我有一张付款表，有正值和负值（即捕获和信用）。我需要确定自上次净正数以来我们收到净正数的点数。例如，如果客户进行这些付款并收到这些信用：
01/01 $100 <-
02/01 -$100
03/01 -$100
04/01 $100
05/01 $100
06/01 $100 <-
......那么积分将是01/01和06/01：从02/01到04/01，他们有负余额，截至05/01他们的余额为零。
我目前的方法是首先使用捕获建立所有日期的结束日期列表，然后计算每个日期的开始日期，最后计算这些期间的净捕获：
Start End NetCaptures
1900/01/01 2011/01/01 $100
2011/01/02 2011/04/01 -$100
2011/04/02 2011/05/01 $100
2011/05/02 2011/06/01 $100
然后我丢弃NetCaptures为$ 0或更少的记录，重新计算开始日期，重新计算网络捕获，并重复直到没有要删除的记录，留下这个。
Start End NetCaptures
1900/01/01 2011/01/01 $100
2011/01/02 2011/06/01 $100
有更好的方法吗？有些聪明地使用分析表达式？这已接近RBAR。在实践中，它运行速度可以接受（500K记录为10分钟，而在我开始以这种方式计算学分之前为1.5）。
*结果 *
虽然Microsoft确实支持优雅的滚动总功能，但是使用这个想法我最终得到了这样的代码：计算所有捕获，计算每个捕获的运行总计，并丢弃那些具有相同的早期记录的那些或者更高的总计。
CREATE TABLE #Sequences
(
OrderID INT NOT NULL,
Sequence INT NOT NULL,
PRIMARY KEY (OrderID, Sequence),
StartDate DATE NOT NULL DEFAULT '1900-01-01',
EndDate DATE NOT NULL,
CapturesThisPeriod DECIMAL(18, 2) NOT NULL DEFAULT 0.00,
)
INSERT INTO #Sequences (OrderID, Sequence, EndDate)
SELECT OrderID, ROW_NUMBER() OVER (PARTITION BY OrderID ORDER BY DateReceived), DateReceived
FROM Receipts
WHERE Amount > 0.00
/* Calculate the start date for each period */
UPDATE S
SET StartDate = DATEADD(D, 1, Prev.EndDate)
FROM
#Sequences AS S
INNER JOIN #Sequences AS Prev ON S.OrderID = Prev.OrderID AND Prev.Sequence = S.Sequence - 1
/* Calculate the cumulative total for each period */
UPDATE M
SET CumulativeReceipts = R.Receipts
FROM
#Sequences AS M
INNER JOIN
(
SELECT
M.OrderID, M.Sequence, SUM(R.Amount) AS Receipts
FROM
#Sequences AS M
INNER JOIN Receipts AS R ON M.OrderID = R.OrderID AND R.DateReceived <= M.EndDate
GROUP BY
M.OrderID, M.Sequence
) AS R ON M.OrderID = R.OrderID AND M.Sequence = R.Sequence
/* Delete sequences with do not represent net positive receipts */
DELETE FROM M
FROM #Sequences AS M
WHERE EXISTS (SELECT * FROM #Sequences AS Prev WHERE M.OrderID = Prev.OrderID AND Prev.Sequence < M.Sequence AND Prev.CumulativeReceipts >= M.CumulativeReceipts)
/* Recalculate sequence numbers and dates */
UPDATE S SET Sequence = NewSequence FROM (SELECT Sequence, ROW_NUMBER() OVER (PARTITION BY OrderID ORDER BY Sequence) AS NewSequence FROM #Sequences) AS S
UPDATE S
SET StartDate = DATEADD(D, 1, Prev.EndDate)
FROM
#Sequences AS S
INNER JOIN #Sequences AS Prev ON S.OrderID = Prev.OrderID AND Prev.Sequence = S.Sequence - 1
END
/* Calculate net captures per period, and continue with analysis */