之前同事发过一个语句，实现的功能比較简单，相似group by的分组计数功能，由于where条件有like，又无法用group by来实现。

SELECT a.N0,b.N1,c.N2,d.N3,e.N4,f.N5,g.N6,h.N7,i.N8,j.N9 from

 (select count(*) N0 from tbl_loginfo_20141110 where keyrecord like '0%' or keyrecord like 'GJ_0%') a,

 (select count(*) N1 from tbl_loginfo_20141110 where keyrecord like '1%' or keyrecord like 'GJ_1%') b,

 (select count(*) N2 from tbl_loginfo_20141110 where keyrecord like '2%' or keyrecord like 'GJ_2%') c,

 (select count(*) N3 from tbl_loginfo_20141110 where keyrecord like '3%' or keyrecord like 'GJ_3%') d,

 (select count(*) N4 from tbl_loginfo_20141110 where keyrecord like '4%' or keyrecord like 'GJ_4%') e,

 (select count(*) N5 from tbl_loginfo_20141110 where keyrecord like '5%' or keyrecord like 'GJ_5%') f,

 (select count(*) N6 from tbl_loginfo_20141110 where keyrecord like '6%' or keyrecord like 'GJ_6%') g,

 (select count(*) N7 from tbl_loginfo_20141110 where keyrecord like '7%' or keyrecord like 'GJ_7%') h,

 (select count(*) N8 from tbl_loginfo_20141110 where keyrecord like '8%' or keyrecord like 'GJ_8%') i,

 (select count(*) N9 from tbl_loginfo_20141110 where keyrecord like '9%' or keyrecord like 'GJ_9%') j;

为了了解语句的性能，我做了例如以下相似的測试：

select * from v$version;   --Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production

drop table a;

create table a as select * from dba_objects where rownum<=50000;

begin

  for x in 1..6 loop

    insert into a select * from a;

  end loop;

  commit;

end;

select count(*) from a;   --3200000

select bytes/1024/1024 from user_segments where segment_name='A';  --357M

alter system flush shared_pool;

alter system flush buffer_cache;

SELECT a.N0,b.N1,c.N2,d.N3,e.N4,f.N5,g.N6,h.N7,i.N8,j.N9 from

 (select count(*) N0 from a where object_name like 'A%' or object_name like 'V%') a,

 (select count(*) N1 from a where object_name like 'B%' or object_name like 'V%') b,

 (select count(*) N2 from a where object_name like 'C%' or object_name like 'V%') c,

 (select count(*) N3 from a where object_name like 'D%' or object_name like 'V%') d,

 (select count(*) N4 from a where object_name like 'E%' or object_name like 'V%') e,

 (select count(*) N5 from a where object_name like 'F%' or object_name like 'V%') f,

 (select count(*) N6 from a where object_name like 'G%' or object_name like 'V%') g,

 (select count(*) N7 from a where object_name like 'H%' or object_name like 'V%') h,

 (select count(*) N8 from a where object_name like 'I%' or object_name like 'V%') i,

 (select count(*) N9 from a where object_name like 'J%' or object_name like 'V%') j;

--58s

alter system flush shared_pool;

alter system flush buffer_cache;

--改写后

select

sum(case when object_name like 'A%' or object_name like 'V%' then 1 else 0 end) N0,

sum(case when object_name like 'B%' or object_name like 'V%' then 1 else 0 end) N1,

sum(case when object_name like 'C%' or object_name like 'V%' then 1 else 0 end) N2,

sum(case when object_name like 'D%' or object_name like 'V%' then 1 else 0 end) N3,

sum(case when object_name like 'E%' or object_name like 'V%' then 1 else 0 end) N4,

sum(case when object_name like 'F%' or object_name like 'V%' then 1 else 0 end) N5,

sum(case when object_name like 'G%' or object_name like 'V%' then 1 else 0 end) N6,

sum(case when object_name like 'H%' or object_name like 'V%' then 1 else 0 end) N7,

sum(case when object_name like 'I%' or object_name like 'V%' then 1 else 0 end) N8,

sum(case when object_name like 'J%' or object_name like 'V%' then 1 else 0 end) N9

from a; 

--19s

--对照运行计划：

--前者运行计划：

SQL> explain plan for

  2  SELECT a.N0,b.N1,c.N2,d.N3,e.N4,f.N5,g.N6,h.N7,i.N8,j.N9 from

  3   (select count(*) N0 from a where object_name like 'A%' or object_name like 'V%') a,

  4   (select count(*) N1 from a where object_name like 'B%' or object_name like 'V%') b,

  5   (select count(*) N2 from a where object_name like 'C%' or object_name like 'V%') c,

  6   (select count(*) N3 from a where object_name like 'D%' or object_name like 'V%') d,

  7   (select count(*) N4 from a where object_name like 'E%' or object_name like 'V%') e,

  8   (select count(*) N5 from a where object_name like 'F%' or object_name like 'V%') f,

  9   (select count(*) N6 from a where object_name like 'G%' or object_name like 'V%') g,

 10   (select count(*) N7 from a where object_name like 'H%' or object_name like 'V%') h,

 11   (select count(*) N8 from a where object_name like 'I%' or object_name like 'V%') i,

 12   (select count(*) N9 from a where object_name like 'J%' or object_name like 'V%') j;

Explained.

