Distinct on 10 million records

Question

Hello everyone.
I have a table called BL_DOCUMENTS which has 10 million records in it.
It has 'DocumentId' column which is the PrimaryKey.
I created an index on three columns:
CREATE INDEX IX_BD_CREA_DOCU_ISDE_ISCUR ON BL_DOCUMENTS (CREATORID ASC, ISDELETED ASC, ISCURRENT ASC)
I run a simple query to get the 'CreatorId' column's values with 'distinct' (there are currently 25,000 different ids):
select O., rownum rnum from(*
SELECT distinct  D.CreatorId  ObjectId FROM BL_DOCUMENTS  D
WHERE D.IsCurrent = 1 AND  D.IsDeleted = 0
*     ORDER BY ObjectId DESC*
*) O where rownum <= 1000*
This query runs for 7-10 seconds.
If I remove the distinct the query runs for 0.007 secs.
How can I improve this?

Answer 1

DevMTs wrote:
I try that but it's not improved.Still changing index column order should give some improvement since INDEX SKIP SCAN DESCENDING is a bit more costly operation than INDEX RANGE SCAN DESCENDING:
SQL> create table BL_DOCUMENTS(CreatorId number,IsCurrent number,IsDeleted number);
Table created.
SQL> CREATE INDEX IX_BD_CREA_DOCU_ISDE_ISCUR ON BL_DOCUMENTS (CREATORID ASC, ISDELETED ASC, ISCURRENT ASC);
Index created.
SQL> exec dbms_stats.gather_table_stats(ownname => user, tabname => 'BL_DOCUMENTS',cascade => true);
PL/SQL procedure successfully completed.
SQL> explain plan for
  2  select O.*, rownum rnum from(
  3  SELECT distinct D.CreatorId ObjectId FROM BL_DOCUMENTS D
  4  WHERE D.IsCurrent = 1 AND D.IsDeleted = 0
  5  ORDER BY ObjectId DESC
  6  ) O where rownum <= 1000
  7  /
Explained.
SQL> @?\rdbms\admin\utlxpls
PLAN_TABLE_OUTPUT
Plan hash value: 4024680831
| Id  | Operation                     | Name                       | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT              |                            |     1 |    13 |     1 (100)| 00:00:01 |
|*  1 |  COUNT STOPKEY                |                            |       |       |            |          |
|   2 |   VIEW                        |                            |     1 |    13 |     1 (100)| 00:00:01 |
|   3 |    SORT UNIQUE NOSORT         |                            |     1 |    39 |     1 (100)| 00:00:01 |
|*  4 |     INDEX SKIP SCAN DESCENDING| IX_BD_CREA_DOCU_ISDE_ISCUR |     1 |    39 |     0   (0)| 00:00:01 |
PLAN_TABLE_OUTPUT
Predicate Information (identified by operation id):
   1 - filter(ROWNUM<=1000)
   4 - access("D"."ISDELETED"=0 AND "D"."ISCURRENT"=1)
       filter("D"."ISCURRENT"=1 AND "D"."ISDELETED"=0)
18 rows selected.
SQL> drop index IX_BD_CREA_DOCU_ISDE_ISCUR;
Index dropped.
SQL> CREATE INDEX IX_BD_CREA_DOCU_ISDE_ISCUR ON BL_DOCUMENTS (ISCURRENT,ISDELETED,CREATORID);
Index created.
SQL> exec dbms_stats.gather_table_stats(ownname => user, tabname => 'BL_DOCUMENTS',cascade => true);
PL/SQL procedure successfully completed.
SQL> explain plan for
  2  select O.*, rownum rnum from(
  3  SELECT distinct D.CreatorId ObjectId FROM BL_DOCUMENTS D
  4  WHERE D.IsCurrent = 1 AND D.IsDeleted = 0
  5  ORDER BY ObjectId DESC
  6  ) O where rownum <= 1000
  7  /
Explained.
SQL> @?\rdbms\admin\utlxpls
PLAN_TABLE_OUTPUT
Plan hash value: 3922126944
| Id  | Operation                      | Name                       | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT               |                            |     1 |    13 |     1 (100)| 00:00:01 |
|*  1 |  COUNT STOPKEY                 |                            |       |       |            |          |
|   2 |   VIEW                         |                            |     1 |    13 |     1 (100)| 00:00:01 |
|   3 |    SORT UNIQUE NOSORT          |                            |     1 |    39 |     1 (100)| 00:00:01 |
|*  4 |     INDEX RANGE SCAN DESCENDING| IX_BD_CREA_DOCU_ISDE_ISCUR |     1 |    39 |     0   (0)| 00:00:01 |
PLAN_TABLE_OUTPUT
Predicate Information (identified by operation id):
   1 - filter(ROWNUM<=1000)
   4 - access("D"."ISCURRENT"=1 AND "D"."ISDELETED"=0)
17 rows selected.
SQL> In any case, post execution plan.
SY.