站长资源数据库

SQL字段拆分优化

整理:jimmy2024/12/23浏览2
简介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.fsnFROM so_orderINNER JOIN so_orderitem ON CHARINDEX(so_Orderitem.fid, SO_Order.fi

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

SQL字段拆分优化

  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

SQL字段拆分优化

  改写后SQL的执行计划如下:

SQL字段拆分优化