Bruce Momjian <bruce@momjian.us> writes:
> Ron Mayer wrote:
>> Seems less fool-proof to me.
> My point is that people should _know_ they are using a cartesian
> product, and a warning would do that for users who have no need for a
> cartesian product and want to be warned about a possible error.
There are quite a lot of standard applications where small cartesian
products make sense --- star schemas are the traditional example.
I recall some discussions awhile back about how to persuade the planner
to consider such cases, in fact. Right now it'll generally use a series
of hash joins where maybe just one would be better.
I concur with Ron that the only merit of this proposal is that it's
(relatively) simple to implement.
regards, tom lane