Hibernate version: 3.2.0
Hibernate tools version: 3.2.0.beta8

In simplified terms I have an object, call it a Forest, that contains (one to many) other objects that are similar to each other, for example Deer and Bear which both extend from the generic class Animal. I am trying to add to my Forest mapping document two collection mapping sets: one for the Deer and one for the Bear inhabitants of the Forest. Here is the gist of the mapping files:



Note that the many-to-one parent reference is in the Animal class, not the derivant Deer and Bear classes.

The problem I'm having is that the hbm2ddl generation causes the Bear and Deer tables to contain columns for "forest_id" but Hibernate core does not expect that to be the case. Basically, my generated schema looks like this:



Which has the "forest_id" column in all three tables (instead of just the Animal table where it belongs). Hibernate core seems to have read my mind when interpreting the mapping declaration and does not include values for "forest_id" when inserting new rows into the Bear or Deer tables. This causes a NULL Constraint violation (since it does not supply values for the "not null" forest_id column) and things break.

So, am I doing something wrong or is this a question better suited for the Tools group?

I think your problem is in the definition of sets "deers" and "bears":

<set name="bears" inverse="true" lazy="true" cascade="all">
<key column="forest_id" not-null="true" />
<one-to-many class="Bear" />
</set>

if you say "<key column="forest_id"" hibernate will create such key in the Bear table.

You should define a "animals" set:

<set name="animals" inverse="true" lazy="true" cascade="all">
<key column="forest_id" not-null="true" />
<one-to-many class="Animal" />
</set>


To get the "deers" and the "bears" maybe you should use the "table per subclass with discriminator" and use the discriminator in a "select" attribute on your sets...

I understand that that would force hbm2ddl into submission at the expense of my having to sort through all the different sub-types but I'm trying to avoid having the "animals" set since that would give me all the chipmunks squirrles and other animals that I don't care about (that I omitted from my sample hbm above). I orignally tried to use the discriminator model but that wouldn't let me include sets inside the derived classes (e.g. the Bears class could not contain a set of Dens that it frequented) so I was forced into the joined-subclass model.

Given that hibernate core seems to be doing the right thing (it works fine if I manually edit the SQL schema) I'm inclined to believe that what I'm doing isn't way off base and that I either have a subtle error in my mapping files or I'm dealing with a bug in the hbm2ddl tool.

What do you mean? If you manually edit the SQL schema, dropping those forest_id, i guess hibernate will do the inserts OK. But, will it give you the Deers of a Forest when you call forest.getDeers()?

Yes, if I remove the forest_id from the Deer and Bear tables (and the corresponding foreign-key contraints). Everything works as expected. the Forest.getDeers() returns just the Deer and Forest.getBears() returns just the Bears. Cascading works just as well.

mmm... in my case:

Hibernate version 3.2 cr4
MySQL 5.0.27 InnoDB

And if i remove those keys manually, as you say, when i call forest.getDeers() i get:

JDBCExceptionReporter:63 - could not initialize a collection: [joinedsubclass.Forest.deers#3] [select deers0_.forest_id as forest2_1_, deers0_.animal_id as animal1_1_, deers0_.animal_id as id1_0_, deers0_1_.forest_id as forest2_1_0_ from Deer deers0_ inner join Animal deers0_1_ on deers0_.animal_id=deers0_1_.id where deers0_.forest_id=?]
java.sql.SQLException: Unknown column 'deers0_.forest_id' in 'field list'

What's your DB? and what SQL hibernate generates when you call forest.getDeers()?

Interesting. Now that I look closer, it was working for me since I was accessing things out of a cache. If i load them from the database, I get the exception as you describe (using PostgreSQL if it makes any difference). So it does populate the database "correctly" (at least as far as I'm concerned) but it does not query properly.

So I guess I'm stuck moving the many-to-one declarations into the joined-subclass definitions, repeating them for each one. Not pretty (and rather redundant) but it seems to work.

Thanks for the help!