SQL字段拆分优化
内容摘要
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_Orde
SELECT SO_Order.fdate ,
SO_Order.fsn
FROM so_order
INNER JOIN so_orderitem ON CHARINDEX(so_Orde
文章正文
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'
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
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
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:
代码注释