基于另一列中的总计的T-SQL不等分十分位数

Question

I didn't see an exact answer for what I was looking for. I have a table with an ID and two values. I need to sort the first value column low to high and then decile the list based on each decile having an equal (or almost equal) total value 2. Here's an example using quartiles for space considerations:

I have:

ID  value1  value2      
1     2      132        
2     6      182        
3     5      195        
4     8      152        
5     3      132        
6     9      129        
7     3      180        
8     9      120        
9     3      172        
10    6      192        
11    9      177        
12    12     151        

Each quartile should about about 478.5

Sorting by value1 gets this but I need to be able to assign my quartile where each is about 478.5. I have manually entered sample quartiles which may or may not be correct based on the calculations

ID  value1  value2  Qtle    
1     2      132      1 
5     3      132      1 
7     3      180      1 
9     3      172      2 
3     5      195      2 
2     6      182      3 
10    6      192      3 
4     8      152      3 
6     9      129      4 
8     9      120      4 
11    9      177      4 
12   12      151      4 

Sorry about the formatting - first post.

Edit 1 - I think I might have solved it, although it's probably not as elegant as it could be

Edit 2 - Added sample quartiles above and fixed the code below to reflect quartiles instead of deciles. Also fixed the sum of value2

SELECT value1
    ,value2
,SUM(value2) OVER (ORDER BY value1 ) CumSum
,CASE
    WHEN SUM(value2) OVER (ORDER BY value1 ) < (Select sum(value2) from table1)/4 Then 1 
   WHEN SUM(value2) OVER (ORDER BY value1 ) < 2 * (Select sum(value2) from 
table1)/4 Then 2 
    WHEN SUM(value2) OVER (ORDER BY value1 ) < 3 * (Select sum(value2) from 
table1)/4 Then 3 
    Else 4 
 End as Quartile
FROM Table1

Answer 1

I hope I've got this correctly...

The following is a generic approach. You can specify the @TileCount with a variable:

DECLARE @Table1 TABLE(ID INT,value1 INT,value2 INT);
INSERT INTO @Table1 VALUES      
 (1,2,132)        
,(2,6,182)        
,(3,5,195)        
,(4,8,152)        
,(5,3,132)        
,(6,9,129)        
,(7,3,180)        
,(8,9,120)        
,(9,3,172)        
,(10,6,192)        
,(11,9,177)        
,(12,12,151);

DECLARE @TileCount INT=4;

WITH Sums AS
(
    SELECT TOP (@TileCount) ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) AS TileRank
              ,A.SumTotal
              ,ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) * (A.SumTotal / CAST(@TileCount AS FLOAT)) AS SumPart  
    FROM master..spt_values
    CROSS APPLY(SELECT (SELECT SUM(value2) FROM @Table1) AS SumTotal)AS A
)
,AddCumSum AS
(
    SELECT value1
          ,value2
          ,SUM(value2) OVER (ORDER BY value1) CumSum
     FROM @Table1
)
SELECT AddCumSum.*
      ,A.SumPart
      ,A.TileRank AS Tile
FROM AddCumSum
OUTER APPLY(SELECT TOP 1 * FROM Sums WHERE CumSum<=SumPart ORDER BY TileRank ASC) AS A;

The result

+--------+--------+--------+---------+------+
| value1 | value2 | CumSum | SumPart | Tile |
+--------+--------+--------+---------+------+
| 2      | 132    | 132    | 478,5   | 1    |
+--------+--------+--------+---------+------+
| 3      | 132    | 616    | 957     | 2    |
+--------+--------+--------+---------+------+
| 3      | 180    | 616    | 957     | 2    |
+--------+--------+--------+---------+------+
| 3      | 172    | 616    | 957     | 2    |
+--------+--------+--------+---------+------+
| 5      | 195    | 811    | 957     | 2    |
+--------+--------+--------+---------+------+
| 6      | 182    | 1185   | 1435,5  | 3    |
+--------+--------+--------+---------+------+
| 6      | 192    | 1185   | 1435,5  | 3    |
+--------+--------+--------+---------+------+
| 8      | 152    | 1337   | 1435,5  | 3    |
+--------+--------+--------+---------+------+
| 9      | 120    | 1763   | 1914    | 4    |
+--------+--------+--------+---------+------+
| 9      | 129    | 1763   | 1914    | 4    |
+--------+--------+--------+---------+------+
| 9      | 177    | 1763   | 1914    | 4    |
+--------+--------+--------+---------+------+
| 12     | 151    | 1914   | 1914    | 4    |
+--------+--------+--------+---------+------+

Some explanation

The CTE Sums computes some values which allows to use them as named variables . The @TileCount is used within the TOP clause in connection with ROW_NUMBER() selecting from master..spt_values . This is nothing else than a well filled table. We are not interested in the values, we just need it as the base to get a running number.

The second CTE AddCumSum returns the result with the running summa.

The final SELECT finds the smallest TileRank fitting to the running summa.