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:
1
2
3
4
5
|
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
1
2
3
4
5
|
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:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
|
; 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: