openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeremy Bauer <>
Subject Re: @PersistentCollection question
Date Mon, 23 Feb 2009 16:17:58 GMT

I think you've found a valid problem in the test cases.  The basic use of
PersistentCollection does not enforce ordering so results will vary for each

Depending on what is being tested you could:

a) Modify the validation code to validate the collection based on
non-deterministic ordering of data.  This may be the best option since the
OpenJPA execution path being tested is the same.

b) Add explicit column ordering using @OrderBy.  This requires the use of a
persistent field/property that is order-able.  Possibly the Id field?  This
also requires that the elements in you your list are in that order.  Might
be get kind of messy - and this changes the OpenJPA execution path.

c) As Pinaki mentioned, use synthetic ordering.  This can be achieved with
@OrderColumn.  An 'order' column will be added to the collection table and
OpenJPA will use this column to manage the order of the collection.  If
changing the execution path isn't a big concern, this is likely the best
option.  However, OpenJPA has OrderColumn specific tests, so I vote for
option a) to keep the non-ordered path(s) covered, if possible.


On Mon, Feb 23, 2009 at 8:41 AM, Pinaki Poddar <> wrote:

> afaicr, OpenJPA was maintaining ordering for elements for field that is
> declared (or even instantiated) as java.util.List. It has the capability of
> creating synthetic column to maintain element ordering even when @OrderBy
> or
> some other ordering spec is absent.
> Tim McConnell wrote:
> >
> > Hi, I see a number of inconsistencies in the OpenJPA Junit testcases
> > concerning
> > the usage of the @PersistentCollection annotation on a java.util.List
> > variable
> > without the @OrderBy annotation. If I'm interpreting this scenario
> > correctly,
> > without the @OrderBy annotation there is no column in the database used
> > for
> > ordering when retrieving the List back into memory, and thus there can be
> > no
> > assumption made about that order in the testcase. This appears to be why
> > these
> > testcases work on some database and not others (e.g., works on Derby,
> > fails on
> > PostgreSQL). Am I interpreting it correctly ?? Or am I missing something
> > obvious ??
> >
> > --
> > Thanks,
> > Tim McConnell
> >
> >
> --
> View this message in context:
> Sent from the OpenJPA Developers mailing list archive at

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