配置单元子查询

Question

我正在使用Hadoop的0.X版本。此版本不允许子查询。有人可以帮助我将下面的SQL转换为不使用子查询。

select 100 * stars / total from 
(select count(rating) as stars 
from ratingshive 
where rating = 5) t1, 
(select count(1) as total 
from ratingshive) t2

Answer 1

您需要条件聚合。 （还要注意，计算列别名不能在同一查询中使用。）

select 100 * sum(cast(rating=5 as int)) / count(*) 
from ratingshive