openjpa-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Dick <>
Subject Re: How work around OpenJPA-1365
Date Tue, 17 Nov 2009 15:48:05 GMT
Hi all,

Daryl's correct, the duplicates returned by a JOIN FETCH clause are correct
behavior. If you don't want the duplicates to be returned you should use the
DISTINCT keyword and OpenJPA will remove duplicates from the list.

There are several outstanding issues though (which I'm working on):

1.) OPENJPA-894:  When results are returned from the database OpenJPA
automatically removes duplicates from the list. If the results are fetched
from memory the duplicates reappear.

2.) OPENJPA-1365: After you apply the fix for OPENJPA-894 the distinct
keyword doesn't work. This is because OpenJPA merely prepends the DISTINCT
keyword to the SQL generated which doesn't work if you're selecting across
multiple tables. Instead we need to filter the result list after retrieving
from the database.

There are two proposed fixes for OPENJPA-894 each of which have some

3a.) Mike's fix : supports pagination but does not support multiple JOIN
FETCH statements (ie SELECT m FROM Manager m JOIN FETCH m.employees JOIN
FETCH m.projects returns the wrong number of results).

3b.) Fay's fix : supported multiple JOIN FETCH statements, but does not
support pagination (ie query.setMaxResults(), query.setFirstResult() doesn't
page forward as expected).

So there is work being done, but it's turned out to be a very ticklish issue
to solve.

At the risk of muddying the waters a bit if you're migrating a JDO
application have you considered using OpenJPA's FetchPlan implementation to
eagerly load some fields? Over medium - large datasets I've found this to be
significantly faster than using a JOIN FETCH, but YMMV.

Hope this helps,


On Tue, Nov 17, 2009 at 8:48 AM, Daryl Stultz <> wrote:

> On Tue, Nov 17, 2009 at 9:27 AM, Michael Simons
> <>wrote:
> > You state, that you're query with distinct and join fetch does work
> > properly. But this would
> > mean OpenJPA-1365 doesn't occur, does it?
> >
> The JIRA states:
> This issue occurs if the proposed fix for OPENJPA-894 is in place.
> So 1365 does not occur unless you've patched your code such that 894 is
> fixed. What version of OpenJPA are you using and do you have any patches in
> place?
> > When we call "select a from A a join fetch B" we get n instances of A,
> with
> > n = numbers of
> > A-B-associations.
> >
> This is the correct behavior. I have found with OpenJPA 1.2.1, I get
> distinct rows of A which sounds like what you (and I) want but is improper.
> 894 shows that a second run of the query in the same EntityManager yields
> duplicates A's (with LEFT JOIN FETCH). What happens when you do this:
> select distinct a from A a join fetch
> ?
> How about these two:
> select a from A a left join fetch
> select distinct a from A a left join fetch
> I don't want to give the impression that I'm an expert on the matter, just
> that I've dealt with this issue and I want to be sure my understanding of
> things is accurate.
> --
> Daryl Stultz
> _____________________________________
> 6 Degrees Software and Consulting, Inc.

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message