oracle bitmap and /and equal

bitmap and 还是 and equal，都是一种将两个索引选择出来的rowid合并的算法，可以理解索引扫描结果合并;

假设这里的条件使用这两个索引选择性也非常之高，那么这里使用bitmap是不会影响效率的。

我们的sql 5分钟没有跑完，查看计划，走了BITMAP CONVERSION TO ROWIDS。

SQL> explain plan for
2     select    a.username ,
3  a.membertype ,
4         a.providername ,
5         a.tradename ,
6         a.actor ,
7         a.telephone ,
8         a.mobilephone  ,
9         a.email ,
10         b.title ,
11         b.starttime ,
12         b.adminname
13    from Seller000 a, Chatroom000 b
14   where a.roomid = b.roomid
15     and b.isshow = '1'
16     and b.roomtype in ('0', '1')
17     and a.signupdate > to_date('2012-08-01', 'yyyy-mm-dd hh24:mi:ss')
18     and a.signupdate < to_date('2012-09-01', 'yyyy-mm-dd hh24:mi:ss')
19   order by b.roomid;

已解释。

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------

Plan hash value: 273647176

----------------------------------------------------------------------------------------------------------
| Id  | Operation                          | Name                | Rows  | Bytes | Cost (%CPU)| Time  |
----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                   |                     |    11 |  1782 |    28  (11)| 00:00:01 |
|*  1 |  FILTER                            |                     |       |       |            |       |
|   2 |   TABLE ACCESS BY INDEX ROWID      | SELLER              |     1 |   112 |    28  (11)| 00:00:01 |
|   3 |    NESTED LOOPS                    |                     |    11 |  1782 |    28  (11)| 00:00:01 |
|*  4 |     TABLE ACCESS BY INDEX ROWID    | CHATROOM            | 10975 |   535K|     3   (0)| 00:00:01 |
|   5 |      INDEX FULL SCAN               | PK_CHATROOM         |     9 |       |     1   (0)| 00:00:01 |

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------

|   6 |     BITMAP CONVERSION TO ROWIDS    |                     |       |       |            |       |
|   7 |      BITMAP AND                    |                     |       |       |            |       |
|   8 |       BITMAP CONVERSION FROM ROWIDS|                     |       |       |            |       |
|*  9 |        INDEX RANGE SCAN            | IDX_SELLER_ROOMID   |   141 |       |     1   (0)| 00:00:01 |
|  10 |       BITMAP CONVERSION FROM ROWIDS|                     |       |       |            |       |
|  11 |        SORT ORDER BY               |                     |       |       |            |       |
|* 12 |         INDEX RANGE SCAN           | I_SELLER_SIGNUPDATE |   141 |       |     2   (0)| 00:00:01 |
----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------

1 - filter(TO_DATE('2012-08-01','yyyy-mm-dd hh24:mi:ss')<TO_DATE('2012-09-01','yyyy-mm-dd
hh24:mi:ss'))
4 - filter("B"."ISSHOW"='1' AND ("B"."ROOMTYPE"='0' OR "B"."ROOMTYPE"='1'))
9 - access("A"."ROOMID"="B"."ROOMID")
12 - access("A"."SIGNUPDATE">TO_DATE('2012-08-01','yyyy-mm-dd hh24:mi:ss') AND
"A"."SIGNUPDATE"<TO_DATE('2012-09-01','yyyy-mm-dd hh24:mi:ss'))

已选择29行。

SQL> explain plan for
2    select  /*+ opt_param('_b_tree_bitmap_plans', 'false') */   a.username ,
3  a.membertype ,
4         a.providername ,
5         a.tradename ,
6         a.actor ,
7         a.telephone ,
8         a.mobilephone  ,
9         a.email ,
10         b.title ,
11         b.starttime ,
12         b.adminname
13    from Seller000 a, Chatroom000 b
14   where a.roomid = b.roomid
15     and b.isshow = '1'
16     and b.roomtype in ('0', '1')
17     and a.signupdate > to_date('2012-08-01', 'yyyy-mm-dd hh24:mi:ss')
18     and a.signupdate < to_date('2012-09-01', 'yyyy-mm-dd hh24:mi:ss')
19   order by b.roomid;

已解释。

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------------------------------------

