Tuning SQL query with similar subqueries for select columns

Question

Hi all,
My query is something like below:
1> SELECT
2> A.COL1,
3> SUM(CASE WHEN A.flag=100 AND NVL(B.flag,0)=0 AND
4> EXISTS (
5> SELECT 'ROW_EXISTS'
6> FROM A A0
7> WHERE A0.COL2=100 AND NVL(A0.flag,0)=0 AND 0.DIRN<>A.DIRN)
8> THEN 1
9> ELSE 0
10> END) SUM_COLUMN1,
11>SUM(CASE WHEN A.flag=100 AND
12> EXISTS (
13> SELECT 'ROW_EXISTS'
14> FROM A A0
15> WHERE A0.COL2=100 AND A0.DIRN<>A.DIRN)
16> THEN 1
17> ELSE 0
18>END) SUM_COLUMN2
19>FROM A,B
20>WHERE A.COL=B.KEY_COL
21>GROUP BY A.COL1
My problem is that I need to index scan the tables A aliased as A0 in the two exist queries. Lines: 5-7 and 13-15.
The major performance degradation i see with the query is that it has to scan A0 twice to compute SUM_COLUMN1 and SUM_COLUMN2.
Is there any way by which i can use the same subquery to scan the table just once to compute the columns.
The table A has more than a 100million records. So i need to improve performance of this query though it is a small issue.
Please feel free to ask me more if i'm not clear with the example.
Thanks in advance,
Raj

Answer 1

Also you could use analytic functions:
SQL> select deptno, sum(case when e.mgr is not null and
  2  exists (select 1 from emp e1 where e1.empno <> e.empno
  3  and e1.deptno = 10 and e1.sal >= 1500)
  4  then 1 else 0 end) sum1,
  5  sum(case when e.mgr is not null and e.sal >= 1200 and
  6  exists (select 1 from emp e1 where e1.empno <> e.empno and e1.deptno = 10)
  7  then 1 else 0 end) sum2
  8  from emp e
  9  group by deptno
10  /
    DEPTNO       SUM1       SUM2
        10          2          2
        20          5          3
        30          6          5
SQL> select deptno, sum(case when mgr is not null and
  2  (d12 != empno or empno !=d11) then 1 else 0 end) sum1,
  3  sum(case when mgr is not null and (d21 != empno or empno !=d22)
  4  and sal >= 1200 then 1 else 0 end) sum2
  5  from (
  6  select deptno, empno, mgr, sal,
  7  min(case when deptno = 10 and sal >= 1500
  8  then empno else -1 end) over(order by null) d11,
  9  max(case when deptno = 10 and sal >= 1500
10  then empno else -1 end) over(order by null) d12,
11  min(case when deptno = 10 then empno else -1 end) over(order by null) d21,
12  max(case when deptno = 10 then empno else -1 end) over(order by null) d22
13  from emp e
14  )
15  group by deptno
16  /
    DEPTNO       SUM1       SUM2
        10          2          2
        20          5          3
        30          6          5In your case it could be something like
SELECT COL1,
SUM( CASE WHEN a_flag = 100 and nvl(b_flag,0)=0 and
(DIRN != MIN_A OR DIRN != MAX_A) THEN 1 ELSE 0 END) SUM_COLUMN1,
SUM( CASE WHEN a_flag = 100 and
(DIRN != MIN_B OR DIRN != MAX_B) THEN 1 ELSE 0 END) SUM_COLUMN2
FROM (
SELECT A.COL1, A.FLAG A_FLAG, B.FLAG B_FLAG, A.DIRN,
MIN(CASE WHEN A.COL2=100 and NVL(A.flag,0) = 0 THEN A.DIRN ELSE null END) OVER(order by null) MIN_A,
MAX(CASE WHEN A.COL2=100 and NVL(A.flag,0) = 0 THEN A.DIRN ELSE null END) OVER(order by null) MAX_A,
MIN(CASE WHEN A.COL2=100 THEN A.DIRN ELSE null END) OVER(order by null) MIN_B,
MAX(CASE WHEN A.COL2=100 THEN A.DIRN ELSE null END) OVER(order by null) MAX_B
FROM A,B
WHERE A.COL=B.KEY_COL
GROUP BY COL1
Rgds.