hive row_number等窗口分析函数

 

一、排序&去重分析

row_number() over(partititon by col1 order by col2) as rn

结果:1,2,3,4

rank() over(partititon by col1 order by col2) as rk

结果:1,2,2,4,5

dense_rank() over(partititon by col1 order by col2) as ds_rk

结果:1,2,2,3,4

select 
        order_id,
        departure_date,
        row_number() over(partition by order_id order by departure_date) as rn,  -- 直排
        rank() over(partition by order_id order by departure_date) as rk,        -- 并列的,下一个数字会跳过
        dense_rank() over(partition by order_id order by departure_date) as d_rk -- 并列的,下一个数据不会跳过
  from ord_test 
 where order_id=410341346
;

运行结果:

 

二、跨行获取  

lag(col1,n,DEFAULT) over(partition by col1 order by col2) as up
用于统计窗口内往上第n行值,第一个参数为列名,第二个参数为往上第n行(可选,默认为1),第三个参数为默认值(当往上第n行为NULL时候,取默认值,如不指定,则为NULL)

lead(col1,n,DEFAULT) over(partition by col1 order by col2) as down
用于统计窗口内往下第n行值,第一个参数为列名,第二个参数为往下第n行(可选,默认为1),第三个参数为默认值(当往下第n行为NULL时候,取默认值,如不指定,则为NULL)

first_value() over(partition by col1 order by col2) as fv
取分组内排序后,截止到当前行,第一个值

last_value() over(partition by col1 order by col2) as lv
取分组内排序后,截止到当前行,第一个值

select 
       order_id,
       departure_date,
       first_value(departure_date) over(partition by order_id order by add_time)as fv,  -- 取分组内第一条
       last_value(departure_date) over(partition by order_id order by add_time)as lv    -- 取分组内最后一条  
  from ord_test
 where order_id=410341346
;

select 
       order_id,
       departure_date,
       lead(departure_date,1) over(partition by order_id order by departure_date)as down_1, -- 向下取一级
       lag(departure_date,1) over(partition by order_id order by departure_date)as up_1     -- 向上取一级
  from ord_test
 where order_id=410341346
;

结果截图:

 

posted @ 2018-02-12 19:46  Kimbo  阅读(3809)  评论(0编辑  收藏  举报