Question

我需要找到每个组中n个最大值的总和。

使用（n = 2）：

group | points  
g1    |  3  
g2    |  3  
g3    |  4  
g1    |  2  
g1    |  4  
g2    |  5  
g2    |  5  
g3    |  1  
g3    |  2

结果

group | sum  
g1    | 7  
g2    | 10  
g3    | 6

使用联接和组的

谢谢

Answer 1

如果RDBMS支持窗口功能，则可以使用ROW_NUMBER()为组中的每个记录分配一个数字（按点排序），然后在外部汇总查询中过滤出每个组的前2条记录。

SELECT grp, SUM(points) total
FROM (
    SELECT grp, points, ROW_NUMBER() OVER(PARTITION BY grp ORDER BY points DESC) rn
    FROM mytable
) x 
WHERE rn <= 2
GROUP BY grp
ORDER BY grp

此 MySQL 8.0 DB Fiddle 和您的示例数据可得出：

| grp | total |
| --- | ----- |
| g1  | 7     |
| g2  | 10    |
| g3  | 6     |

Answer 2

示例数据集：

create table #temp (name varchar(20), value int)
insert into #temp values ('g1',2),('g2',2),('g3',2),('g2',7),
                         ('g3',9),('g1',4),('g2',8),('g3',1),('g1',3),('g1',11)

处理此问题的另一种方法是使用“交叉应用”，例如下面的Sql Server：

--This returns top 2 rows for each group where its value is highest.
SELECT x.*
FROM ( SELECT DISTINCT name FROM #temp ) c
    CROSS APPLY ( SELECT TOP 2 * FROM #temp t WHERE c.name = t.name order by value desc ) x

--This returns sum of top 2 value of each group   
SELECT x.name, SUM(x.Value) as Total
FROM ( SELECT DISTINCT name FROM #temp ) c
    CROSS APPLY ( SELECT TOP 2 * FROM #temp t WHERE c.name = t.name order by value desc ) x
group by x.name

由于这里的@n值不是静态的，并且将根据用户的选择进行更改，因此可以使用如下所示的动态查询：

declare @n int = 2;
declare @sql nvarchar(max) = 'SELECT x.name, SUM(x.Value) as Total
FROM ( SELECT DISTINCT name FROM #temp ) c
       CROSS APPLY ( SELECT TOP '+cast(@n as nvarchar(10))+' * FROM #temp t 
       WHERE c.name = t.name order by value desc ) x
       group by x.name'
exec sp_executesql @sql

sql：每个组中最高n值的总和

2 个答案: