Say I have a join table that joins table1 and table2. This join table is going to define equals and compareTo based on table1 and table2, so I would probably be tempted to write this:



Say I query for a list of these objects, then throw them all in a TreeSet. Will I then get two selects for every object because TreeSet is called getTable1() and getTable2() all over the place? If so, is there a way to define equality on the IDs, rather then the full objects?

Yes, just call the identifier getter.

.append(getTable1().getId(), rhs.getTable1().getId())

Invoking the identifier getter does not cause initialization of uninitialized proxies, since the identifier value of a many-to-one is loaded along with the parent entity, even if the object model tends to obscure that.

In truth, you'd probably do something like this:

private static Long getId(Identifiable obj)
{
return obj == null ? null : obj.getId();
}
...
.append(getId(getTable1()), getId(rhs.getTable1())

So as to avoid NPEs when sorting transient objects.

using id's for equals/hashcode has issues - see wiki about equals and hashcode.