Elapsed: 00:00:00.15

SQL> @getplan

'general,outline,starts'

Enter value for plan type:

PLAN_TABLE_OUTPUT

-----------------------------------------------------------------------------------------

Plan hash value: 2527411742

-------------------------------------------------------------------------------------

| Id  | Operation                    | Name | Rows  | Bytes | Cost (%CPU)| Time     |

-------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT             |      |     1 |   130 |   123K  (1)| 00:24:46 |

|   1 |  NESTED LOOPS                |      |     1 |   130 |   123K  (1)| 00:24:46 |

|   2 |   NESTED LOOPS               |      |     1 |   117 |   111K  (1)| 00:22:17 |

|   3 |    NESTED LOOPS              |      |     1 |   104 | 99032   (1)| 00:19:49 |

|   4 |     NESTED LOOPS             |      |     1 |    91 | 86653   (1)| 00:17:20 |

|   5 |      NESTED LOOPS            |      |     1 |    78 | 74274   (1)| 00:14:52 |

|   6 |       NESTED LOOPS           |      |     1 |    65 | 61895   (1)| 00:12:23 |

|   7 |        NESTED LOOPS          |      |     1 |    52 | 49516   (1)| 00:09:55 |

|   8 |         NESTED LOOPS         |      |     1 |    39 | 37137   (1)| 00:07:26 |

|   9 |          NESTED LOOPS        |      |     1 |    26 | 24758   (1)| 00:04:58 |

|  10 |           VIEW               |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  11 |            SORT AGGREGATE    |      |     1 |    66 |            |          |

|* 12 |             TABLE ACCESS FULL| A    | 91587 |  5903K| 12379   (1)| 00:02:29 |

|  13 |           VIEW               |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  14 |            SORT AGGREGATE    |      |     1 |    66 |            |          |

|* 15 |             TABLE ACCESS FULL| A    |   137K|  8831K| 12379   (1)| 00:02:29 |

|  16 |          VIEW                |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  17 |           SORT AGGREGATE     |      |     1 |    66 |            |          |

|* 18 |            TABLE ACCESS FULL | A    | 85818 |  5531K| 12379   (1)| 00:02:29 |

|  19 |         VIEW                 |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  20 |          SORT AGGREGATE      |      |     1 |    66 |            |          |

|* 21 |           TABLE ACCESS FULL  | A    |   111K|  7158K| 12379   (1)| 00:02:29 |

|  22 |        VIEW                  |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  23 |         SORT AGGREGATE       |      |     1 |    66 |            |          |

|* 24 |          TABLE ACCESS FULL   | A    | 86539 |  5577K| 12379   (1)| 00:02:29 |

|  25 |       VIEW                   |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  26 |        SORT AGGREGATE        |      |     1 |    66 |            |          |

|* 27 |         TABLE ACCESS FULL    | A    | 91587 |  5903K| 12379   (1)| 00:02:29 |

|  28 |      VIEW                    |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  29 |       SORT AGGREGATE         |      |     1 |    66 |            |          |

|* 30 |        TABLE ACCESS FULL     | A    |   228K|    14M| 12379   (1)| 00:02:29 |

|  31 |     VIEW                     |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  32 |      SORT AGGREGATE          |      |     1 |    66 |            |          |

|* 33 |       TABLE ACCESS FULL      | A    | 87981 |  5670K| 12379   (1)| 00:02:29 |

|  34 |    VIEW                      |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  35 |     SORT AGGREGATE           |      |     1 |    66 |            |          |

|* 36 |      TABLE ACCESS FULL       | A    | 84376 |  5438K| 12379   (1)| 00:02:29 |

|  37 |   VIEW                       |      |     1 |    13 | 12379   (1)| 00:02:29 |

|  38 |    SORT AGGREGATE            |      |     1 |    66 |            |          |

|* 39 |     TABLE ACCESS FULL        | A    |   112K|  7251K| 12379   (1)| 00:02:29 |

-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

---------------------------------------------------

  12 - filter("OBJECT_NAME" LIKE 'J%' OR "OBJECT_NAME" LIKE 'V%')

  15 - filter("OBJECT_NAME" LIKE 'I%' OR "OBJECT_NAME" LIKE 'V%')

  18 - filter("OBJECT_NAME" LIKE 'H%' OR "OBJECT_NAME" LIKE 'V%')

