Question

我在同一张桌子上有2个查询，我预计会有类似的表现，但我仍然会有很大的差异 - 30秒对比~1秒。

查询1：为当月计算用户数

b()

查询2：计算本周的用户数

SELECT datepart(year,getdate()) as TheYear, 
       DATEPART(month, GETDATE()) AS TheMonth, 
       COUNT(DISTINCT Username) AS TheUsers
FROM dbo.[AP-Data] 
WHERE datepart(year,RequestDate) =  datepart(year,getdate()) 
  AND datepart(month,RequestDate) = datepart(month,getdate())

该表有大约350万条记录，并由SELECT datepart(year,getdate()) as TheYear, DATEPART(week, GETDATE()) AS TheWeek, COUNT(DISTINCT Username) AS TheUsers FROM dbo.[AP-Data] WHERE datepart(year,RequestDate) = datepart(year,getdate()) AND datepart(week,RequestDate) = datepart(week,getdate())编制索引。本月的计数正在进行全表扫描，因此需要30秒，本周的一次是在~1秒内进行RID查询 - 此信息来自执行计划。

知道为什么会有差异吗？

Answer 1

尝试这些查询怎么样？

SELECT year(getdate()) as TheYear, month(GETDATE()) AS TheMonth,
       COUNT(DISTINCT Username) AS TheUsers
FROM dbo.[AP-Data]
WHERE RequestDate >= dateadd(month, datediff(month, 0, getdate()), 0);

和

SELECT year(getdate()) as TheYear, month(GETDATE()) AS TheMonth,
       COUNT(DISTINCT Username) AS TheUsers
FROM dbo.[AP-Data]
WHERE RequestDate >= dateadd(week, datediff(week, 0, getdate()), 0);

在列上使用函数通常会阻止编译器使用索引。以上假设您没有将来的请求日期。

SQL Server：类似查询的不同执行计划和性能

1 个答案: