从 SQL Server 中 4 列的表中获取每组的前 N ​​行

Question

假设我有一个有 4 列的表：

Col1   Col2   Col3   Col4  

我的初始查询是：

SELECT Col1, Col2, Col3, Col4  
FROM myTable  
ORDER BY Col1, Col2, Col3 DESC, Col4 

我想要的结果是所有 4 列，但是在这种情况下，当 Col1、Col2 相等时，Top N Col3 不同的行。

N=2 的示例：

表格样本数据：

Col1 Col2 Col3  Col4  
---------------------
1    a    2000  s  
1    a    2002  c  
1    a    2001  b  
2    b    1998  s  
2    b    2002  c  
2    b    2000  b  
3    c    2000  b  
1    f    1998  n  
1    g    1999  e

想要的结果：

1    a    2002  c  
1    a    2001  b  
1    f    1998  n  
1    g    1999  e  
2    b    2002  c  
2    b    2000  b  
3    c    2000  b

另外一种描述，当(col1, col2)在多条记录中重复出现时，按照Col3降序排列，只导出这些记录的前N行。

我可以用 SQL 脚本做到这一点，而不用硬编码吗？

Answer 1

declare @t table (Col1 int, Col2 char, Col3 int,  Col4 char);

insert into @t values  
(1,    'a',    2000,  's'),  
(1,    'a',    2002,  'c'),  
(1,    'a',    2001,  'b'),  
(2,    'b',    1998,  's'),  
(2,    'b',    2002,  'c'),  
(2,    'b',    2000,  'b'),  
(3,    'c',    2000,  'b'),  
(1,    'f',    1998,  'n'),  
(1,    'g',    1999,  'e');


declare @N int = 2; -- number per "top"

with cte as
(
select *,
       row_number() over(partition by col1, col2 order by col3 desc) as rn
from @t
)

select *
from cte c
where rn <= @N;

Answer 2

我认为下面的代码符合预期

declare @tab table (Col1 int, Col2 char(1), Col3 int, Col4 char(1))
declare @N int
insert into @tab
select 1,    'a'   , 2000,  's' 
union all
select 1 ,   'a'  ,  2002 , 'c'  
union all
select 1 ,   'a'  ,  2001 , 'b'  
union all
select 2 ,   'b'  ,  1998 , 's'  
union all
select 2 ,   'b'  ,  2002  ,'c'  
union all
select 2 ,   'b'  ,  2000  ,'b'  
union all
select 3 ,   'c'  ,  2000  ,'b'  
union all
select 1 ,   'f'  ,  1998  ,'n'  
union all
select 1 ,   'g'  ,  1999  ,'e'

;with tab as
(
select ROW_NUMBER() over(partition by t.col1,t.col2 order by t.col3 desc) as row,t.* 
from @tab t
)

select Col1,Col2,Col3,Col4 
from tab
where row < 3

输出

Col1    Col2    Col3    Col4
1        a      2002    c
1        a      2001    b
1        f      1998    n
1        g      1999    e
2        b      2002    c
2        b      2000    b
3        c      2000    b

Answer 3

方法 1- 适用于 MSSQL

http://sqlfiddle.com/#!6/4bda39/6

with a as (
    select ROW_NUMBER() over(partition by t.col1,t.col2 order by t.col3 desc) as row,t.* 
    from myTable as t) 
select * from a where a.row <= 2

替换 a.row （2 用你的 N）

方法 2- FOR MYSQL

http://sqlfiddle.com/#!9/79e81a/63

SELECT   myTable.Col1, myTable.Col2, myTable.Col3, myTable.Col4
FROM ( 
    Select Col1 as Col1, Col2 as Col2, count(Col1) as cc, AVG(Col3) as aa 
    From myTable 
    group by Col1, Col2) as tt
join  myTable on myTable.Col1 = tt.Col1 and myTable.Col2 = tt.Col2
where myTable.Col3 >= tt.aa
Order by Col1 ,Col2 ,Col3 Desc,Col4 

方法 3- FOR MYSQL

http://sqlfiddle.com/#!9/79e81a/79

SELECT * FROM (
    SELECT CASE Col1
        WHEN @Col1 THEN
            CASE Col2 
                WHEN @Col2 THEN @curRow := @curRow + 1 
                ELSE @curRow := 1
            END
        ELSE @curRow :=1
    END AS rank,
    @Col1 := Col1 AS Col1,
    @Col2 := Col2 AS Col2, 
    Col3, Col4
    FROM myTable p
    JOIN (SELECT @curRow := 0, @Col1 := 0, @Col2 := '') r
    ORDER BY Col1, Col2, Col3 DESC) as tt 
WHERE tt.rank <= 2

替换 tt.rank 将 2 替换为您想要的索引