Re: sql question regarding count(*)

On Fri, 7 Mar 2003, Rajesh Kumar Mallah wrote:

> On Friday 07 March 2003 11:41 am, Stephan Szabo wrote:
> > On Fri, 7 Mar 2003, Rajesh Kumar Mallah wrote:
> > > Tomasz did you notice the two queries are totally differet on different
> > > tables?
> >
> > A count() on a non-grouped table returns 1 row.
> >
> > A grouped table returns 1 row per group.  If there are no groups, no rows
> > are returned.  If the groups are filtered by a having clause, it might
> > also return no rows.
>
> True the second query does not return any row.
> even in first case no rows are returned.
>
> where is the difference?
>

I don't think I understand the question.

select userid from users where userid=-1
returns 1 row for each row the where clause returns true
and 0 rows if there are none.

select count(*) from users where userid=-1
returns 1 row (it's not a grouped table)

select count(*) from users where userid=-1 group by userid
returns 1 row if there is at least 1 row having userid=-1or 0 rows otherwise (either there is a group or there isn't)

select count(*) from users where userid=-1 group by useridhaving userid!=-1
should return 0 rows because even if there is a group it'll fail the
having clause.