Using a dummy where-clause to force the use of an index

Question

I notice that an index only gets used when I use the index key in the where clause. Should I use a dummy where clause to force the index to be used?
SQL> create table emp (
2 empno NUMBER(5),
3 empname VARCHAR2(15) );
Table created.
SQL> create index idx_emp_no on emp (empno);
Index created.
SQL> insert into emp (empno, empname) values (1, 'Peter');
1 row created.
SQL> set autotrace traceonly;
SQL> select empno from emp;
Execution Plan
Plan hash value: 3956160932
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 1 | 13 | 3 (0)| 00:00:01 |
| 1 | TABLE ACCESS FULL| EMP | 1 | 13 | 3 (0)| 00:00:01 |
SQL> select empno from emp where empno > 0;
Execution Plan
Plan hash value: 434430053
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 1 | 13 | 1 (0)| 00:00:01 |
|* 1 | INDEX RANGE SCAN| IDX_EMP_NO | 1 | 13 | 1 (0)| 00:00:01 |
As you can see, using a dummy where clause (empno > 0), I manage to reduce the cost from 3 to 1.

Answer 1

Again. Be careful with comparisons. But I like the discussion. It is usefull to carefully look at what happens "behind the scenes" and find a sensible explaination for that.
Your new scenario now is different. By selecting one specific row instead of selection all you already made an assumption. That only one or few of the rows are returned. Andre's statement still is correct. Now the CBO will prefere the index scan.
The CBO now has to decide between two options.
Option a) use the index.
This means =>
Step 1: Access the index to find the value
Step 2: Access the table using the rowid that was found in the index leaf entry.
Option b) scan the full table
This means =>
Step 1: Iterate over all rows of the table
Step 2: Apply a filter condition (empno = :x)
We can see this execution plan by adding a hint to the query.
select /*+ FULL(emp) */ * from emp where empno = 1;This hint will force the CBO to access the table emp using a full table scan.
The cost in my environment for option B is 3.
Option A has a cost with 2. Therefore the CBO prefers Option A.
However this can easily change.
Consider the following
begin
  for i in 1..1000 loop
    insert into emp (empno, empname) values (1, 'Employee');
  end loop;
end;
commit;
execute dbms_stats.gather_table_stats(user,'EMP');The run the select again.
SQL> set autotrace traceonly
SQL> select * from emp where empno = 1;
1001 Zeilen ausgewõhlt.
Ausf³hrungsplan
Plan hash value: 3956160932
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |  1001 | 12012 |     3   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| EMP  |  1001 | 12012 |     3   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter("EMPNO"=1)
Statistiken
          1  recursive calls
          0  db block gets
         74  consistent gets
          0  physical reads
          0  redo size
      18356  bytes sent via SQL*Net to client
       1250  bytes received via SQL*Net from client
         68  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
       1001  rows processedThe CBO now prefers the full table scan and not the index access.
If we force it to use an index then the cost increases.
SQL> set linesize 100
SQL> select /*+INDEX(emp) */ * from emp where empno = 1
1001 Zeilen ausgewõhlt.
Ausf³hrungsplan
Plan hash value: 2426388914
| Id  | Operation                   | Name      | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |           |  1001 | 12012 |     5   (0)| 00:00:01 |
|   1 |  TABLE ACCESS BY INDEX ROWID| EMP       |  1001 | 12012 |     5   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | IDX_EMPNO |  1001 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - access("EMPNO"=1)
Statistiken
          0  recursive calls
          0  db block gets
        139  consistent gets
          0  physical reads
          0  redo size
      30550  bytes sent via SQL*Net to client
       1250  bytes received via SQL*Net from client
         67  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
       1001  rows processed
SQL>The cost using the index range scan is now 5. WHile the FTS is still 3. Therefore the CBO prefers to use the FTS.