Re: rule weirdness

On Mon, 2007-08-20 at 11:36 -0700, David E. Wheeler wrote:
> Yes, that makes sense. They should work if you run them in reverse
> order though, correct?
>
> create or replace rule rr_upd_rule as on update to rr do instead (
> UPDATE rr2 SET i=NEW.i, d2=NEW.d2 WHERE i=OLD.i
> UPDATE rr1 SET i=NEW.i, d1=NEW.d1 WHERE i=OLD.i;
> );
>

Without testing, I think symmetry would then shift the problem to a
statement such as:

update rr set i=41, d1=43 where i=11;
^

Also, I think you would need to make the FK deferrable in that case.

For a particular application, and with a certain amount of care, rules
can be very powerful. I am just mentioning that there is a lot of danger
lurking in rules as well, and they can't really create perfect
abstractions from physical relations.

I really wish that there was a system to create a more ideal abstraction
from physical relations. Rules are better than what most RDBMSs give
you.

Regards,
Jeff Davis