价值分摊的T-SQL查询

Question

我有一个需求，我需要在多个记录中均匀分配一个值。下面是我的数据看起来像

Relationship Table 
------------------------------------------------------- 
CategoryID   SubCategoryID 
------------------------------------------------------- 
101    1 
101    2 
101    3 
102    4 
102    5  

Table 1 
------------------------------------------------------- 
BatchLine ID  CategoryID  Amount 
------------------------------------------------------- 
1     101    10 
2     102    100 
3     101    100 

Table 2 
------------------------------------------------------------------- 
LineID BatchLineID CategoryID SubCategoryID LineValue 
------------------------------------------------------------------- 
1   1    101   1    3.33 
2   1    101   2    3.33 
3   1    101   3    3.33 
4   2    102   4    66.66 
5   2    102   5    33.33 
6   3    101   1    33.33 
7   3    101   2    33.33 
8   3    101   3    33.33  

在“表2”中的“LineValue”是从“表1”上的“金额”字段进行的计算的结果，然后将最终值四舍五入到两位小数点。在某些情况下，舍入过程会导致与“表格1”中的“金额”字段存在一些差异。

Difference 
------------------------------------------------------------------ 
BatchLineID CategoryID Amount SumofLineValue 
------------------------------------------------------------------ 
1    101   10  9.99 
2    102   100  99.99 
3    101   100  99.99 

为了从“表1”匹配“金额”值，我们需要与类别ID组中添加“0.01”到每个“LineID”直到的“LineValue”等于它的总和。

还有一个条件，在'CategoryID'101的情况下'表2'中有两组数据。我们将“0.01”添加到LineID 1，这将使总和与“表1”中的“金额”字段匹配。当我们需要分摊第二组时，需要以“线路ID”7而不是6开始。下面是我预期的数据。

Expected Output 
------------------------------------------------------------------- 
LineID BatchLineID CategoryID SubCategoryID LineValue 
------------------------------------------------------------------- 
1   1    101   1    3.33 + 0.01 
2   1    101   2    3.33 
3   1    101   3    3.33 
4   2    102   4    66.66 + 0.01 
5   2    102   5    33.33 
6   3    101   1    33.33 
7   3    101   2    33.33 + 0.01 
8   3    101   3    33.33 

Answer 1

只要没有标志哪些记录必须更新，该解决方案就会更新每个组的第一条记录。

with rn as 
(
    select t2.LineID, t1.BatchLineID, t1.CategoryID, t1.Amount, t2.SubCategoryID, 
      sum(t2.LineValue) over (partition by t2.BatchLineID, t2.CategoryID) as ActSum, 
      row_number() over (partition by t2.BatchLineID, t2.CategoryID order by SubCategoryID) as RNum 
    from  t1 
    left join t2 
    on  t1.BatchLineID = t2.BatchLineID 
    and  t1.CategoryID = t2.CategoryID 
) 
update  t2 
set  LineValue = LineValue + (Amount - ActSum) 
from  t2 tt2 
inner join rn 
on   tt2.LineID = rn.LineID 
where  RNum = 1 
and  Amount <> ActSum; 
GO 

select * from t2; 
GO 

 
LineID | BatchLineID | CategoryID | SubCategoryID | LineValue 
-----: | ----------: | ---------: | ------------: | :-------- 
    1 |   1 |  101 |    1 | 3.34  
    2 |   1 |  101 |    2 | 3.33  
    3 |   1 |  101 |    3 | 3.33  
    4 |   2 |  102 |    4 | 66.67  
    5 |   2 |  102 |    5 | 33.33  
    6 |   3 |  101 |    1 | 33.34  
    7 |   3 |  101 |    2 | 33.33  
    8 |   3 |  101 |    3 | 33.33  

dbfiddle here

Answer 2

适应的fiddl从McNets E，你可以使用dense_rank()以确定哪一行应该被更新：

with cte1 as (
    select t2.LineID, t1.BatchLineID, t1.CategoryID, t1.Amount, t2.SubCategoryID, 
      sum(t2.LineValue) over (partition by t2.BatchLineID, t2.CategoryID) as ActSum, 
      row_number() over (partition by t2.BatchLineID, t2.CategoryID order by SubCategoryID) as RNum, 
      dense_rank() over (partition by t1.CategoryID order by t1.BatchLineID) as drank 
    from  t1 
    left join t2 
    on  t1.BatchLineID = t2.BatchLineID 
    and  t1.CategoryID = t2.CategoryID 
) 
select a.LineID, a.BatchLineID, a.CategoryID, a.SubCategoryID, case when drank = rnum then a.LineValue + (b.Amount - b.ActSum) else a.LineValue end as LineValue 
from t2 a 
left join cte1 b 
    on a.LineID = b.LineID 

需要注意的是，你可以运行到随后的群体没有足够的行，因此没有行被更新的问题。