SQL Server-基於集合的瀑布

Question

我有一個表，用於在不同的存儲桶（存儲桶1、2和3）中存儲數字余額。 在系統中過帳金額后，必須將金額分配到存儲桶。 事務類型決定了它如何影響存儲桶。 類型1、2和3增加相應的存儲桶（無瀑布），而類型0使用瀑布減少存儲桶。 這意味着類型0的數量首先減少存儲區1，然后任何剩余的存儲區減少存儲區2，任何剩余的存儲區減少存儲區3。

我目前使用一個游標應用金額，該游標遍歷每個已過帳的金額。 由於瀑布的緣故，關鍵是要應用的金額順序。 顯然，在處理成千上萬的金額時，它相當慢。 我想知道是否有基於集合的方法來執行此操作？

生成表和記錄的代碼：

http://pastebin.com/XgKrKkbm

我的分配數量的光標如下：

DECLARE @Instrument int,
        @Type int,
        @Amount numeric(19,2),
        @NewAmount numeric(19,2),
        @Seq int,
        @Bucket1 numeric(19,2),
        @Bucket2 numeric(19,2),
        @Bucket3 numeric(19,2)

DECLARE Waterfall CURSOR Fast_Forward FOR 
        SELECT InstrumentID, TypeID, Amount, Sequence
        FROM Amount
        WHERE EffectiveDate between '2015-06-20' and '2015-06-30'
        ORDER BY InstrumentID, Sequence

OPEN Waterfall;
FETCH NEXT from Waterfall into @Instrument, @Type, @Amount, @Seq

WHILE @@FETCH_STATUS = 0
    BEGIN

    SELECT @Bucket1 = Bucket1, @Bucket2 = Bucket2, @Bucket3 = Bucket3
    FROM Buckets WHERE InstrumentID = @Instrument

        IF @Type > 0 /*Increase Buckets, no waterfall*/
        BEGIN
            SET @Bucket1 = CASE WHEN @Type = 1 THEN @Bucket1 + @Amount ELSE @Bucket1 END
            SET @Bucket2 = CASE WHEN @Type = 2 THEN @Bucket2 + @Amount ELSE @Bucket2 END
            SET @Bucket3 = CASE WHEN @Type = 3 THEN @Bucket3 + @Amount ELSE @Bucket3 END
        END

        ELSE /*Decrease buckets with waterfall*/
        BEGIN

            SET @NewAmount  = CASE WHEN @Amount >= @Bucket1 THEN @Amount - @Bucket1 ELSE 0 END
            SET @Bucket1    = CASE WHEN @Amount >= @Bucket1 THEN 0 ELSE @Bucket1 - @Amount END
            SET @Amount     = @NewAmount
            SET @NewAmount  = CASE WHEN @Amount >= @Bucket2 THEN @Amount - @Bucket2 ELSE 0 END
            SET @Bucket2    = CASE WHEN @Amount >= @Bucket2 THEN 0 ELSE @Bucket2 - @Amount END
            SET @Amount     = @NewAmount
            SET @Bucket3    = CASE WHEN @Amount >= @Bucket3 THEN 0 ELSE @Bucket3 - @Amount END
        END

/*Record effect of each amount on the balances for audit/undo*/
    UPDATE  Amount
    SET     Bucket1 = @Bucket1 - Buckets.Bucket1, 
            Bucket2 = @Bucket2 - Buckets.Bucket2, 
            Bucket3 = @Bucket3 - Buckets.Bucket3
    FROM    Buckets 
        inner join Amount 
            on Amount.InstrumentID = Buckets.InstrumentID
    where Sequence = @Seq

/*update bucket values in table*/    
    UPDATE Buckets
        SET Bucket1 = @Bucket1,
            Bucket2 = @Bucket2,
            Bucket3 = @Bucket3
    WHERE InstrumentID = @Instrument

FETCH NEXT from Waterfall INTO @Instrument, @Type, @Amount, @Seq
END

CLOSE Waterfall
DEALLOCATE Waterfall

創建和插入腳本位於PasteBin，因為這樣，當包含它們時，我不會讓我發布問題（文本長度？）。

---

清楚說明瀑布序列至關重要的示例：

從以下內容開始：

Instrument | Bucket1 | Bucket2 | Bucket 3
1          |     500 |     200 |     3000

如下適用兩個數額

Instrument | Sequence|    Type |   Amount
1          |       1 |       0 |     800  /*this decreases buckets via waterfall*/
1          |       2 |       1 |     500  /*this increases bucket 1*/

如果按順序應用金額，則結果如下：

Instrument | Bucket1 | Bucket2 | Bucket 3
1          |     500 |       0 |     2900

但是，金額按相反順序應用，結果不正確：

Instrument | Bucket1 | Bucket2 | Bucket 3
1          |     200 |     200 |     3000

Answer 1

這樣的東西就是您想要的。 比CURSOR效率要高得多。

UPDATE
  b
SET
  Bucket1 = CASE WHEN a.[Type] = 1 THEN b.Bucket1 + a.Amount
                 WHEN a.[Type] = 0 AND Amount >= b.Bucket1 THEN 0
                 WHEN a.[Type] = 0 AND Amount <  b.Bucket1 THEN b.Bucket1 - a.Amount
                 ELSE Bucket1  
            END,
  Bucket2 = CASE WHEN a.[Type] = 2 THEN b.Bucket2 + a.Amount
                 WHEN a.[Type] = 0 AND Amount >= b.Bucket1 + Bucket2 THEN 0
                  WHEN a.[Type] = 0 AND Amount BETWEEN  b.Bucket1 AND Bucket2 THEN Bucket2 - (a.Amount - b.Bucket1)
                 ELSE Bucket2
            END,
  Bucket3 = CASE WHEN a.[Type] = 3 THEN b.Bucket3 + a.Amount
                 WHEN a.[Type] = 0 AND Amount >= b.Bucket1 + Bucket2 THEN Bucket3  - (a.Amount - b.Bucket1 -  Bucket2)
                 ELSE Bucket3
            END
FROM
  dbo.Buckets AS b
  JOIN 
  (SELECT TOP 10000000 * FROM dbo.Amounts ORDER BY Sequence) AS a
    ON b.InstrumentID = a.InstrumentID