Question

我是HiveQL的初学者级别，我正在尝试编写更快，更高效的查询，但我遇到了麻烦。有人可以帮我改写这个查询吗？您可以提供任何改进我的查询的提示。

select "AUDIOONLYtopctrbyweek37Q32015", weekofyear(day),op.order_id,oppty_amount, mv.order_start_date, mv.order_end_date, count(distinct rdz.listener_id) as listeners,  sum(impressions) , sum(clicks), (sum(clicks)/sum(impressions)) as ctr, sum(oline_net_amount)
from ROLLUP_PST rdz
join dfp2ss mv on (rdz.order_id = mv.dfp_order_id)
join oppty_order_oline op on (mv.order_id = op.order_id)
where day >= '2015-09-07'
and day <= '2015-09-13'
and creative_size in ('2000x132','134x1285','2000x114')
group by "AUDIOONLYtopctrbyweek37Q32015", weekofyear(day),op.order_id,oppty_amount, mv.order_start_date, mv.order_end_date

由ctr desc订购限制150;

Answer 1

请尝试以下修改过的查询。它会对你有用。

select "AUDIOONLYtopctrbyweek37Q32015",week_of_year,order_id,oppty_amount,order_start_date,order_end_date, count(distinct listener_id) over (partition by "AUDIOONLYtopctrbyweek37Q32015",week_of_year,order_id,oppty_amount,order_start_date,order_end_date) from (select "AUDIOONLYtopctrbyweek37Q32015", weekofyear(day) as week_of_year,op.order_id as order_id,
oppty_amount, mv.order_start_date as order_start_date, mv.order_end_date as order_end_date,rdz.listener_id as listener_id
from 
ROLLUP_PST rdz,
dfp2ss mv,
oppty_order_oline op where rdz.order_id = mv.dfp_order_id and mv.order_id = op.order_id and day >= '2015-09-07' and day <= '2015-09-13' 
and creative_size in ('2000x132','134x1285','2000x114')) z

编写子查询时遇到困难

1 个答案: