通过SQL选择TOP 2 MAX值组

Question

我有下表：

-----------------------------------------
xDate        xItem       xCount
-----------------------------------------
2018-01-01   A           100
2018-01-01   B           200
2018-01-01   D           500
2018-01-02   C           200
2018-01-02   E           800

我想在xCount字段的MAX值上为每个日期选择TOP 2值。 因此，结果应为：

-----------------------------------------
xDate        xItem      xCount
-----------------------------------------
2018-01-01   D          500
2018-01-01   B          200
2018-01-02   E          800
2018-01-02   C          200

有人对此案有想法吗？
干杯，

Answer 1

您可以尝试使用RANK窗口函数，如果两行中的xCount相同，则您希望全部获取它们。

您可以尝试使用dense_rank而不是RANK

SELECT xDate,xItem,xCount
FROM (
    SELECT *,RANK() OVER(PARTITION BY xDate ORDER BY xCount DESC) rn
    FROM T
) t1
WHERE t1.rn <= 2

Answer 2

您可以使用ROW_NUMBER()并根据xDate对其进行分区， xDate进行xCount以获取所需的内容。

select X.xDate, 
       X.xItem, 
       X.xCount
  from 
       (select xDate, 
               xItem, 
               xCount, 
               row_number() over (partition by xDate order by xCount desc) rank_of_count
         from table_name) X
 where rank_of_count < 3

Answer 3

好吧，我不会使用任何功能。 对我来说看起来很直白。 最快的是

SELECT * FROM   #temp s
WHERE ( SELECT  COUNT(*)  FROM    #temp  f WHERE f.xDate = s.xDate AND  f.xCount >= s.xCount ) <= 2
Order by xDate, xCount desc

在此处检查完整的示例代码：

create table #temp (xDate datetime, xItem nvarchar(max), xCount int);

insert into #temp
select
'2018-01-01','A', 100 union all 
select '2018-01-01','B', 200 union all 
select '2018-01-01','D', 500 union all 
select '2018-01-02','C', 200 union all 
select '2018-01-02','E', 800


SELECT * FROM   #temp s
WHERE ( SELECT  COUNT(*)  FROM    #temp  f WHERE f.xDate = s.xDate AND  f.xCount >= s.xCount ) <= 2
Order by xDate, xCount desc

drop table #temp;

Answer 4

另一种方法是使用TOP，CROSS APPLY。

;WITH CTE AS
(
select 
distinct xDate
from Your_Table
)
SELECT 
T.xDate,
T1.xItem,
T1.xCount
FROM CTE T
CROSS APPLY (SELECT TOP 2 xItem,xCount FROM Your_Table WHERE xDate=T.xDate ORDER BY xCount DESC ) T1

Answer 5

另一个使用DENSE_RANK()建议：

DECLARE @FooTable TABLE
(
    xDate VARCHAR(25),
    xItem VARCHAR(10),
    xCount INT  
)

INSERT INTO @FooTable
(
    xDate,
    xItem,
    xCount
)
VALUES
  ('2018-01-01',   'A',           100)
, ('2018-01-01',   'B',           200)
, ('2018-01-01',   'D',           500)
, ('2018-01-02',   'C',           200)
, ('2018-01-02',   'E',           800)

SELECT 
s.*
FROM 
(
    SELECT 
      ft.xDate
    , ft.xItem
    , ft.xCount
    --, ROW_NUMBER() OVER(PARTITION BY ft.xDate ORDER BY ft.xCount DESC) rn
    , DENSE_RANK() OVER (PARTITION BY ft.xDate ORDER BY ft.xCount desc) dr
    FROM @FooTable ft
)s
WHERE s.dr < 3