Hive分组取前N条记录

2019独角兽企业重金招聘Python工程师标准>>>

需求：根据主叫被叫通话记录，取出每个号码呼出前三的记录

--建表，calling主叫,called被叫
create external table beijiao(calling string,called string) row format delimited 
fields terminated by '%' location 'beijiao/'

--通过group统计出被叫号码呼叫次数，rnk是对分组数据进行标号1,2,3......,分组是根据calling
select a.calling,a.called,count(*) num,row_number() over(partition by calling) as rnk 
from beijiao a group by a.calling,a.called;

--上面语句rnk分组的序号不是按呼叫次数排序的，所以还要执行下面的语句，用num进行排序，重新打上标号，并建成一张表
select a.calling,a.called,num,row_number() over(partition by calling order by num desc) 
as rnk2 from cl_bj_2 a;

--根据rnk2取出前3条记录
create table cl_bj_4 as select called from cl_bj_3 where rnk2<4;

 

转载于:https://my.oschina.net/iamchenli/blog/1545267