Stephan Szabo <sszabo@megazone23.bigpanda.com> writes:
> On Mon, 9 Sep 2002, Bruce Momjian wrote:
>> All the problems here are coming from INSTEAD rules. We don't have
>> INSTEAD triggers or contraints.
> Sure we do, well sort of. :)
> Make a before trigger that does a different statement and returns NULL
> to abort the original action on that row.
I think we can reasonably leave the side-effects of triggers out of the
discussion. PQcmdStatus numbers have never included side-effects of
triggers in the past, and I see no reason for them to start now.
I think it's reasonable to exclude both triggers and non-INSTEAD rules
from the status count, on the grounds that these normally represent
"add-on" actions and not the "real" action. The cases that get
interesting are those that involve multiple INSTEAD actions (either from
multiple INSTEAD rules, or a single rule with multiple commands in its
body) and those cases where the INSTEAD action is a different type from
the original command (eg, ON UPDATE DO INSTEAD INSERT...).
regards, tom lane