如何使用 Max(date) 获取记录，然后比较值以获得结果

Question

我正在尝试获取每个 MeterNumber 的 Max(StartDate) 记录，然后尝试获取特定 AccountNumber 的 startdate、enddate 和 RateCode 不同的记录。

示例数据脚本如下

create table Meter 
(
    AccountNumer varchar(50), 
    MeterNumber varchar(50), 
    StartDate date, 
    EndDate date, 
    RateCode Varchar(50)
)

Insert into Meter Values('0142628117','123470203','4/22/2020','12/31/9999','UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20')
Insert into Meter Values('0142628117','123470203','4/10/2019', '4/9/2020', '***Custom***')
Insert into Meter Values('0142628117','123470205','4/22/2020','12/31/9999','UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20')
Insert into Meter Values('0142628117','123470205','4/10/2019', '4/9/2020', '***Custom***')
    
Insert into Meter Values('0500000178767001363445','TCA105238304','02/25/2016','04/22/2016', '***Custom***')
Insert into Meter Values('0500000178767001363445','TCA105238304','10/2/2018','08/11/2019', '***Custom***')
Insert into Meter Values('0500000178767001363445','TCA130359929','8/12/2019','12/31/9999', '***Custom***')

RowNo   AccountNumer            MeterNumber     StartDate   EndDate     RateCode
    1   0142628117              123470203       2020-04-22  9999-12-31  UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20
    2   0142628117              123470203       2019-04-10  2020-04-09  ***Custom***
    3   0142628117              123470205       2020-04-22  9999-12-31  UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20
    4   0142628117              123470205       2019-04-10  2020-04-09  ***Custom***
    5   0500000178767001363445  TCA105238304    2016-02-25  2016-04-22  ***Custom***
    6   0500000178767001363445  TCA105238304    2018-10-02  2019-08-11  ***Custom***
    7   0500000178767001363445  TCA130359929    2019-08-12  9999-12-31  ***Custom***

首先，我需要为每个MeterNumber 找到Max(StartDate)，以获取特定的AccountNumber。输出应该是这样的：

    RowNo   AccountNumer            MeterNumber        StartDate   EndDate     RateCode
        1   0142628117              123470203       2020-04-22  9999-12-31  UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20
        3   0142628117              123470205       2020-04-22  9999-12-31  UGE-PECO-E-R-V-GRN-HOLDOVER-ETF0_APR20
        6   0500000178767001363445  TCA105238304    2018-10-02  2019-08-11  ***Custom***
        7   0500000178767001363445  TCA130359929    2019-08-12  9999-12-31  ***Custom***

然后我试图从（两行）获取帐号的开始日期、结束日期和费率代码不同的记录。所以我期望的最终结果如下表所示。

   RowNo   AccountNumer            MeterNumber        StartDate   EndDate     RateCode
       6    0500000178767001363445  TCA105238304    2018-10-02  2019-08-11  ***Custom***
       7    0500000178767001363445  TCA130359929    2019-08-12  9999-12-31  ***Custom***

任何帮助将不胜感激！

Answer 1

这是一个 3 步过程，首先使用 ROW_NUMBER() 对每个 Account/Meter 组合的记录进行排名

SELECT *,
      RowNumber = ROW_NUMBER() OVER(PARTITION BY AccountNumber, MeterNumber 
                                    ORDER BY EndDate DESC)
FROM Meter

输出

AccountNumber	MeterNumber	StartDate	EndDate	RateCode	RowNumber
0142628117	123470203	2020-04-22	9999-12-31	ETF0_APR20	1
0142628117	123470203	2019-04-10	2020-04-09	***Custom***	2
0142628117	123470205	2020-04-22	9999-12-31	ETF0_APR20	1
0142628117	123470205	2019-04-10	2020-04-09	***Custom***	2
1363445	105238304	2018-10-02	2019-08-11	***Custom***	1
1363445	105238304	2016-02-25	2016-04-22	***Custom***	2
1363445	130359929	2019-08-12	9999-12-31	***Custom***	1

_{注意一些数据被缩短以更好地显示}

然后您可以过滤 RowNumber = 1 以获得每个仪表的最新结束日期。

接下来您需要计算不同的 EndDate/RateCode 组合，您不能在窗口函数中使用 COUNT(DISTINCT ...)，但是您可以使用 DENSE_RANK() 来模拟：

SELECT *,
        CntDistinct = DENSE_RANK() OVER(PARTITION BY AccountNumber 
                                                ORDER BY EndDate, RateCode) 
                        + DENSE_RANK() OVER(PARTITION BY AccountNumber 
                                            ORDER BY EndDate DESC, RateCode DESC) - 1
FROM (  SELECT *,
                RowNumber = ROW_NUMBER() OVER(PARTITION BY AccountNumber, MeterNumber 
                                                ORDER BY EndDate DESC)
        FROM Meter AS m
    ) AS m
WHERE m.RowNumber = 1;

输出

AccountNumber	MeterNumber	StartDate	EndDate	RateCode	CntDistinct
0142628117	123470203	2020-04-22	9999-12-31	ETF0_APR20	1
0142628117	123470205	2020-04-22	9999-12-31	ETF0_APR20	1
1363445	130359929	2019-08-12	9999-12-31	***Custom***	2
1363445	105238304	2018-10-02	2019-08-11	***Custom***	2

_{注意一些数据被缩短以更好地显示}

最后将所有这些放入进一步的子查询中，并限制在 EndDate/RateCode 的唯一组合不止一个的地方：

SELECT AccountNumber, MeterNumber, StartDate, EndDate, RateCode
FROM (  SELECT *,
                CntDistinct = DENSE_RANK() OVER(PARTITION BY AccountNumber 
                                                    ORDER BY EndDate, RateCode) 
                                + DENSE_RANK() OVER(PARTITION BY AccountNumber 
                                                    ORDER BY EndDate DESC, RateCode DESC) - 1
            FROM (  SELECT *,
                            RowNumber = ROW_NUMBER() OVER(PARTITION BY AccountNumber, MeterNumber 
                                                        ORDER BY EndDate DESC)
                    FROM Meter AS m
                ) AS m
            WHERE m.RowNumber = 1
    ) AS m
WHERE m.CntDistinct > 1;

输出

AccountNumber	MeterNumber	StartDate	EndDate	RateCode
0500000178767001363445	TCA105238304	2018-10-02	2019-08-11	***Custom***
0500000178767001363445	TCA130359929	2019-08-12	9999-12-31	***Custom***

Example on db<>Fiddle