Re: Understanding behavior of SELECT with multiple unnested columns

On Wed, Mar 27, 2013 at 9:03 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Gavin Flower <GavinFlower(at)archidevsys(dot)co(dot)nz> writes:
>> The rule appears to be,
>> where N_x & N_y are the number of entries returned for x & y:
>> N_result = is the smallest positive integer that has N_x & N_y as factors.
>
> Right: if there are multiple set-returning functions in a SELECT list,
> the number of rows you get is the least common multiple of their
> periods. (See the logic in ExecTargetList that cycles the SRFs until
> they all report "done" at the same time.) I guess there's some value
> in this for the case where they all have the same period, but otherwise
> it's kind of bizarre. It's been like that since Berkeley days though,
> so I doubt we'll consider changing it now. Rather, it'll just be
> quietly deprecated in favor of putting SRFs into FROM (with LATERAL
> where needed).

It's a neat way to make a query that doesn't terminate (which AFAIK is
impossible in vanilla SQL):

create sequence s;
select generate_series(1,nextval('s')), generate_series(1,nextval('s'));

merlin