  21 - filter("OBJECT_NAME" LIKE 'G%' OR "OBJECT_NAME" LIKE 'V%')

  24 - filter("OBJECT_NAME" LIKE 'F%' OR "OBJECT_NAME" LIKE 'V%')

  27 - filter("OBJECT_NAME" LIKE 'E%' OR "OBJECT_NAME" LIKE 'V%')

  30 - filter("OBJECT_NAME" LIKE 'D%' OR "OBJECT_NAME" LIKE 'V%')

  33 - filter("OBJECT_NAME" LIKE 'C%' OR "OBJECT_NAME" LIKE 'V%')

  36 - filter("OBJECT_NAME" LIKE 'B%' OR "OBJECT_NAME" LIKE 'V%')

  39 - filter("OBJECT_NAME" LIKE 'A%' OR "OBJECT_NAME" LIKE 'V%')

--后者运行计划：

SQL> explain plan for

  2  select

  3  sum(case when object_name like 'A%' or object_name like 'V%' then 1 else 0 end) N0,

  4  sum(case when object_name like 'B%' or object_name like 'V%' then 1 else 0 end) N1,

  5  sum(case when object_name like 'C%' or object_name like 'V%' then 1 else 0 end) N2,

  6  sum(case when object_name like 'D%' or object_name like 'V%' then 1 else 0 end) N3,

  7  sum(case when object_name like 'E%' or object_name like 'V%' then 1 else 0 end) N4,

  8  sum(case when object_name like 'F%' or object_name like 'V%' then 1 else 0 end) N5,

  9  sum(case when object_name like 'G%' or object_name like 'V%' then 1 else 0 end) N6,

 10  sum(case when object_name like 'H%' or object_name like 'V%' then 1 else 0 end) N7,

 11  sum(case when object_name like 'I%' or object_name like 'V%' then 1 else 0 end) N8,

 12  sum(case when object_name like 'J%' or object_name like 'V%' then 1 else 0 end) N9

 13  from a;

Explained.

Elapsed: 00:00:00.01

SQL> @getplan

'general,outline,starts'

Enter value for plan type:

PLAN_TABLE_OUTPUT

--------------------------------------------------------------------------------------------

Plan hash value: 3918351354

---------------------------------------------------------------------------

| Id  | Operation          | Name | Rows  | Bytes | Cost (%CPU)| Time     |

---------------------------------------------------------------------------

|   0 | SELECT STATEMENT   |      |     1 |    66 | 12349   (1)| 00:02:29 |

|   1 |  SORT AGGREGATE    |      |     1 |    66 |            |          |

|   2 |   TABLE ACCESS FULL| A    |  3097K|   194M| 12349   (1)| 00:02:29 |

---------------------------------------------------------------------------

Note

-----

   - dynamic sampling used for this statement (level=2)  

能够看出，前者10次全表扫描，后者1次全表扫描。

从而时间上也大大减少了。由58s减少到19s。

优化这个sql主要还是思路的转换。难点在于如何把10次全表扫描转化成1次全表扫描。

在OLAP中，能够加并行使sql速度更快。

最新文章

  1. DDD~DDD从零起步架构说明
  2. 双节点(nginx+keepalived)为两台apache服务器提供负载均衡
  3. Android Studio项目结构
  4. flex布局滑动页面
  5. SOCKET:SO_LINGER 选项
  6. ECSide标签属性说明之&lt;ec:table&gt;
  7. linxu php连接sqlserver
  8. BZOJ 2243: [SDOI2011]染色 (树链剖分+线段树合并)
  9. php对UTF8字体串进行单字分割返回数组
  10. cocos2d-x3.1.1 step by step 学习笔记1----- 软件的安装与HelloWord
  11. Masonry 控件详解
  12. 笔记28 mssql的update :from语法
  13. Python 爬虫：把廖雪峰教程转换成 PDF 电子书
  14. php面试之四-Linux部分
  15. Vue单页面骨架屏实践
  16. CENTOS6.6下zabbix2.4.7搭建
  17. django 模型层（2）
  18. 人生就像一条加速奔向死亡的贪吃蛇【winform版】
  19. NodeJS 连接接MySQL
  20. Java.lang.IllegalStateException: Cannot call sendRedirect() after the response has been committed

热门文章

  1. [HNOI2006]超级英雄（二分+网络流）
  2. .ashx 实现自动路由和参数填充
  3. No WebApplicationContext found: no ContextLoaderListener registered?报错解决
  4. CF Mike and Feet （求连续区间内长度为i的最小值）单调栈
  5. sass02
  6. codeforces Gym100589H Count Subarrays 树状数组/线段树+离散化
  7. 是时候抛弃web.xml了？
  8. Jesse&#39;s Code
  9. python获取教务管理系统的MM照片
  10. Bootstrap Table 的用法