On Thu, 2006-05-25 at 10:33 +0200, Thomas Hallgren wrote:
> Rafal Pietrak wrote:
> > I'd like to propose a 'syntax/semantics' of such trigger:
> >
> > Triggers normally execute inside of a transaction.
> >
> > A COMMIT within a trigger could mean: "do a fork: fork-1) return to the
> > main and schedule COMMIT there, fork-2) continue in bacground".
> >
> And what if fork-1) returns to the main, attempts the COMMIT but instead and
> rolls back due
> to a violated constraint? Where does that leave fork-2?
>
> Regards,
> Thomas Hallgren
No problem at all (at least in particular case of an application I have
in mind :). The precedure that remains within fork-2 just does a time
consuming housekeeping. Like a cleanup - always succeeds, even if
sometimes is not really necesary (like in case of main rolling-back).
And that's exacly why I thing that it should be 'released to run' by
RDBMS *after* the main COMMITS (or ROLLES-BACK). It should be run on
COMMITED (visible to the world) changes, not on session trancients.
-R
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
