SQL有外连接的时候注意过滤条件位置

2013年/5月/2日

奶奶的，为啥现在五一节只放3天，5月的天气最适合出游了，不过俺们这些苦逼的IT男是没法享受了。

一来到公司，项目经理就找到开发leader，说我们网站 页面很慢，让他排查原因。

一听说 网站慢，页面慢 哥 就来精神了，哥的老本行就是 解决 “慢” 的问题。

开发leader 很郁闷的说，我们已经加了 memcache了，20分钟 cache一次，咋个还是慢呢，

于是哥就问，那个网页跑了哪些SQL？ 能抓出来让我看看吗？ 开发Leader 果断的把SQL 抓了出来。

经过排查，我们发现了一个SQL确实跑得慢。该SQL 如下

select *
  from (select u.NAME UniversityName,
               u.id UniversityId,
               count(a.SIGNUPNUMBER) playercnt
          from T_B_UNIVERSITY u
          left join T_D_EDUCATION e
            on e.UNIVERSITY_ID = u.id
          left join T_D_VIDEO_PLAYER a
            on a.USER_ID = e.user_id
           and e.ISDEFAULT = 1
           and e.ISVALID = 1
           and a.AUDITSTATUS = 1
           and a.ISVALID = 1
          left join T_D_USER c
            on a.USER_ID = c.id
           and c.ISVALID = 1
         where u.REGION_CODE like '43%'
         group by u.NAME, u.id)
 order by playercnt desc;

执行计划如下

执行计划
----------------------------------------------------------
Plan hash value: 3938743742

--------------------------------------------------------------------------------------------
| Id  | Operation               | Name             | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT        |                  |   142 | 10366 |   170   (3)| 00:00:03 |
|   1 |  SORT ORDER BY          |                  |   142 | 10366 |   170   (3)| 00:00:03 |
|   2 |   HASH GROUP BY         |                  |   142 | 10366 |   170   (3)| 00:00:03 |
|*  3 |    HASH JOIN RIGHT OUTER|                  |   672 | 49056 |   168   (2)| 00:00:03 |
|*  4 |     TABLE ACCESS FULL   | T_D_USER         |   690 |  5520 |     5   (0)| 00:00:01 |
|   5 |     NESTED LOOPS OUTER  |                  |   672 | 43680 |   162   (1)| 00:00:02 |
|*  6 |      HASH JOIN OUTER    |                  |   672 | 37632 |    14   (8)| 00:00:01 |
|*  7 |       TABLE ACCESS FULL | T_B_UNIVERSITY   |    50 |  2050 |     8   (0)| 00:00:01 |
|   8 |       TABLE ACCESS FULL | T_D_EDUCATION    |   672 | 10080 |     5   (0)| 00:00:01 |
|   9 |      VIEW               |                  |     1 |     9 |     0   (0)| 00:00:01 |
|* 10 |       FILTER            |                  |       |       |            |          |
|* 11 |        TABLE ACCESS FULL| T_D_VIDEO_PLAYER |     1 |    15 |     3   (0)| 00:00:01 |
--------------------------------------------------------------------------------------------

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

   3 - access("A"."USER_ID"="C"."ID"(+))
   4 - filter("C"."ISVALID"(+)=1)
   6 - access("E"."UNIVERSITY_ID"(+)="U"."ID")
   7 - filter("U"."REGION_CODE" LIKE '43%')
  10 - filter("E"."ISVALID"=1 AND "E"."ISDEFAULT"=1)
  11 - filter("A"."USER_ID"="E"."USER_ID" AND "A"."AUDITSTATUS"=1 AND
              "A"."ISVALID"=1)

大家能发现这个SQL 的问题吗？ 这个 SQL 之所以跑得慢是因为开发人员把SQL的条件写错位置了

正确的写法应该是 下面这样的

select *
  from (select u.NAME UniversityName,
               u.id UniversityId,
               count(a.SIGNUPNUMBER) playercnt
          from T_B_UNIVERSITY u
          left join T_D_EDUCATION e
            on e.UNIVERSITY_ID = u.id
           and e.ISDEFAULT = 1
           and e.ISVALID = 1
          left join T_D_VIDEO_PLAYER a
            on a.USER_ID = e.user_id    
           and a.AUDITSTATUS = 1
           and a.ISVALID = 1
          left join T_D_USER c
            on a.USER_ID = c.id
           and c.ISVALID = 1
         where u.REGION_CODE like '43%'
         group by u.NAME, u.id)
 order by playercnt desc; 

执行计划如下

执行计划
----------------------------------------------------------
Plan hash value: 2738827747

---------------------------------------------------------------------------------------------
| Id  | Operation                | Name             | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT         |                  |   142 | 11218 |    25  (16)| 00:00:01 |
|   1 |  SORT ORDER BY           |                  |   142 | 11218 |    25  (16)| 00:00:01 |
|   2 |   HASH GROUP BY          |                  |   142 | 11218 |    25  (16)| 00:00:01 |
|*  3 |    HASH JOIN RIGHT OUTER |                  |   301 | 23779 |    23   (9)| 00:00:01 |
|*  4 |     TABLE ACCESS FULL    | T_D_USER         |   690 |  5520 |     5   (0)| 00:00:01 |
|*  5 |     HASH JOIN RIGHT OUTER|                  |   301 | 21371 |    17   (6)| 00:00:01 |
|*  6 |      TABLE ACCESS FULL   | T_D_VIDEO_PLAYER |    78 |  1170 |     3   (0)| 00:00:01 |
|*  7 |      HASH JOIN OUTER     |                  |   301 | 16856 |    14   (8)| 00:00:01 |
|*  8 |       TABLE ACCESS FULL  | T_B_UNIVERSITY   |    50 |  2050 |     8   (0)| 00:00:01 |
|*  9 |       TABLE ACCESS FULL  | T_D_EDUCATION    |   301 |  4515 |     5   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------

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

   3 - access("A"."USER_ID"="C"."ID"(+))
   4 - filter("C"."ISVALID"(+)=1)
   5 - access("A"."USER_ID"(+)="E"."USER_ID")
   6 - filter("A"."AUDITSTATUS"(+)=1 AND "A"."ISVALID"(+)=1)
   7 - access("E"."UNIVERSITY_ID"(+)="U"."ID")
   8 - filter("U"."REGION_CODE" LIKE '43%')
   9 - filter("E"."ISDEFAULT"(+)=1 AND "E"."ISVALID"(+)=1)

之前SQL要跑至少5秒以上，现在0.1秒能出结果。

各位童鞋，SQL 有外连接的时候，要注意过滤条件的位置，记住啦！！！

 有SQL 需要优化的 欢迎加入  QQ 群 220761024 申请注明 来自CSDN