In my specific case, I am making use of a discriminator column strategy. This means that my JPA implementation (Hibernate) creates a users table with a special DTYPE column. This column contains the class name of the entity. For example, my users table can have subclasses of TrialUser and PayingUser. These class names would be in the DTYPE column so that when the EntityManager loads the entity from the database, it knows which type of class to instantiate.
I've tried two ways of converting Entity types and both feel like dirty hacks:
- Use a native query to manually do an UPDATE on the column, changing its value. This works for entities whose property constraints are similar.
- Create a new entity of the target type, do a BeanUtils.copyProperties() call to move over the properties, save the new entity, then call a named query which manually replaces the new Id with the old Id so that all the foreign key constraints are maintained.
The problem with #1 is that when you manually change this column, JPA doesn't know how to refresh/reattach this Entity to the Persistance Context. It expects a TrialUser with Id 1234, not a PayingUser with Id 1234. It fails out. Here I could probably do an EntityManager.clear() and detach all Entities/clear the Per. Context, but since this is a Service bean, it would wipe pending changes for all users of the system.
The problem with #2 is that when you delete the TrialUser all of the properties you have set to Cascade=ALL will be deleted as well. This is bad because you're only trying to swap in a different User, not delete all the extended object graph.
Update 1: The problems of #2 have made it all but unusable for me, so I've given up on trying to get it to work. The more elegant of the hacks is definitely #1, and I have ma