通过拆分字段优化SQL

数据库环境:SQL SERVER 2008R2

  今天看到一条用函数处理连接的SQL,是群里某位网友的,SQL语句如下:

SELECT  SO_Order.fdate ,
        SO_Order.fsn
FROM    so_order
        INNER JOIN so_orderitem ON CHARINDEX(so_Orderitem.fid, SO_Order.fid) >= 1
WHERE   so_order.FOrderDate = '2015-09-06'
View Code

  语句不算复杂,只是执行比较慢,下面是关于这SQL语句的一些信息:

--1.SQL执行203条数据
--2.耗时12秒
--3.so_order表的fid字段是字符串集合, --由1-2个字符串组成,用','分隔 SELECT COUNT(*) FROM SO_Order WHERE so_order.FOrderDate = '2015-09-06'--24 SELECT COUNT(*) FROM so_Orderitem--414154

  执行计划

  通过分析执行计划,so_order和so_orderitem走嵌套循环是正确的,查询的瓶颈是在so_orderitem的索引

扫描上。因此,我们要通过改写SQL,达到so_orderitem走索引查找的目的。

  将so_order表的fid字段拆分成2个字段,然后union all成一个结果集,再和so_orderitem关联,即可让

so_orderitem表走索引查找。改写的SQL如下:

;WITH    x0
          AS ( SELECT   fdate ,
                        fsn ,
                        LEFT(fid, CASE WHEN CHARINDEX(',', fid, 1) = 0 THEN 0
                                       ELSE CHARINDEX(',', fid, 1) - 1
                                  END) AS fid
               FROM     so_order
               WHERE    FOrderDate = '2015-09-06'
               UNION ALL
               SELECT   fdate ,
                        fsn ,
                        RIGHT(fid, LEN(fid) - CHARINDEX(',', fid, 1)) AS fid
               FROM     so_order
               WHERE    FOrderDate = '2015-09-06'
             )
    SELECT  SO_Order.fdate ,
            SO_Order.fsn
    FROM    x0 so_order
            INNER JOIN so_orderitem ON so_Orderitem.fid = SO_Order.fid
View Code

  下面是SQL改写后,网友的反馈截图

  改写后SQL的执行计划如下:

posted on 2015-09-06 18:55  ToBeHJH  阅读(1357)  评论(4编辑  收藏  举报

导航