例1:
hive -e"
select
type
,status
,count(1)
from
usr_info
where pt=‘2015-09-14‘
group by type,status
grouping sets ((type,status),( type),());
">one.txt
Grouping sets按照各种指定聚类汇总方式,如group by type,status
grouping sets ((type,status),( type),())
表示group by type,status union all group by type union all group by ()
得到
type status _c2
NULL NULL 69467
1 NULL 68216
1 1 63615
1 2 540
1 4 4061
2 NULL 891
2 1 873
2 2 18
3 NULL 360
3 1 340
3 4 20
例2:
hive -e"
select
type
,status
,count(1)
from
usr_info
where pt=‘2015-09-14‘
group by type,status with rollup;
">two.txt
group by type,status with rollup按照以type为主的固定聚类汇总方式,如同group by type,status grouping sets ((type,status),( type),()) ,不过形式已经固定了
表示group by type,status union all group by type union all group by ()
得到
Type status _c2
NULL NULL 69467
1 NULL 68216
1 1 63615
1 2 540
1 4 4061
2 NULL 891
2 1 873
2 2 18
3 NULL 360
3 1 340
3 4 20
例3:
hive -e"
select
type
,status
,count(1)
from
usr_info
where pt=‘2015-09-14‘
group by type,status with cube;
">three.txt
group by type,status with cube按照以type和status为主的固定聚类汇总方式,如同group by type,status grouping sets ((type,status),( type),(status),()) ,不过形式已经固定了
表示group by type,status union all group by type union all group by status union all group by ()
得到
Type status _c2
NULL NULL 69467
NULL 1 64828
NULL 2 558
NULL 4 4081
1 NULL 68216
1 1 63615
1 2 540
1 4 4061
2 NULL 891
2 1 873
2 2 18
3 NULL 360
3 1 340
3 4 20
例4:
hive -e"
select
type
,status
,grouping__id
,count(1)
from
usr_info
where pt=‘2015-09-14‘
group by type,status with cube;
">five.txt
type
,status
,grouping__id
grouping__id(两条横线)函数判断其参数是否参与了分组,如果参与则返回1,
如果没有参与了分组则返回0
而其多个参数的形式则将其每个参数进行grouping__id运算后返回的值拼成二进制后转换为十进制返回,
grouping_id(argn,...,arg2,arg1)=grouping_id(argn)*2^(n-1)+...+grouping_id(arg2)*2^1+grouping_id(arg1)*2^0(‘^‘表示幂运算)。
Hive中grouping__id不带参数,用法见例子。
得到
type status grouping__id _c3
NULL NULL 0 69467
NULL 1 2 64828
NULL 2 2 558
NULL 4 2 4081
1 NULL 1 68216
1 1 3 63615
1 2 3 540
1 4 3 4061
2 NULL 1 891
2 1 3 873
2 2 3 18
3 NULL 1 360
3 1 3 340
3 4 3 20