SQL查詢以查找group by子句中所有行的總和以及行的部分總和

Question

我正在嘗試為SQL Server 2008數據庫建立查詢。 該查詢將對SID和Date列進行分組，並對Profit列求和。 該查詢還需要在一天結束時找到具有相同單位的所有行的利潤總和。 在這種情況下，單位值1從下午12:54的-1切換為1，並保持不變，直到當天的最后一條記錄為下午1:15。

SID Date               Profit    Units   RowID
-------------------------------------------------------
1   7/26/10 1:15 PM     -25          1        1
1   7/26/10 12:54 PM    -12.5        1        2
1   7/26/10 12:54 PM    0            0        3
1   7/26/10 12:54 PM    -125         -1       4
1   7/26/10 12:00 PM    100          -1       5
1   7/26/10 10:31 AM    -50          -1       6

單位為0的行表示發生了切換。 在第三行中，單位並不總是變為0。 例如，

SID Date             Profit     Units
--------------------------------------------------------
1   7/26/10 2:15 PM    -37.5        -1
1   7/26/10 2:06 PM    -125         -1
1   7/26/10 2:00 PM    -12.5        -1
1   7/26/10 2:00 PM    0             0
1   7/26/10 2:00 PM    -75           1
1   7/26/10 12:45 PM   -12.5         1
1   7/26/10 12:45 PM    0            0
1   7/26/10 12:45 PM   -125         -1
1   7/26/10 12:00 PM    100         -1
1   7/26/10 9:55 AM     -50         -1

第一個數據集的輸出如下所示：

SID      Date     TotalProfit    ProfitforEndofDayUnits 
-------------------------------------------------------
1        7/26/10      -112.5                  -37.5

謝謝。

Answer 1

declare @T table( 
  [SID] int,
  [Date] datetime,
  Profit money,
  Units int)

insert into @T values
(1,   '7/26/10 2:15 PM',    -37.5,        -1),
(1,   '7/26/10 2:06 PM',    -125,         -1),
(1,   '7/26/10 2:00 PM',    -12.5,        -1),
(1,   '7/26/10 2:00 PM',    0,             0),
(1,   '7/26/10 2:00 PM',    -75,           1),
(1,   '7/26/10 12:45 PM',   -12.5,         1),
(1,   '7/26/10 12:45 PM',    0,            0),
(1,   '7/26/10 12:45 PM',   -125,         -1),
(1,   '7/26/10 12:00 PM',    100,         -1),
(1,   '7/26/10 9:55 AM',     -50,         -1),
(2,   '7/26/10 1:15 PM',     -25,          1),
(2,   '7/26/10 12:54 PM',    -12.5,        1),
(2,   '7/26/10 12:54 PM',    0,            0),
(2,   '7/26/10 12:54 PM',    -125,         -1),
(2,   '7/26/10 12:00 PM',    100,          -1),
(2,   '7/26/10 10:31 AM',    -50,          -1)


;with cte as
(
  select
    [SID],
    cast([Date] as Date) as [Date],
    Profit,
    Units,
    row_number() over(partition by SID, cast([Date] as Date) 
                      order by [Date] desc) as rn
  from @T
)
select 
  C1.[SID],
  C1.[Date],
  sum(C1.Profit) as TotalProfit,
  sum(case when C2.Units = C1.Units and
                C1.rn < S.rn 
        then C1.Profit
        else 0.0
      end) ProfitforEndofDayUnits
from cte as C1
  inner join cte as C2
    on C1.[SID] = C2.[SID] and
       C1.[Date] = C2.[Date] and
       C2.rn = 1
  cross apply
    (select min(C3.rn) as rn
     from cte as C3
     where C1.[SID] = C3.[SID] and
           C1.[Date] = C3.[Date] and
           C3.Units <> C2.Units) as S      
group by C1.[SID], C1.[Date]

結果：

SID         Date       TotalProfit           ProfitforEndofDayUnits
----------- ---------- --------------------- -----------------------
1           2010-07-26 -337,50               -175.0000
2           2010-07-26 -112,50               -37.5000