所以我在使用AdventureWorks 2012进行在线帮助教程后，正在使用SQL窗口函数。
我从以下查询中选择了数据到新表
select *,
MIN(Amount) OVER(Partition by AccountID order by transactionDate) AS MinOrderForDate
from dbo.transactions
Order by AccountID, transactionID desc
然后我运行了以下查询：
AccountID TransactionID TransactionDate Amount MinOrderForDate
11000 63804 2007-11-04 00:00:00.000 53.990000 4.990000
11000 63803 2007-11-04 00:00:00.000 34.990000 4.990000
11000 63802 2007-11-04 00:00:00.000 4.990000 4.990000
11000 63801 2007-11-04 00:00:00.000 28.990000 4.990000
11000 63800 2007-11-04 00:00:00.000 2384.070000 4.990000
11000 38716 2007-07-22 00:00:00.000 21.980000 21.980000
11000 38715 2007-07-22 00:00:00.000 2319.990000 21.980000
11000 449 2005-07-22 00:00:00.000 3399.990000 3399.990000
11001 115673 2008-06-12 00:00:00.000 34.990000 4.990000
11001 115672 2008-06-12 00:00:00.000 8.990000 4.990000
11001 115671 2008-06-12 00:00:00.000 4.990000 4.990000
11001 115670 2008-06-12 00:00:00.000 539.990000 4.990000
11001 38639 2007-07-20 00:00:00.000 8.990000 4.990000
11001 38638 2007-07-20 00:00:00.000 53.990000 4.990000
11001 38637 2007-07-20 00:00:00.000 9.990000 4.990000
11001 38636 2007-07-20 00:00:00.000 4.990000 4.990000
11001 38635 2007-07-20 00:00:00.000 21.980000 4.990000
11001 38634 2007-07-20 00:00:00.000 2319.990000 4.990000
11001 423 2005-07-18 00:00:00.000 3374.990000 3374.990000
这给了我预期的结果：
select *,
MAX(Amount) OVER(Partition by AccountID order by transactionDate) AS MaxOrderForDate
from dbo.transactions
Order by AccountID, transactionID desc
AccountID TransactionID TransactionDate Amount MaxOrderForDate
11000 63804 2007-11-04 00:00:00.000 53.990000 3399.990000
11000 63803 2007-11-04 00:00:00.000 34.990000 3399.990000
11000 63802 2007-11-04 00:00:00.000 4.990000 3399.990000
11000 63801 2007-11-04 00:00:00.000 28.990000 3399.990000
11000 63800 2007-11-04 00:00:00.000 2384.070000 3399.990000
11000 38716 2007-07-22 00:00:00.000 21.980000 3399.990000
11000 38715 2007-07-22 00:00:00.000 2319.990000 3399.990000
11000 449 2005-07-22 00:00:00.000 3399.990000 3399.990000
11001 115673 2008-06-12 00:00:00.000 34.990000 3374.990000
11001 115672 2008-06-12 00:00:00.000 8.990000 3374.990000
11001 115671 2008-06-12 00:00:00.000 4.990000 3374.990000
11001 115670 2008-06-12 00:00:00.000 539.990000 3374.990000
11001 38639 2007-07-20 00:00:00.000 8.990000 3374.990000
11001 38638 2007-07-20 00:00:00.000 53.990000 3374.990000
11001 38637 2007-07-20 00:00:00.000 9.990000 3374.990000
11001 38636 2007-07-20 00:00:00.000 4.990000 3374.990000
11001 38635 2007-07-20 00:00:00.000 21.980000 3374.990000
11001 38634 2007-07-20 00:00:00.000 2319.990000 3374.990000
11001 423 2005-07-18 00:00:00.000 3374.990000 3374.990000
但是当我用MAX替换MIN时，我会继续在帐户的整个数据范围内获取MAX值，我不明白为什么？
load_and_authorize_resource我这样做错了吗？
答案 0 :(得分：2)
如果您使用非唯一列进行排序，则无法保证结果集。
在您的情况下，结果正是您所要求的，累积的MIN / MAX，由于ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW，您的查询会自动添加ORDER BY。此外，窗口化聚合函数中的ORDER BY与最终ORDER BY不同，这会增加混淆。
您没有明确说明您想要获得的内容。根据列别名，您可能需要一个“组”MIN / MAX：
MIN(Amount) OVER(Partition by AccountID, transactionDate) AS MinOrderForDate
答案 1 :(得分：1)
您对自己的意图并不十分明确，但听起来您希望每个min/max分组都account/transactionDate。如果是这种情况，您可能并不意味着将transactionDate放在窗口函数的order by子句中。
我认为你的意思是这样做：
MIN(Amount) OVER(Partition by AccountID, transactionDate) AS MinOrderForDate
MAX(Amount) OVER(Partition by AccountID, transactionDate) AS MaxOrderForDate
答案 2 :(得分：0)
您正在通过AccountID进行分区。它为您提供了该AccountID的MAX。对于AccountID 11000，MAX为3399，对于AccountID 11001，为3374。