Hello!

This is (in standard and programmatic Hibernate), what I need:


But when using the Persistence API from EJB3, I don't see the cfg object. So, I like to pass the name of my strategy class as a property in /META-INF/persistence.xml ... but I can't find the string constant pointing to a strategy class name. At least the Environment class doesn't provide one, does it?

Something like:



Or do I have to roll my own service provider? Nah... then I'll file a feature request against HibernatePersistence - it should provide a protected method like customize(AnnotatedConfiguration cfg) which can be overriden by subclasses. Line 368 in revision 1.25 is a good insertion point. Just before the fixed EJB3 properties are set the session factory is created... :)

Cheers,
Christian

Open a JIRA request

http://opensource.atlassian.com/projects/hibernate/browse/EJB-35

Done.

Hoped to see this issue solved in Hibernate EntityManager Changelog from 3.1beta3 (14-09-2005) ... maybe beta4? :)

Cheers,
Christian

I do have some conceptual issues that prevented me to close it.

Emmanuel,

Has there been any decisions made about this naming stratgey issue? I'm upgrading our product to start using entityManager beta 3 (w/ annotations beta 5) to be EJB3.0 compliant but none of the EJBQLs will work because of the mismatches in all the column names (we have our own NamingStrategy implementation).

Ejb3Configuration class also doesn't allow me to get to the AnnotationConfiguration instance so there is no way for me to do a temporary change for now. Are there any temporary workaround solutions that I can use ?

Thanks!
Aye

No sorry.
It's pretty trivial to add a setNAmingStrategy() to the Ejb3Configuration and delegate the call to the underlying annotationconfiguration.
But until I've solve the conceptual issues, I'll not add it to the release.

I would happily live with the constraint that columnName(propertyToCoreColumnName()) == propertyToColumnName() if it meant I could use propertyToColumnName() with my Hibernate Annotations beans.

All the columnName() and tableName() implementations I've written just return their inputs anyway, so that constraint is already always true for me; my feeling is if you're going to bother specifying a table or column name explicitly in an annotation or a mapping document, you may as well specify the actual name from the database to make it clear to a reader what you mean. So the change favored by Emmanuel in the mailing list message would be zero loss to me. My hunch is that columnName() and tableName() are probably not that widely used, but I'm sure if I'm wrong I'll get plenty of replies telling me so.

The "no per-property annotations needed by default" feature of EJB3 is basically nonexistent for me at the moment -- I have to have @Column(name="xxx") on the majority of my properties because I can't define a naming strategy. In that respect Hibernate Annotations is a step backwards for me, since I make extensive use of naming strategies in Hibernate 3.0.

I'm currently working on the naming strategy support, but it's a real PITA. It actually requires the notion of logical naming and physical naming.

I'm glad to hear it's being worked on. By the way, my "step backwards" comment was a bit harsh, so I apologize for that. Hibernate Annotations is an impressive piece of work -- the naming strategy problem is the only thing I've found wrong with it. Even with the @Column annotations it's a lot cleaner than all the XML files I used to have to maintain. It will be downright amazing once this problem is out of the way: for leaf-node classes, just one @Entity annotation at the top of the file and I'm done. You guys are doing a great job.

Check out Hibernate3 and Hibernate annotations from CVS, the naming strategy is implemented

did you or do you plan to add hibernate.naming_strategy support too ?

That's a different beast.

it was the original question ;)

Damn it! So I guess I'll have to support that ;-)