Count(*) with nested query

Question

Hi,
I have a question about the count(*) with nested query.
I have a table T1 with these columns:
C1 number
C2 number
C3 number
C4 number
C5 number
(The type of each column is not relevant for the example.)
This query:
select C1, C2, C3, C4
from T1
group by C1, C2
it's not correct becausa C3 and C4 are not columns specified in the GROUP BY expression.
If if run this query:
select count(*)
from (select C1, C2, C3, C4
from T1
group by C1, C2)
I haven't an error message (the result is correctly the number of records).
Why?
Thanks.
Best regards,
Luca

Answer 1

Because you are just selecting count(*) and none of the columns from the subquery, Oracle is optimising it by ignoring the selected columns and just running the sub query with the group by columns. I know it seems odd, but if you take a basic example:
SQL> ed
Wrote file afiedt.buf
  1  select count(*)
  2  from (select empno, sal, mgr, deptno
  3  from emp
  4* group by deptno)
SQL> /
  COUNT(*)
         3... all columns but deptno are ignored
... but if you include one of the other columns, even if you group by that column...
SQL> ed
Wrote file afiedt.buf
  1  select count(*), empno
  2  from (select empno, sal, mgr, deptno
  3  from emp
  4  group by deptno)
  5* group by empno
SQL> /
group by empno
ERROR at line 5:
ORA-00979: not a GROUP BY expression
SQL>... the error returns, because you're forcing oracle to include the column in the subquery.