SQL

Question

图片 1：示例数据

图 2：输出不正确

图 3：所需输出

查询：我正在尝试通过 Class_Name 和客户（image1 样本数据）

从列 (Median_Percentage) 中查询最大值

问题：查询显示所有客户，而不是具有最大中值的客户（image2 结果不正确）。它正在正确计算 Max() 但查询将所有客户的值而不是在 Class_Name

内具有该最大值的客户的值

我需要的只是Class_Name有Max(Median_Percentage)的给客户看。（image3 所需的输出）

Select  
       distinct  
        C.Class_Name,
        C.Customer,
        C.Max_Median_Percentage

        

FROM (
   
    SELECT 
        
        B.Class_Name,
        
        case (when B.Median_Percentage =  Max(B.Median_Percentage) OVER(PARTITION By B.Class_Name ORDER BY B.Median_Percentage desc  )
            then B.Customer
        end as Customer,

        Max(B.Median_Percentage) OVER(PARTITION By B.Class_Name ORDER BY B.Median_Percentage desc  ) as Max_Median_Percentage

    FROM (
        
        SELECT 
  
            
            A.Class_Name,
            A.Customer,
            A.Date_Time
            
            A.Median_Percentage

        From table1 as A

    ) as B


) as C

Answer 1

如果您的数据库不直接支持“中值”函数，您可以使用 percentile_cont()：

select t.*,
       boot_time / percentile_cont(0.5) within group (order by boot_time) over (partition by classid)
from t;

如果您的数据库没有 percentile_cont() 或 percentile_disc() 函数，您可以使用简单的 ntile():

获得非常接近的结果

select t.*,
       boot_time / max(case when tile = 1 then boot_time end) over (partition by classid)
from (select t.*,
             ntile(2) over (partition by classid order by boot_time) as tile
      from t
     ) t

如果 classid 中的行数为奇数，这将完全有效。对于偶数，它相差 1。您可以轻松处理，但更复杂：

select t.*,
       (boot_time /
        (( max(case when tile_asc = 1 then boot_time end) over (partition by classid) / 2 +
           max(case when tile_desc = 1 then boot_time end) over (partition by classid)
         ) / 2
        )
       )
from (select t.*,
             ntile(2) over (partition by classid order by boot_time) as tile_asc,
             ntile(2) over (partition by classid order by boot_time desc) as tile_desc
      from t
     ) t

Answer 2

也许这有用 -

加载提供的测试数据

val df = spark.sql(
      """
        |select Class_Name, Customer, Date_Time, Median_Percentage
        |from values
        |   ('ClassA', 'A', '6/13/20', 64550),
        |   ('ClassA', 'B', '6/6/20', 40200),
        |   ('ClassB', 'F', '6/20/20', 26800),
        |   ('ClassB', 'G', '6/20/20', 18100)
        |  T(Class_Name, Customer, Date_Time, Median_Percentage)
      """.stripMargin)
    df.show(false)
    df.printSchema()

    /**
      * +----------+--------+---------+-----------------+
      * |Class_Name|Customer|Date_Time|Median_Percentage|
      * +----------+--------+---------+-----------------+
      * |ClassA    |A       |6/13/20  |64550            |
      * |ClassA    |B       |6/6/20   |40200            |
      * |ClassB    |F       |6/20/20  |26800            |
      * |ClassB    |G       |6/20/20  |18100            |
      * +----------+--------+---------+-----------------+
      *
      * root
      * |-- Class_Name: string (nullable = false)
      * |-- Customer: string (nullable = false)
      * |-- Date_Time: string (nullable = false)
      * |-- Median_Percentage: integer (nullable = false)
      */

通过 `Class_Name`

找到最大 Median_Percentage 行

    df.groupBy("Class_Name")
      .agg(max(struct($"Median_Percentage", $"Date_Time", $"Customer")).as("struct"))
      .selectExpr("Class_Name", "struct.Customer", "struct.Date_Time", "struct.Median_Percentage")
      .show(false)

    /**
      * +----------+--------+---------+-----------------+
      * |Class_Name|Customer|Date_Time|Median_Percentage|
      * +----------+--------+---------+-----------------+
      * |ClassA    |A       |6/13/20  |64550            |
      * |ClassB    |F       |6/20/20  |26800            |
      * +----------+--------+---------+-----------------+
      */

SQL - 组中的 Max() 值无效

SQL - Max() value from the group is not working

max

window-functions

apache-spark-sql

加载提供的测试数据

通过 `Class_Name`