Question

当我在查询中间使用字符串作为日期时，我在查询中看到性能下降（速度慢2倍-3x），而不是先将字符串值放入类型为smalldatetime的变量中。所以例如

declare @rptDate AS smalldatetime
set @rtpDate = CAST('2014-06-22' AS smalldatetime)
SELECT SUM(tons) FROM sales
 WHERE ItemClass IN (1, 3, 5, 7) 
   AND AcctNo LIKE '31_-30[12]0'
   AND YEAR(InvDate) = YEAR(@rptDate) 
   AND MONTH(InvDate) = MONTH(@rptDate) '
   AND DAY(InvDate) <= DAY(@rptDate)

肯定比

更快

SELECT SUM(tons) FROM sales
 WHERE ItemClass IN (1, 3, 5, 7) 
   AND AcctNo LIKE '31_-30[12]0'
   AND YEAR(InvDate) = YEAR('2014-06-22') 
   AND MONTH(InvDate) = MONTH('2014-06-22') 
   AND DAY(InvDate) <= DAY('2014-06-22')

那么为什么“正确的优化”不会发生在这里呢？

Answer 1

我会再往前走了

declare @rptDate AS smalldatetime
declare @year as int 
declare @month as int
declare @day as int
set @rtpDate = CAST('2014-06-22' AS smalldatetime)
set @year  = year(@rtpDate)
set @month = month(@rtpDate)
set @day   = day(@rtpDate)

SELECT SUM(tons) FROM sales
 WHERE ItemClass IN (1, 3, 5, 7) 
   AND AcctNo LIKE '31_-30[12]0'
   AND YEAR(InvDate)  = @year 
   AND MONTH(InvDate) = @month
   AND DAY(InvDate)  <= @day

不首先转换为datetime时性能显着下降

1 个答案: