Question

I have data like this:

column1    column2   column3
A          V         10
A          Z         11
A          X         11

And for each data in column1I want to find the maximum value in column3 and the corresponding value in column2. How do I do this in HIVE?

The closest thing I have is to do select column1, max(column3) from table group by column1. But, this doesn't add the corresponding information from column2. How do I get that as well?

Upon a tie in column3, I really don't care which value we pull from column2. Thanks.

I want the result to be:

column1   column2   column3
A         Z         11

Answer 1

一种方法是使用row_number。在绑定的情况下，您将获得column2的任意值。

select column1,column2,column3
from (
select t.*,row_number() over(partition by column1 order by column3 desc) as rn
from tablename t
) x
where rn=1

Answer 2

以下是相同概念的一些变体

select      column1           ,
           ,max(named_struct('column3',column3,'column2',column2)).column2
           ,max(column3) as column3

from        mytable

group by    column1
;

+---------+---------+---------+
| column1 | column2 | column3 |
+---------+---------+---------+
| A       | Z       |      11 |
+---------+---------+---------+

select      column1
           ,max(struct(column3,column2)).col2  as column2
           ,max(column3)                       as column3

from        mytable

group by    column1
;

+---------+---------+---------+
| column1 | column2 | column3 |
+---------+---------+---------+
| A       | Z       |      11 |
+---------+---------+---------+

select      column1
           ,col.column2
           ,col.column3

from       (select      column1
                       ,max(named_struct('column3',column3,'column2',column2)) as col

            from        mytable

            group by    column1
            ) t
;

+---------+---------+---------+
| column1 | column2 | column3 |
+---------+---------+---------+
| A       | Z       |      11 |
+---------+---------+---------+

How to group by one column, take max for each group, and retain information of the other groups?

2 个答案: