Bill Moseley <moseley@hank.org> writes:
> On Fri, Mar 09, 2007 at 06:50:39PM -0500, Tom Lane wrote:
>> This is not going to work because the row's not there yet.
> This is a BEFORE *UPDATE* trigger, not a BEFORE INSERT, so the row is
> there. The audit table is written when the primary record changes
> and the old version is written to the audit table, not the new
> version.
Well, if you want to write the old data, it's still a silly way to do
it: write the OLD.* tuple instead of forcing a fresh search of the
table.
> Ok, but as the id is a sequence. I need to test if NEW.id is set after
> the insert -- seems like not, IIRC, and I'd need to use curval().
You're confusing rules with triggers. In a trigger, NEW.* and OLD.*
are physical rows and you don't need to worry about multi evaluation
or anything like that.
regards, tom lane