Comments on: Why cannot I use subquery there?

By: Laurent Schneider

Laurent Schneider — Mon, 28 Jan 2008 14:50:52 +0000

Oracle fails in noticing the scalar subquery is constant and not correlated.

I filed a documentation bug since GROUP BY () is not documented in the SQL Reference but it is implictely documented Oracle® Database Data Warehousing Guide, SQL for Aggregation in Data Warehouses
GROUP BY ROLLUP(a, (b, c))
This is equivalent to:

GROUP BY a, b, c UNION ALL
GROUP BY a UNION ALL
GROUP BY ()

note the latest group by !

in the following query, the group by clause is mandatory, because the subquery is correlated

select
    (select dname from dept where emp.deptno=dept.deptno),
  count(*)
  from emp
  group by deptno;

(SELECTDNAMEFR   COUNT(*)
-------------- ----------
SALES                   6
RESEARCH                5
ACCOUNTING              3

By: Martin Smevik

Martin Smevik — Mon, 28 Jan 2008 13:27:56 +0000

I would guess that the reason for the apparent difference is that you can’t have an aggregate function/or group function in a non-group query. The function count() requires that only one row is returned, while the sub-query will return an unknown amount of rows.

This goes for this example also:
SELECT COUNT(*), TABLE_NAME
FROM USER_TABLES;

ERROR at line 1:
ORA-00937: not a single-group group function

I would like to state that this is merely a undocumented theory. Anyone have a better explanation?

By: Laurent Schneider

Laurent Schneider — Fri, 25 Jan 2008 05:53:18 +0000

me neither

By: Chen Shapira

Chen Shapira — Fri, 25 Jan 2008 04:13:45 +0000

Isn’t this query:

SQL> SELECT (SELECT COUNT(*) FROM EMP),
COUNT(*) FROM DEPT ;

equivalent to this:

SQL> SELECT (SELECT COUNT(*) FROM EMP),
COUNT(*) FROM DEPT GROUP BY ();

Don’t see the reason to fail one and allow the other.