Aggregate functions, %notfound and group by

Question

Hi,
Below are 2 versions of a script that query the employees table and get the max salary by specific departments. Department ids are held in a varray.
With 'group by':
declare
    type c is ref cursor;
    c1 c;
    sal employees.salary%type;
    type rid is varray(3) of number;
    r_id rid:=rid();
begin
r_id.extend(3);
r_id(1):=50;
r_id(2):=80;
r_id(3):=200;
for y in 1..3
loop
open c1 for select max(salary) from employees
where department_id =r_id(y)
group by department_id;
                    fetch c1 into sal;
                    if c1%notfound then                         
                    dbms_output.put_line('not found');
                    else
                    dbms_output.put_line(sal);
                    end if;
    close c1;
end loop;
end;No 'group by':
declare
    type c is ref cursor;
    c1 c;
    sal employees.salary%type;
    type rid is varray(3) of number;
    r_id rid:=rid();
begin
r_id.extend(3);
r_id(1):=50;
r_id(2):=80;
r_id(3):=200;
for y in 1..3
loop
open c1 for select max(salary) from employees
where department_id =r_id(y);
                    fetch c1 into sal;
                    if c1%notfound then                         
                    dbms_output.put_line('not found');
                    else
                    dbms_output.put_line(sal);
                    end if;
    close c1;
end loop;
end;The version with no group by clause always returns false for cursor attribute %notfound. Whereas in the version with the clause %notfound is true when an non-existent department id is entered. Why is this?
Product info: Oracle Database 10g Enterprise Edition Release 10.2.0.1.0
PL/SQL Release 10.2.0.1.0
Thanks.

Answer 1

<strike>Because the group by works on the rows that are returned from the where clause, and then the columns are applied. If you pass a condition into the where clause that causes no rows to be returned, then there's nothing to group by,
Whereas when you select an aggregate query without a group by, you always get a row returned.
eg: select count(*) from (
select max(dummy)
from   dual
where  dummy = 'Y')
  COUNT(*)
         1
select count(*) from (
select max(dummy)
from   dual
where  dummy = 'Y'
group by dummy)
  COUNT(*)
         0</strike>
Edited by: Boneist on 19-Aug-2009 15:21
Hmm... the documention seems to indicate that you should always get a row back, even if it's null.
I have no idea why adding the group by causes that statement to not be true, then, unless I'm misunderstanding the documentation!!