Re: Is drop/restore trigger transactional?

On Tue, Aug 7, 2012 at 2:39 PM, Craig James <cjames@emolecules.com> wrote:
> On Tue, Aug 7, 2012 at 1:45 PM, Jeff Janes <jeff.janes@gmail.com> wrote:
>> On Tue, Aug 7, 2012 at 1:15 PM, Merlin Moncure <mmoncure@gmail.com> wrote:
>>>
>>> IF current_user = 'bulk_writer' THEN
>>>   return new;
>>> END IF;
>>> <expensive stuff>
>>
>> I don't know Craig's case, but often the most expensive of the
>> "expensive stuff" is the bare fact of firing a trigger in the first
>> place.
>
> My use case is pretty simple: Copy some already-validated user data
> from one schema to another.  Since the trigger has already been
> applied, we're guaranteed that the data is already in the form we
> want.
>
> For your amusement:

Thanks.  That was probably more amusing to me in particular than to most
pgsql hackers, as I think I've been a victim of your trigger.


...
>
> Obviously this is a very expensive trigger, but one that we can drop
> in a very specific circumstance.  But we NEVER want to drop it for
> everyone.  It seems like a very reasonable use-case to me.

And since the query is absolutely expensive, not just expensive
relative to a no-op, then Merlin's suggestion seems entirely suitable
for your use-case.

Cheers,

Jeff