Bruno Wolff III <bruno@wolff.to> writes:
> Jon Lapham <lapham@jandr.org> wrote:
>> When using queries with aggregate functions, is there any way to not
>> have to have to explicitly write all the columns names after the GROUP
>> BY ? I would like to use a wildcard "*".
> Don't those tables have primary keys? Grouping by the primay key of each
> table will produce the same result set as grouping by all of the columns.
Unfortunately, PG will still make him GROUP BY everything he wants to
use as a non-aggregated output column. This behavior is per SQL92
spec. SQL99 added some verbiage to the effect that you only need to
GROUP BY columns that the rest are functionally dependent on (this
covers primary keys and some other cases); but we haven't got round
to implementing that extension.
regards, tom lane
