递归调用同一列中的滞后值，或者有两个相互依赖的列

Question

第一个查询没有运行，因为我在 LAG() 中有一个 SUM()，这是非法的，但我将其包含在内只是为了概念化这个想法。

SELECT
     GUID_Key
    ,AFE_Number
    ,UWI
    ,Date
    ,ACT_NET_AMOUNT
    ,Cum_Act
    ,FC_NET_AMOUNT
    ,Cum_FC
    ,ROWNUM
    ,IIF(Cum_FC > Cum_Act, FC_NET_AMOUNT, Cum_Act-ISNULL(LAG(SUM(Cum_Act)OVER(PARTITION BY Date))OVER (ORDER BY ROWNUM),0)) AS Result

FROM ready_set

或者

SELECT
     GUID_Key
    ,AFE_Number
    ,UWI
    ,Date
    ,ACT_NET_AMOUNT
    ,Cum_Act
    ,FC_NET_AMOUNT
    ,Cum_FC
    ,ROWNUM
    ,IIF(Cum_FC > Cum_Act, FC_NET_AMOUNT, IIF((LAG(CumRes) OVER( ORDER BY ROWNUM)) IS NULL, Cum_Act, Cum_Act - LAG(CumRes) OVER( ORDER BY ROWNUM))) AS Result
    ,IIF((LAG(CumRes) OVER( ORDER BY ROWNUM)) IS NULL ,/*Result*/IIF(Cum_FC > Cum_Act, FC_NET_AMOUNT, IIF((LAG(CumRes) OVER( ORDER BY ROWNUM)) IS NULL, Cum_Act, Cum_Act - LAG(CumRes) OVER( ORDER BY ROWNUM))), /*Result*/(IIF(Cum_FC > Cum_Act, FC_NET_AMOUNT, IIF((LAG(CumRes) OVER( ORDER BY ROWNUM)) IS NULL, Cum_Act, Cum_Act - LAG(CumRes) OVER( ORDER BY ROWNUM))) + Lag(CumRes) OVER ( ORDER BY ROWNUM))) AS CumRes
FROM ready_set

现在我应该提到在“ready_set”中有两个空列，result 和 CumRes，这就是第二个查询甚至运行的原因。

这是 Excel 电子表格中显示的确切想法。原则上它并不是非常复杂，但我猜它是在编程中！ Excel Formula of the same idea

编辑：Excel results of the formula

同时，我试图在 CumRes 下保持结果的运行总计，并根据之前的累积计算结果。它们是相互依赖的，所以我不能简单地将结果相加来得到 CumRes 列。我并不完全了解 SQL 如何生成行，如果我试图滞后于正在制作而不是预先存在的列，我可能会采用完全错误的方法。你看到我的问题或思维错误了吗？

Answer 1

如果你想要一个累计和：

select rs.*, sum(fc_net_amount) over (order by rownum)
from ready_set rs

Answer 2

create table tableX (rownum int not null,
    ACC_NET_ int not null, CumACT int not null,
    FC_NET_AMOUNT int not null, CumFC int not null);

insert into tableX (rownum, ACC_NET_, CumACT, FC_NET_AMOUNT, CumFc) values
    (1, 40, 40, 50, 50), (2, 164, 204, 2050, 2100), (3, 1513, 1717, 0, 2100),
    (4, 1000, 2717, 500, 2600), (5, 100, 2817, 0, 2600);    
go

with iterated as (
    select rownum, ACC_NET_, CumACT, FC_NET_AMOUNT, CumFc,
        iif(CumFc > CumACT, FC_NET_AMOUNT, CumACT - 0) as result,
        0 as CumRes
    from tableX
    where rownum = 1
    union all
    select x.rownum, x.ACC_NET_, x.CumACT, x.FC_NET_AMOUNT, x.CumFc,
        f.result,
        f.result + anchor.CumRes
    from iterated as anchor inner join tableX x
            on x.rownum = anchor.rownum + 1
        cross apply (
            select iif(x.CumFc > x.CumACT, x.FC_NET_AMOUNT, x.CumACT - anchor.CumRes)
        ) as f(result)
)
select * from iterated;

递归查询适用于您的公式。如果问题比您上面的数据所建议的要大，您可能需要阅读有关该主题的更多信息。此外，我确实必须添加一个 rownum 列才能按顺序对行进行排序。

编辑：我错误地播种了一个初始值并且已修复。

https://rextester.com/UXXY92525

Answer 3

这似乎符合您的公式：

with cte as 
 (
   select ACC_NET_AMOUNT, fc_NET_AMOUNT
     -- cumulative sums up to the previous rows
     , coalesce(sum(ACC_NET_AMOUNT) over (order by rownum rows between unbounded preceding and 1 preceding) ,0) as cumacc
     , coalesce(sum( FC_NET_AMOUNT) over (order by rownum rows between unbounded preceding and 1 preceding), 0) as cumfc
   from tableX
 )
select ACC_NET_AMOUNT, fc_NET_AMOUNT
  , ca.CumResult - case when cumfc > cumacc then cumfc else cumacc end as Result
  , ca.CumResult
from cte 
cross apply
 ( select
      case when cumfc+FC_NET_AMOUNT > cumacc + ACC_NET_AMOUNT 
           then cumfc+FC_NET_AMOUNT 
           else cumacc+ACC_NET_AMOUNT 
      end as CumResult) as ca

见fiddle