hive 建表中group by 和over partition by运行效率比较

[xiaoran_Li]

数据量级大概4kw左右，我想要聚合不同维度的统计值，简单来说，想要聚合：
本公司销售件数；
同行业平均销售件数；
本省同行业平均销售件数；
本市同行业平均销售件数：

两种方法：1.group by
分别left join 4次，大致内容：

（select company,count(1) from t
group by company) t1
on 条件1
left join
（select category,count(1) from t
group by category) t2
on 条件2
left join
（select province,category,count(1) from t
group by province,category) t3
on 条件3
left join
（select city,category,count(1) from t
group by city,category) t4
on 条件4



2.over partition by

select company,
count(1) over (partition by name) ,
count(1) over (partition by category),
count(1) over (partition by province,category),
count(1) over (partition by city,category)
from t

求指教哪一种效率更高呢？为啥啊。。。如果想写的内容少一点的话肯定会用 partition by....