SQL字段拆分优化
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：
