Re: altering foreign key without a table scan

Vincent de Phily <vincent(dot)dephily(at)mobile-devices(dot)fr> writes:
> On Friday 19 August 2011 11:52:50 Tom Lane wrote:
>> IIRC, there are fields of pg_constraint that are copied into the
>> pg_trigger rows for the supporting triggers, so as to save one catalog
>> lookup at run time. If you diddle one manually, you'd better diddle
>> both.

> Some relid values are indeed duplicated in pg_constraint and pg_trigger, but
> it doesn't look like I need to fiddle with those ?

> I'm only touching pg_trigger.tgfoid and pg_constraint.confdeltype/confupdtype
> (which indeed seem to say the same thing in a different way). Do you know if
> there is something else I've missed ?

Yeah, that seems to be it except for the deferrable/deferred fields,
which match up in the obvious way. I had been thinking the RI triggers
avoided doing a lookup in pg_constraint, but that was mistaken. (I
think we used to store all that info in tgargs, but we evidently don't
anymore.)

regards, tom lane