I have spent the best part of 3 hours this morning looking around the web, JIRA and forums, and whilst finding related topics, I have not found the definitive answer.

I have a very basic Parent 1..N Child association and I have developed a search function that allows a user to search on both parent and child properties at the same time, e.g the parent might need to be restricted to a department of say I.T, and the child's category to Hardware.

The Criteria API has been ideal for this, and I have the full querying setup. The problem is in paginating these results.

When I set a row count onto the query, Hibernate sends a big join between Parent and Child with a TOP 10 prepended to the statement. The trouble is, many of these 10 are duplicates in terms of the root entity Parent.

So I found the ResultsTransformer and by applying a distinct root entity, I can "collapse" the results down to be distinct which is great except that the list I get back from Hibernate is not 10 at all, but 4 because of this distinctifying AFTER the query.

It is this which appears to be at the heart of all the posts on this topic and no real answer exists or appears to be on the way.

I don't really want to keep querying to fill up my list of results to 10, so I am not sure what direction to take here.

Any ideas would be great.

I spent a lot more time on this question and came to the same result as you did.
It's a pitty that this issue has not been resolved in the latest release (3.2.0). To me it's essential. JIRA HB-520 exists for it (http://opensource.atlassian.com/project ... wse/HB-520), but it is only minor. So there is no hope that this will be resolved soon.
However, you could vote for it.