We have a rather complex mapping-job, where the classes >person<, >department< and >firm<, are extending the class >contact<.
A person belongs to a firm or a department.
A department belongs to a firm or a parent-department.
and a firm stands for its own.

What i tried to do first is to do a mapping through joined subclasses which worked fine.
In our second approach we tried to use the (recommended) subclass-strategy which permanently threw
WrongClassExceptions when loading the object-tree of firms.
We used this mapping (well, a bit stripped down):

------------
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE hibernate-mapping PUBLIC "-//Hibernate/Hibernate Mapping DTD 2.0//EN"
"http://hibernate.sourceforge.net/hibernate-mapping-2.0.dtd">
<hibernate-mapping>
<class name="Contact" table="T_CONTACT" >
<id column="UNIQUE_ID" name="uniqueId" type="long">
<generator class="sequence">
<param name="sequence">SEQ_UNID</param>
</generator>
</id>
<discriminator column="DISCRIMINATOR" type="string" />
<property name="name" column="NAME" type="string" />
<subclass name="Person" discriminator-value="Person" >
<property name="firstName" column="FIRSTNAME" type="string" />
<many-to-one name="parentFirm" column="FIRM_ID" />
<many-to-one name="parentDepartment" column="DEPARTMENT_ID" />
</subclass>
<subclass name="Firm" discriminator-value="Firm" >
<set name="employees" >
<key column="FIRM_ID"/>
<one-to-many class="Person"/>
</set>
<set name="subDepartments" >
<key column="FIRM_ID"/>
<one-to-many class="Department"/>
</set>
</subclass>
<subclass name="Department" discriminator-value="Department" >
<many-to-one name="parentDepartment" column="DEPARTMENT_ID" />
<many-to-one name="parentFirm" column="FIRM_ID" />
<set name="subDepartments" >
<key column="DEPARTMENT_ID"/>
<one-to-many class="Department"/>
</set>
<set name="employees" >
<key column="DEPARTMENT_ID"/>
<one-to-many class="Person"/>
</set>
</subclass>
</class>
</hibernate-mapping>
--------------
I finally found a solution in not reusing the foreign-keys between the used
classes and now use the following mapping without getting WrongClassExceptions.
---------------
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE hibernate-mapping PUBLIC "-//Hibernate/Hibernate Mapping DTD 2.0//EN"
"http://hibernate.sourceforge.net/hibernate-mapping-2.0.dtd">
<hibernate-mapping>
<class name="Contact" table="T_CONTACT" >
<id column="UNIQUE_ID" name="uniqueId" type="long">
<generator class="sequence">
<param name="sequence">SEQ_UNID</param>
</generator>
</id>
<discriminator column="DISCRIMINATOR" type="string" />
<property name="name" column="NAME" type="string" />
<subclass name="Person" discriminator-value="Person" >
<property name="firstName" column="FIRSTNAME" type="string" />
<many-to-one name="parentFirm" column="PERSON_FIRM_ID" />
<many-to-one name="parentDepartment" column="PERSON_DEPARTMENT_ID" />
</subclass>
<subclass name="Firm" discriminator-value="Firm" >
<set name="employees" >
<key column="PERSON_FIRM_ID"/>
<one-to-many class="Person"/>
</set>
<set name="subDepartments" >
<key column="DEPARTMENT_FIRM_ID"/>
<one-to-many class="Department"/>
</set>
</subclass>
<subclass name="Department" discriminator-value="Department" >
<many-to-one name="parentDepartment" column="DEPARTMENT_DEPARTMENT_ID" />
<many-to-one name="parentFirm" column="DEPARTMENT_FIRM_ID" />
<set name="subDepartments" >
<key column="DEPARTMENT_DEPARTMENT_ID"/>
<one-to-many class="Department"/>
</set>
<set name="employees" >
<key column="PERSON_DEPARTMENT_ID"/>
<one-to-many class="Person"/>
</set>
</subclass>
</class>
</hibernate-mapping>
---------------------
Does this work as designed?
Why does the hibernate-documentation recommend, not to use joined-subclasses, what sounds more straightforward?
And what still bothers me is that hibernate permanently uses (self-)joins when trying to retrieve the object-tree.
This does not make sense as each retrieved object can be mapped to a single row of T_CONTACT and is a performance-killer.

Thanks in advance

Peter K

yes. the alternative option is to do something like this:






It performs better, and the generated SQL is more understandable.




Eh?

I don't know what you are saying .... but you do know how to disable outer join fetching, right?

Now I know ;)

Thanks a lot

Peter K