In Oracle, which is best practice for 'NOT EQUAL TO'

Question

I need to check for where a decimal value is not zero.
Am I better to use
'less than symbol or greater than symbol' 0
or
!= 0or does Oracle substitute 'less than symbol or greater than symbol' to != effectively ?
In my mind if it doesn't the 'less than symbol or greater than symbol' would be slower because it has to check if the value is greater than OR less than.
I appreciate != is not SQL-92 compliant.
I appreciate it would be less than a nano second difference but they all add up!
Thanks
The 'less than symbol or greater than symbol' do not show it seems as they are used in HTML.
Edited by: cubmar on Jun 5, 2009 11:16 AM
Edited by: cubmar on Jun 5, 2009 11:21 AM

Answer 1

Even though The CBO is cleverly written program it does makes its assumptions. When ever the Optimizer finds a NOT operator it assumes that is going to fetch a large portion of the data and goes for a FULL TABLE SCAN. INDEX SCAN Is not an option.
And hence if you are not going to select a large portion of the data using NOT can be expensive.
Below is an example.
i have created a table which has one record with value as 1 and 99999 records with value as 0.
SQL> create table t
  2  as
  3  select decode(level,1,1,0) no, rpad('*',100,'*') name
  4    from dual
  5  connect by level <= 100000
  6  /
Table created.
SQL> create index t_idx on t(no)
  2  /
Index created.
SQL> exec dbms_stats.gather_table_stats(user,'T',cascade=>true)
PL/SQL procedure successfully completed.so when i use the condition no != 0 i am gong to select only 1 record. So i expect the optimizer to go for a INDEX RANGE SCAN but see what happens.
SQL> explain plan
  2  for
  3  select * from t where no != 0
  4  /
Explained.
SQL> select * from table(dbms_xplan.display)
  2  /
PLAN_TABLE_OUTPUT
Plan hash value: 1601196873
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |     1 |   103 |   345   (2)| 00:00:05 |
|*  1 |  TABLE ACCESS FULL| T    |     1 |   103 |   345   (2)| 00:00:05 |
Predicate Information (identified by operation id):
PLAN_TABLE_OUTPUT
   1 - filter("NO"<>0)
13 rows selected.But with the use of < or > you can go for a INDEX RANGE SCAN.
SQL> explain plan
  2  for
  3  select * from t where no >0 or no <0
  4  /
Explained.
SQL> select * from table(dbms_xplan.display)
  2  /
PLAN_TABLE_OUTPUT
Plan hash value: 4259936809
| Id  | Operation                    | Name  | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |       |     2 |   206 |     6   (0)| 00:00:01 |
|   1 |  CONCATENATION               |       |       |       |            |          |
|   2 |   TABLE ACCESS BY INDEX ROWID| T     |     1 |   103 |     3   (0)| 00:00:01 |
|*  3 |    INDEX RANGE SCAN          | T_IDX |     1 |       |     2   (0)| 00:00:01 |
|   4 |   TABLE ACCESS BY INDEX ROWID| T     |     1 |   103 |     3   (0)| 00:00:01 |
|*  5 |    INDEX RANGE SCAN          | T_IDX |     1 |       |     2   (0)| 00:00:01 |
PLAN_TABLE_OUTPUT
Predicate Information (identified by operation id):
   3 - access("NO"<0)
   5 - access("NO">0)
       filter(LNNVL("NO"<0))
19 rows selected.So you must be cautious when you use NOT operator.