load sample data for testing? I'd be willing to build it if there are other people interested.

Currently I'm working together with Ara Abrahamian to implement something like this.

Ara is working on getting the XML databinding stuff finished, and I have implemented a Session.replicate() method.

So in future, you should be able to completely reproduce an object graph in the database, starting from an XML document.

I am using Hibernate-3 beta_1 and using SchemeExportTask. The generated schema doesnt have the "on delete cacade".

Just want to know if this is going to be fixed in Hibernate 3? We have a workaround but just wanted to see there is anything in the pipeline for Hibernate3.

Thanks.

Use <key on-delete="cascade"/>

Thanks Gavin.

The generated schema has the "on delete cascade" added for the foreign key constraint after I added the <key on-delete="cascade"/>.

But the problem is that I get the mapping exception [only inverse one-to-many associations may use on-delete="cascade"] Collection.java:215 when the session is being initialized.

Does the relationship has to have [inverse="true"] to get this working? Why is this restriction?

Thanks again.