The Hermit Hacker <scrappy@hub.org> writes:
> Anyone want to comment on this one? Just tested with v6.5.0 and it still
> exists there...
> vhosts=> create table test ( a int, b char );
> CREATE
> vhosts=> insert into test values ( 1, 'a' );
> INSERT 149258 1
> vhosts=> select a from test group by a having a > 0;
> ERROR: SELECT/HAVING requires aggregates to be valid
That's not a bug, it means what it says: HAVING clauses should contain
aggregate functions. Otherwise they might as well be WHERE clauses.
(In this example, flushing rows with negative a before the group step,
rather than after, is obviously a win, not least because it would
allow the use of an index on a.)
However, I can't see anything in the SQL92 spec that requires you to
use HAVING intelligently, so maybe this error should be downgraded to
a notice? "HAVING with no aggregates would be faster as a WHERE"
(but we'll do it anyway to satisfy pedants...)
regards, tom lane