Hive分组统计前top N条记录

参考这篇博客而写：http://blog.csdn.net/longshenlmj/article/details/50525385

本 Hive 语句的目的是统计中国每个省份下所有城市记录出现总次数为前5的结果。

hive -e "
select bb.* 
from 
(
select country_name,
       province_name,
       city_name,
       pv_cnt,
       row_number() over (partition by country_name,province_name order by pv_cnt desc) rank_code
from
    (select b.country_name,
            b.province_name,
            b.city_name,
            count(1) as pv_cnt
     from table_a a
     join table_b b
     on a.area_id = b.area_id
     where a.dt = '某天' and b.country_name = '中国'
     group by b.country_name,
	      b.province_name,
	      b.city_name ) aa 
) bb
where bb.rank_code <= 5;"