mapping 2 versions of a POJO: POJO and POJOAggr

[Clark Butler]

I think this is a simple, real-world case that I cannot find documented anywhere in HBM.

I have a POJO which also contains a Set of dependent objects. For performance, I would like to have 2 ways of accessing - the POJO with and without the Set. The POJO version is for list results, then the POJO + Set version is for details of a single instance.

Normally I would represent these objects as a POJO and its Aggregate, like so (please imagine getters/setters for brevity):

    
public class Result {
   private Long resultID;
   private String resultName;
}
 
public class ResultAggr extends Result {
   private Set resultElements;
}
 
public class ResultElement {
   private Long resultElementID;
   private String data;
}

I cannot figure out how to model this in Hibernate.

If I map 2 separate objects to the same table, when I query for a List of search results, HBM calculates results for both objects (Result and ResultAggr), even though I am matching the proper class name in the query.

The HBM inheritance examples cover polymorphism very well, but does not cover a case as (seemingly) simple as this.

I think I am going to have to manage the sub-list entirely independently. But then I will lose the cascading and many-to-many abilities of HBM for the Aggregated case, which I don't want to do.

I very much appreciate any ideas. If it means anything, I'm using HBM3 via Spring's template wrappers.

[Clark Butler]

I should post my current Aggregated HBM descriptor:

    
    <class name="ResultAggr" table="RESULT">
        <id name="resultID" column="RESULTID">
<generator class="identity"/>
</id>
         <property name="resultType" column="REPORTTYPE" />
         <property name="resultDate" column="REPORTDATE" />
         <property name="description" column="DESCRIPTION" />
         <property name="outFile" column="OUTFILE" />
         <property name="status" column="STATUS" />
 
        <set name="resultElements" inverse="true" cascade="all-delete-orphan">
            <key column="RESULTID"/>
            <one-to-many class="ResultElement"/>
</set>
 
</class>

And so if I try to add

    
 
    <class name="ResultJBean" table="RESULT">
        <id name="resultID" column="RESULTID">
<generator class="identity"/>
</id>
         <property name="resultType" column="REPORTTYPE" />
         <property name="resultDate" column="REPORTDATE" />
         <property name="description" column="DESCRIPTION" />
         <property name="outFile" column="OUTFILE" />
         <property name="status" column="STATUS" />
 
</class>

HBM maps both, always.

But since there is no required discriminator, the only difference between the types being aggregation of dependent data, I can not find an inheritence-based HBM pattern that will implement this persistence.

Thanks again -

[Jeff Albertson]

Have you tried the following, which sounds like the simplest approach?

1. Have one POJO class with a Set-valued property.
2. Map the set lazily (the default in Hibernate 3.0):

<set name="resultElements" lazy="true">
...

In your code, these dependent object will only be selected from the database if you attempt to access them or ask Hibernate to explicitly initialize them.