Query performance question!

Question

I have 2 SQL shown below. the first has no index and the others have 2 indexes. It's a huge table with no primary key. From below explain plan, which is more efficient? or someone can help to rewrite the query. I just use 'COUNT' to easily get the explain plan instead of selecting columns for tables. Any input is appreciated.
SQL> l
  1  select count(*)
  2    from  etl_orcl.tmp_trade_lot34 tt
  3*  where  not (nvl(trim(status_code), 'NA') = 'OO' or nvl(trim(status_code), 'NA') = 'PC' or nvl(trim(m_status_code), 'NA') = 'PC')
SQL> /
  COUNT(*)
   2574571
Elapsed: 00:00:47.74
Execution Plan
| Id  | Operation          | Name            | Rows  | Bytes | Cost (%CPU)|
|   0 | SELECT STATEMENT   |                 |     1 |     7 | 28330   (6)|
|   1 |  SORT AGGREGATE    |                 |     1 |     7 |            |
|   2 |   TABLE ACCESS FULL| TMP_TRADE_LOT34 |   322 |  2254 | 28330   (6)|
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
SQL> l
  1  select count(*)
  2    from  etl_orcl.tmp_trade_lot34 tt
  3*  where  not (status_code in ('OO','PC') or m_status_code = 'PC')
SQL> /
  COUNT(*)
   2574571
Elapsed: 00:00:24.24
Execution Plan
| Id  | Operation               | Name              | Rows  | Bytes | Cost (%CPU)|
|   0 | SELECT STATEMENT        |                   |     1 |     7 |  7798   (6)|
|   1 |  SORT AGGREGATE         |                   |     1 |     7 |            |
|   2 |   VIEW                  | index$_join$_001  |  2574K|    17M|  7798   (6)|
|   3 |    HASH JOIN            |                   |       |       |            |
|   4 |     INDEX FAST FULL SCAN| IDX#2_TRADE_LOT34 |  2574K|    17M|  1659   (7)|
|   5 |     INDEX FAST FULL SCAN| IDX#1_TRADE_LOT34 |  2574K|    17M|  1768   (7)|
----------------------------------------------------------------------------------Edited by: gonzroman on May 7, 2010 12:51 PM
Edited by: gonzroman on May 7, 2010 12:52 PM

Answer 1

From below explain plan, which is more efficient?Which one finishes first?
where not (nvl(trim(status_code), 'NA') = 'OO' or nvl(trim(status_code), 'NA') = 'PC' or nvl(trim(m_status_code), 'NA') = 'PC')The NVL functions are causing the optimizer to bypass the index as the results of the function are not indexed.
Keep in mind that counts can use a different execution path than returning data if the index can be used;
Sample table with index;
create table t as (
   select rownum rn, rownum*10 val
   from dual
   connect by rownum <= 100000);
Table created
create index t_i on t(rn);
Index createdExplain plan for COUNT;
explain plan for
   select count(*)
   from t
   where rn between 10 and 20;
Explained
select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
Plan hash value: 4142320527
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |     1 |    13 |     2   (0)| 00:00:01 |
|   1 |  SORT AGGREGATE   |      |     1 |    13 |            |          |
|*  2 |   INDEX RANGE SCAN| T_I  |    11 |   143 |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - access("RN">=10 AND "RN"<=20)
Note
   - dynamic sampling used for this statement
18 rows selectedExplain plan for query that returns data from the table;
explain plan for
   select val
   from t
   where rn between 10 and 20;
Explained
select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
Plan hash value: 1658221841
| Id  | Operation                   | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |      |    11 |   286 |     3   (0)| 00:00:01 |
|   1 |  TABLE ACCESS BY INDEX ROWID| T    |    11 |   286 |     3   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | T_I  |    11 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - access("RN">=10 AND "RN"<=20)
Note
   - dynamic sampling used for this statement
18 rows selected