Hibernate version:
3.1.1

Mapping documents:
<class name="ca.ubc.nmc.contacts.db.NcContact" table="NC_CONTACT">
<many-to-one name="orgUnitIdAndorgUnitType" class="ca.ubc.nmc.contacts.db.NcOrgUnit" fetch="select">
<column name="ORG_UNIT_ID" length="10" />
<column name="ORG_UNIT_TYPE" length="4" />
</many-to-one>
</class>

<class name="ca.ubc.nmc.contacts.db.NcOrgUnit" table="NC_ORG_UNIT">
<composite-id name="id" class="ca.ubc.nmc.contacts.db.NcOrgUnitId">
<key-property name="orgUnitId" type="string">
<column name="ORG_UNIT_ID" length="10" />
</key-property>
<key-property name="orgUnitTypeCde" type="string">
<column name="ORG_UNIT_TYPE_CDE" length="4" />
</key-property>
</composite-id>
</class>

Because of some strange application programming, the value
that would normally be in the column ORG_UNIT_TYPE
is determined at run time so I'm wondering if it is possible
to modify the query that is used to retrieve the NcOrgUnit
row when I do a "from NcContact" so when I reference
NcContact.orgUnitIdAndorgUnitType I get the appropriate value?

I don't think I fully understand the question, but is it possible that you want to use the formula attribute instead of the column attribute? This will make the column read-only, though, so you may want two mappings for the same column, one using formula and one using column. Alternatively, you could keep the existing mapping and add a new member to the pojo that implements the heuristics for you. I.e. the pojo cound have a getORG_UNIT_TYPE and a getORG_UNIT_TYPERuntime.

I'm hibernate newbie so I'm not up the forumula use.

To rephrase my question:

I would like some way to modify the query used by a lazy load
for a compound key to use both a value from the parent row
and a value to be provided by my code.

Thanks

I'd have to save that that's not something that data access layers do. That's busniess logic. You could fake it like this: use hibernate to map the value in question to an "internal" value that you don't advertise in your class' interface (getInternalOrgUnitType(), or whatever). Then write your class' accessor (getOrgUnitType()) with whatever logic you need to determine the value you want to return. Something like

I realize it is a strange situation (and not one I would have designed if
I had written the initial implementation).

I don't really care about orgUnitType being available to
to the "outside world". What I do want is the ncOrgUnit value
for the NC_CONTACT table which right now is always null
because the lazy load query always fails due to orgUnitType
having the "wrong" value for the SELECT.

Right, I think I'm all caught up. You've got a relationship based on a two-column key, but the type column in NcContact, despite having the right name and type, will (almost) always have the wrong value.

There's probably no simple mapping that will fix this. The only simple in-mapping solution that I can think of applies only if the type value that you want is constant, or is stored in some other column in NcContact. In this case, you can change the many-to-one mapping in NcContact to use the "formula" element instead of the "column" element. A formula element passes simple SQL (not HQL) to the query, so you could replace <column name="ORG_UNIT_TYPE"/> in NcContact with <formula>4</formula> if the value was a constant 4; or you could use <formula>ORG_UNIT_OTHER_COLUMN_NAME</formula> if the value is in another column. You can also put other SQL in there; a UDF, a simple mathematical formula, etc. Obviously, this doesn't allow for a user-supplied value.

The in-java solution would be more flexible, but with a bit of an overhead. Map the relationship as a many-to-many with just the ORG_UNIT_ID column specified, and store the resulting NcOrgUnits in a Map. Then you can implement something like this: