join fetch and scroll

taz.climber · **Posted:** Fri Jan 11, 2008 2:34 pm

Hello,

I have a model where I have a big collection of an object type which has childs. I want hibernate to get a part of my collection and all childs in one select.

So I thought about "join fetch" which avoids n select to get the child of each object in the collection (where n is the size of the collection).

I didn't want to get the whole collection in my select so I think about setMaxResult() and better scroll() because it tells you if there are more objects.

But "join fetch" and scroll() don't work together...

I tried to understand why : If i take a

Code:

Collection<A>

and A has an attribute :

Code:

public class A{
          private Collection<Thing> things;

and the HQL query :

Code:

"from A a left join fetch a.things where a member of ..."

then Hibernate can't sure that every element A of the collection get all its Thing before he reaches the end of the ResultSet of the query.

So if you use a scroll and stop scrolling before the end of the ResultSet, you may have invalid objects.

Am I right ?

It's important for me to understand why "join fetch" and scroll() don't work together because it can help me to solve my problem.

Thanks.

Fabrice

Hibernate version: 3.2.2.ga

gonzao_diaz · **Joined:** Mon Nov 26, 2007 2:29 pm **Posts:** 443

The scrolling/pagination capabilities rely on ROWNUM-type functions, they work with the size of your resultset. They wrap your original query on a simple outer query that says

Code:

select * from (<original query>) where rownum ...

If you do fetch joins, the size of your resultset contains rows belonging to a number of children per each parent, therefore the number of rows does not reflect the quantity of "parent" entities anymore.

scrolling/pagination is only possible when Hibernate is able to work keyset-style, with one row representing each top-level entity of your query.