Plan hash value: 4238594111

----------------------------------------------------------------------------------------------------
| Id  | Operation                      | Name              | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT               |                   |    11 |  1782 |   240   (0)| 00:00:03 |
|*  1 |  FILTER                        |                   |       |       |            |          |
|*  2 |   TABLE ACCESS BY INDEX ROWID  | SELLER            |     1 |   112 |    34   (0)| 00:00:01 |
|   3 |    NESTED LOOPS                |                   |    11 |  1782 |   240   (0)| 00:00:03 |
|*  4 |     TABLE ACCESS BY INDEX ROWID| CHATROOM          | 10975 |   535K|     3   (0)| 00:00:01 |
|   5 |      INDEX FULL SCAN           | PK_CHATROOM       |     9 |       |     1   (0)| 00:00:01 |

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------------------------------------

|*  6 |     INDEX RANGE SCAN           | IDX_SELLER_ROOMID |   141 |       |     1   (0)| 00:00:01 |
----------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - filter(TO_DATE('2012-08-01','yyyy-mm-dd
hh24:mi:ss')<TO_DATE('2012-09-01','yyyy-mm-dd hh24:mi:ss'))
2 - filter("A"."SIGNUPDATE">TO_DATE('2012-08-01','yyyy-mm-dd hh24:mi:ss') AND
"A"."SIGNUPDATE"<TO_DATE('2012-09-01','yyyy-mm-dd hh24:mi:ss'))
4 - filter("B"."ISSHOW"='1' AND ("B"."ROOMTYPE"='0' OR "B"."ROOMTYPE"='1'))

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------------------------------------

6 - access("A"."ROOMID"="B"."ROOMID")

已选择23行。

出现这样的情况，是因为表中存在不适当的索引，这些索引列的唯一度不高，oracle就有可能选择两个这样的索引转为bitmap来执行，

根据这两个索引的值再确认共同有的ROWID，最后再通过ROWID回表提取符合条件的数据。

可以使用/*+ opt_param('_b_tree_bitmap_plans', 'false') */ hint在sql级消除bitmap。

最终解决方法，删除唯一度低的index，建立组合index。

删除索引IDX_SELLER_ROOMID，如果I_SELLER_SIGNUPDATE索引确认无用，也可以删除，IND_ROOMID_SIGNUPDATE是我们建立的组合索引，这样消除了BITMAP CONVERSION TO ROWIDS，30秒出结果。以下是新的执行计划：

SQL> explain plan for
2    select   a.username ,
3  a.membertype ,
4         a.providername ,
5         a.tradename ,
6         a.actor ,
7         a.telephone ,
8         a.mobilephone  ,
9         a.email ,
10         b.title ,
11         b.starttime ,
12         b.adminname
13    from Seller000 a, Chatroom000 b
14   where a.roomid = b.roomid
15     and b.isshow = '1'
16     and b.roomtype in ('0', '1')
17     and a.signupdate > to_date('2012-08-01', 'yyyy-mm-dd hh24:mi:ss')
18     and a.signupdate < to_date('2012-09-01', 'yyyy-mm-dd hh24:mi:ss')
19   order by b.roomid;

已解释。

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------

Plan hash value: 2603971898

--------------------------------------------------------------------------------------------------------
| Id  | Operation                      | Name                  | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT               |                       |    11 |  1672 |     9   (0)| 00:00:01 |
|*  1 |  FILTER                        |                       |       |       |            |          |
|   2 |   TABLE ACCESS BY INDEX ROWID  | SELLER                |     1 |   102 |     1   (0)| 00:00:01 |
|   3 |    NESTED LOOPS                |                       |    11 |  1672 |     9   (0)| 00:00:01 |
|*  4 |     TABLE ACCESS BY INDEX ROWID| CHATROOM              |  9978 |   487K|     3   (0)| 00:00:01 |
|   5 |      INDEX FULL SCAN           | PK_CHATROOM           |     9 |       |     1   (0)| 00:00:01 |

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------

|*  6 |     INDEX RANGE SCAN           | IND_ROOMID_SIGNUPDATE |     1 |       |     1   (0)| 00:00:01 |
--------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

1 - filter(TO_DATE('2012-08-01','yyyy-mm-dd hh24:mi:ss')<TO_DATE('2012-09-01','yyyy-mm-dd
hh24:mi:ss'))
4 - filter("B"."ISSHOW"='1' AND ("B"."ROOMTYPE"='0' OR "B"."ROOMTYPE"='1'))
6 - access("A"."ROOMID"="B"."ROOMID" AND "A"."SIGNUPDATE">TO_DATE('2012-08-01','yyyy-mm-dd
hh24:mi:ss') AND "A"."SIGNUPDATE"<TO_DATE('2012-09-01','yyyy-mm-dd hh24:mi:ss'))

已选择22行。