Re: Detecting when outer joins are needed

From: Andrus Adamchik (andru..bjectstyle.org)
Date: Tue Aug 29 2006 - 05:20:29 EDT

  • Next message: Øyvind Harboe: "Re: Detecting when outer joins are needed"

    -1.

    My argument still stands - unless a user tells us what (s)he wants to
    do, there is no good way to guess it. Ok, we've even identified a few
    specific cases. But we still can't guess it in a general case.
    Changing the algorithm without understanding it is asking for trouble.

    > Also, how do I know that I've got the complete list of gotchas?

    How do you know that we have it handled correctly in Cayenne ? ;-)
    That's my point exactly - this requires analysis - Cayenne can't do
    it automagically.

    Andrus

    On Aug 28, 2006, at 9:54 AM, Øyvind Harboe wrote:

    > On 8/25/06, Andrus Adamchik <andru..bjectstyle.org> wrote:
    >>
    >> On Aug 23, 2006, at 11:35 PM, Mike Kienenberger wrote:
    >>
    >> > Here's where I'm finding that I need outer joins:
    >> >
    >> > 1) If two qualifiers for a non-mandatory relationship path are or'd
    >> > together. This is also what Øyvind Harboe reported. It may
    >> be that
    >> > the non-mandatory part is irrelevent.
    >> >
    >> > 2) If using noMatchExp on a non-null value for a non-mandatory
    >> > relationship path.
    >>
    >> Thanks Mike - that makes sense. I think "non mandatory" part is
    >> relevant. Not sure that we need to catch it though. IMO this should
    >> be left as an exercise to the user.
    >
    > I could really do with having this exercise clearly formulated in
    > doc/wiki...
    >
    > I have a hard time explaining my peers what precisely it is that they
    > should be looking for.
    >
    > Also, how do I know that I've got the complete list of gotchas?
    >
    > Is it in theory possible to detect this during runtime?
    >
    > It would be nice to be able to add some runtime checks that generated
    > expressions aren't running into this problem.
    >
    > --
    > Øyvind Harboe
    > http://www.zylin.com
    >



    This archive was generated by hypermail 2.0.0 : Tue Aug 29 2006 - 05:22:41 EDT