IN Vs Exists

Question

Hi ,
Recently I was asked in an interview: When we would "IN" clause would be preferred to use over "EXISTS" clause. I had told the interviewer that "EXISTS" will ignore NULL value if the table has any. Wheter was I correct? Please provide valueable inputs.
Regards,
Achyut K

Answer 1

S10390 wrote:
Check this Tom's post
http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:953229842074
I think that was on a very old version of oracle (8.1.6). Oracle has changed a lot over the time.
Here is a simple example.
I have EMP and DEPT table like this. With relation defined as following
SQL> select table_name, constraint_name, constraint_type, r_constraint_name
  2    from user_constraints
  3   where table_name in ('EMP', 'DEPT');
TABLE_NAME                     CONSTRAINT_NAME                C R_CONSTRAINT_NAME
DEPT                           DEPT_PK                        P
EMP                            EMP_FK_DEPT_NO                 R DEPT_PK
EMP                            EMP_PK                         PNow a simple query using IN
SQL> select * from emp e where e.deptno in (select d.deptno from dept d);
     EMPNO ENAME  JOB              MGR HIREDATE         SAL        COM     DEPTNO
      7369 SMITH  CLERK           7902 02-APR-81       2975          0         20
      7499 ALLEN  SALESMAN        7698 20-FEB-81       1600        300         30
      7521 WARD   SALESMAN        7698 22-FEB-81       1250        500         30
      7566 JONES  MANAGER         7839 02-APR-81       2975          0         20
      7654 MARTIN SALESMAN        7698 28-SEP-81       1250       1400         30
      7698 BLAKE  MANAGER         7839 01-MAY-81       2850          0         30
      7782 CLARK  MANAGER         7839 09-JUN-81       2450          0         10
      7788 SCOTT  ANALYST         7566 19-APR-87       3000          0         20
      7839 KING   PRESIDENT            17-NOV-81       5000          0         10
      7844 TURNER SALESMAN        7698 08-SEP-81       1500          0         30
      7876 ADAMS  CLERK           7788 23-MAY-87       1100          0         20
11 rows selected.
SQL> select * from table(dbms_xplan.display_cursor);
PLAN_TABLE_OUTPUT
SQL_ID  f7h123zffyf6y, child number 0
select * from emp e where e.deptno in (select d.deptno from dept d)
Plan hash value: 2872589290
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |       |       |     2 (100)|          |
|*  1 |  TABLE ACCESS FULL| EMP  |    11 |   418 |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter("E"."DEPTNO" IS NOT NULL)
18 rows selected.The amazing thing is that the DEPT table is not scanned at all. I guess the CBO is using the referencial
integrity to determine scanning fo EMP table would be sufficient.
Now using EXISTS
SQL> select * from emp e where exists (select null from dept d where d.deptno = e.deptno);
     EMPNO ENAME  JOB              MGR HIREDATE         SAL        COM     DEPTNO
      7369 SMITH  CLERK           7902 02-APR-81       2975          0         20
      7499 ALLEN  SALESMAN        7698 20-FEB-81       1600        300         30
      7521 WARD   SALESMAN        7698 22-FEB-81       1250        500         30
      7566 JONES  MANAGER         7839 02-APR-81       2975          0         20
      7654 MARTIN SALESMAN        7698 28-SEP-81       1250       1400         30
      7698 BLAKE  MANAGER         7839 01-MAY-81       2850          0         30
      7782 CLARK  MANAGER         7839 09-JUN-81       2450          0         10
      7788 SCOTT  ANALYST         7566 19-APR-87       3000          0         20
      7839 KING   PRESIDENT            17-NOV-81       5000          0         10
      7844 TURNER SALESMAN        7698 08-SEP-81       1500          0         30
      7876 ADAMS  CLERK           7788 23-MAY-87       1100          0         20
11 rows selected.
SQL> select * from table(dbms_xplan.display_cursor);
PLAN_TABLE_OUTPUT
SQL_ID  8x38asybd9j0v, child number 0
select * from emp e where exists (select null from dept d where
d.deptno = e.deptno)
Plan hash value: 2610556560
| Id  | Operation          | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT   |         |       |       |     3 (100)|          |
|   1 |  NESTED LOOPS SEMI |         |    11 |   451 |     3   (0)| 00:00:01 |
|   2 |   TABLE ACCESS FULL| EMP     |    11 |   418 |     2   (0)| 00:00:01 |
|*  3 |   INDEX UNIQUE SCAN| DEPT_PK |     4 |    12 |     1   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - access("D"."DEPTNO"="E"."DEPTNO")
21 rows selected.Here the DEPT table is scanned. So as already said it all deptnds. And need to tested for every case.