蜂巢:查找前20%的记录

时间:2019-02-10 17:31:01

标签: hadoop hive bigdata hiveql

我有一些数据,例如:-

ID  PRICE
1   100
2   200
3   120
4   130
5   320
6   300
7   200
8   100
9   120
10  250

我需要找到20%的最高价格。

预期输出:-

ID  PRICE
5   320
6   300

3 个答案:

答案 0 :(得分:1)

您可以不加入联接而这样做。使用解析函数计算max(price),取80%,然后使用过滤器价格> 80%:

with your_data as ( --this is your data
select stack(10,
1 ,  100,
2 ,  200,
3 ,  120,
4 ,  130,
5 ,  320,
6 ,  300,
7 ,  200,
8 ,  100,
9 ,  120,
10,  250) as (ID,  PRICE)
)

select id, price 
from
(
select d.*, max(price) over()*0.8 as pct_80 from your_data d
)s where price>pct_80

结果:

OK
id      price
6       300
5       320

使用表格代替WITH子查询,并根据需要通过 ID 添加订单。

答案 1 :(得分:0)

下面是查询-

with top_20 as (
  select 
    max(price)*0.8 as price1 
  from 
    <tableName>
)
select * from <tableName> t1 , top_20 t2 where t1.price > t2.price1;

select 
 name, 
 price 
from 
 (select 
    name, 
    price, 
    max(price)*0.8 over (order by price) as top_20 
 from <tableName>
 ) t1 
where 
 t1.price > t1.top_20;

以下查询在蜂巢中不起作用-

select * from <tableName> where price > (select max(salary)*0.8 from <tableName>)

select * from <tableName> t1 where exists (select salary from <tablename> t2 where t1.salary > t2.salary*0.8)

原因-Hive不支持条件相同的where子句中的子查询,它仅支持IN,NOT IN,EXISTS和NOT EXISTS。

即使存在且不存在,它也仅支持Equijoin,有关更多详细信息,请参考https://cwiki.apache.org/confluence/display/Hive/LanguageManual+SubQueries#LanguageManualSubQueries-SubqueriesintheWHEREClause

希望这会有所帮助。

答案 2 :(得分:0)

这里提供了一种无需使用join就能做到的方式。

Select id,price from (select id,price, row_number() over(order by price desc) r1,count(*) over()*(20/100) ct from table_name)final where r1<=ct ;