前行的总和并与另一列的值相乘

Question

我有一张大约有 80000 条记录的表。 这是一个例子：

开始年	分配	帐户	预算	预报
2021	40	4100	5122952,22	0,012306656
2022	40	4100	0	0,011424198
2023	40	4100	0	0,010431491
2024	40	4100	0	0,009311863
2025	40	4100	0	0,008324122
2026	40	4100	0	0,007763793
2027	40	4100	0	0,007557735
2028	40	4100	0	0,007357883
2029	40	4100	0	0,007160051
2030	40	4100	0	0,006953345
2031	40	4100	0	0,006737952
2032	40	4100	0	0,006535297
2033	40	4100	0	0,006364179
2034	40	4100	0	0,006213237
2035	40	4100	0	0,006085724
2036	40	4100	0	0,005944279
2037	40	4100	0	0,005758285
2038	40	4100	0	0,005559474
2039	40	4100	0	0,005360105
2040	40	4100	0	0,005163794
2041	40	4100	0	0,004972228

我正在尝试从第 1 行的预算开始计算一个新列。所以 2021 年将是相同的。 第一个计算是正确的。 但是计算的rest忽略了之前的变化。 这是代码：

WITH cte AS (
SELECT startyear, division, account, forecast, SUM(budget) AS asum
FROM @table
GROUP BY startyear,
         division,
         account,
         forecast
), cteRanked AS (
SELECT cte.startyear,
             cte.division,
             cte.account,
             cte.forecast,
             cte.asum AS asum
FROM cte
)
SELECT c1.startyear,
          c1.division,
          c1.account,
          c1.forecast,
          c1.asum , ISNULL((SELECT SUM(c2.asum)*c1.forecast 
FROM cteRanked c2 
WHERE c1.startyear > 2021 
    AND c2.startyear < c1.startyear 
    AND c2.division = c1.division 
    AND c2.account = c1.account), c1.asum) AS Bud
FROM cteRanked c1
ORDER BY c1.division, c1.account, c1.startyear

给出这个结果：

开始年	分配	帐户	预报	预算	结果	正确的
2021	40	4100	0,012306656	5122952,22	5122952,22	5122952,22
2022	40	4100	0,011424198	0	58525,62051	58525,62
2023	40	4100	0,010431491	0	53440,02998	54050,54
2024	40	4100	0,009311863	0	47704,22923	48752,52
2025	40	4100	0,008324122	0	42644,07928	43987,00
2026	40	4100	0,007763793	0	39773,54059	41367,57

如您所见，此查询在查询运行时不会累积结果。 有人有想法么？

Answer 1

虽然您的帖子缺少一些重要信息……但我不一定要责怪您，因为这是一个很难解释和解决的问题。

看起来你基本上是在想出一个复利计算器。 除了这种情况，费率每年都在变化。

为了计算您的预测的 PRODUCT 聚合，我找到了这篇博文： https://blog.jooq.org/2018/09/21/how-to-write-a-multiplication-aggregate-function-in-sql/

它只需要一点点调整。

这是我的回答：

DECLARE @StartingYear int = 2021,
        @StartingBudget decimal(12, 2);

SELECT @StartingBudget = yt.budget
FROM #YourTable yt
WHERE yt.startyear = @StartingYear

SELECT yt.startyear, yt.division, yt.account, yt.budget, yt.forecast, YearBudget = yt.budget, PrevYearDiff = yt.budget
FROM #YourTable yt
WHERE yt.startyear = @StartingYear
UNION ALL
SELECT x.startyear, x.division, x.account, x.budget, x.forecast
     , YearBudget = CONVERT(decimal(10,2), x.SumProdBudget)
     , PrevYearDiff = CONVERT(decimal(10,2), x.SumProdBudget - LAG(x.SumProdBudget,1,@StartingBudget) OVER (ORDER BY x.startyear))
FROM (
    SELECT yt.startyear, yt.division, yt.account, yt.budget, yt.forecast
        , SumProdBudget = EXP(SUM(LOG(1+yt.forecast)) OVER (ORDER BY yt.startyear)) * @StartingBudget
    FROM #YourTable yt
    WHERE yt.startyear > @StartingYear
) x

回报：

| startyear | division | account | budget     | forecast    | YearBudget | PrevYearDiff | 
|-----------|----------|---------|------------|-------------|------------|--------------| 
| 2021      | 40       | 4100    | 5122952.22 | 0.012306656 | 5122952.22 | 5122952.22   | 
| 2022      | 40       | 4100    | 0.00       | 0.011424198 | 5181477.84 | 58525.62     | 
| 2023      | 40       | 4100    | 0.00       | 0.010431491 | 5235528.38 | 54050.54     | 
| 2024      | 40       | 4100    | 0.00       | 0.009311863 | 5284280.90 | 48752.52     | 
| 2025      | 40       | 4100    | 0.00       | 0.008324122 | 5328267.90 | 43987.00     | 
| 2026      | 40       | 4100    | 0.00       | 0.007763793 | 5369635.47 | 41367.57     | 

整个事情的关键是这一行：

SELECT SumProdBudget = EXP(SUM(LOG(1+yt.forecast)) OVER (ORDER BY yt.startyear)) * @StartingBudget

这就是说返回所有先前预测的运行产品，然后将其乘以原始预算。 这将产生每年的预算，技术上基于上一年的预算。

然后，一旦你有了它，我就把它放在一个子查询中，以找出当年和上一年之间的差异。

---

样本数据：

IF OBJECT_ID('tempdb..#YourTable','U') IS NOT NULL DROP TABLE #YourTable; --SELECT * FROM #YourTable
CREATE TABLE #YourTable (
    startyear   int             NOT NULL,
    division    int             NOT NULL,
    account     int             NOT NULL,
    budget      decimal(12, 2)  NOT NULL,
    forecast    decimal(10, 9)  NOT NULL,
);

INSERT INTO #YourTable (startyear, division, account, budget, forecast)
VALUES (2021, 40, 4100, 5122952.22, 0.012306656)
     , (2022, 40, 4100, 0         , 0.011424198)
     , (2023, 40, 4100, 0         , 0.010431491)
     , (2024, 40, 4100, 0         , 0.009311863)
     , (2025, 40, 4100, 0         , 0.008324122)
     , (2026, 40, 4100, 0         , 0.007763793)
     , (2027, 40, 4100, 0         , 0.007557735)
     , (2028, 40, 4100, 0         , 0.007357883)
     , (2029, 40, 4100, 0         , 0.007160051)
     , (2030, 40, 4100, 0         , 0.006953345)
     , (2031, 40, 4100, 0         , 0.006737952)
     , (2032, 40, 4100, 0         , 0.006535297)
     , (2033, 40, 4100, 0         , 0.006364179)
     , (2034, 40, 4100, 0         , 0.006213237)
     , (2035, 40, 4100, 0         , 0.006085724)
     , (2036, 40, 4100, 0         , 0.005944279)
     , (2037, 40, 4100, 0         , 0.005758285)
     , (2038, 40, 4100, 0         , 0.005559474)
     , (2039, 40, 4100, 0         , 0.005360105)
     , (2040, 40, 4100, 0         , 0.005163794)
     , (2041, 40, 4100, 0         , 0.004972228);