group by ，grouping sets ，cube ，rollup区别

数据准备

创建表

create table if not exists cookies(
     cookieid     string,
     cookie_day   string,
     cookie_month  string,
     cookie_status  bigint
)comment 'group测试'
PARTITIONED BY (ds string) lifecycle 1000;

数据

insert into cookies PARTITION(ds = '20210401')
values ('10001','2021-2-12','2021-2','1'),
       ('10002','2021-2-12','2021-2','2'),

       ('10003','2021-2-13','2021-2','1'),
       ('10004','2021-2-13','2021-2','2'),
       ('10005','2021-2-13','2021-2','3'),

       ('10006','2021-2-14','2021-2','1'),
       ('10007','2021-2-14','2021-2','2'),
       ('10008','2021-2-14','2021-2','3'),
       ('10009','2021-2-14','2021-2','4'),
       ('10010','2021-2-14','2021-2','5'),
       ('10011','2021-2-14','2021-2','6'),
       ('10012','2021-2-14','2021-2','7'),

       ('10013','2021-3-1','2021-3','1'),

       ('10014','2021-3-2','2021-3','1'),
       ('10015','2021-3-2','2021-3','2'),
       ('10016','2021-3-2','2021-3','3'),
       ('10017','2021-3-2','2021-3','4'),
       ('10018','2021-3-2','2021-3','5'),
       ('10019','2021-3-2','2021-3','6'),

       ('10020','2021-3-3','2021-3','1'),
       ('10021','2021-3-3','2021-3','2'),
       ('10022','2021-3-3','2021-3','3'),

       ('10020','2021-3-4','2021-3','1'),
       ('10021','2021-3-4','2021-3','2'),
       ('10022','2021-3-4','2021-3','3'),
       ('10023','2021-3-4','2021-3','4'),
       ('10024','2021-3-4','2021-3','5'),

       ('10025','2021-4-5','2021-4','1'),
       ('10026','2021-4-5','2021-4','2'),
       ('10027','2021-4-5','2021-4','3'),
       ('10028','2021-4-5','2021-4','4'),
       ('10029','2021-4-5','2021-4','5'),
       ('10030','2021-4-5','2021-4','6'),
       ('10031','2021-4-5','2021-4','7'),
       ('10032','2021-4-5','2021-4','8'),
       ('10033','2021-4-5','2021-4','9');


alter table daole_cookies_log drop partition (ds='20210401');

select * from daole_cookies_log;

（1）group by

select cookie_month,cookie_date,count(cookie_id) as total
  from cookies
 where ds = '20210401'
 group by cookie_month,cookie_date

结果：

（2）grouping sets

在一个GROUP BY查询中，根据不同的维度组合进行聚合，等价于将不同维度的GROUP BY结果集进行UNION ALL。

group by A,B,C
grouping sets (A,B,C)

• （A）
• （B）
• （C）

group by A,B,
grouping sets (A,B,(A,B))

• （A）
• （B）
• （A）（B）

例如：

select
       cookie_month,
       cookie_date,
       count(cookie_id) as uv
 from cookies
group by cookie_month,cookie_date
grouping sets (cookie_month,cookie_date)

等价于：

SELECT cookie_month,NULL,COUNT(cookie_id) AS uv 
  FROM cookies
 GROUP BY cookie_month
 UNION ALL
SELECT NULL,cookie_date,COUNT(cookie_id) AS uv 
  FROM cookies
 GROUP BY cookie_date

再比如：

 select cookie_month,cookie_date,count(cookie_id) as total
  from cookies
 where ds = '20210401'
 group by cookie_month,cookie_date
 grouping sets (cookie_month,
                cookie_date,
                (cookie_month,cookie_date)
                );

等价于：

SELECT cookie_month,NULL,COUNT(cookie_id) AS uv 
  FROM cookies GROUP BY cookie_month
 UNION ALL
SELECT NULL,cookie_date,COUNT(cookieid) AS uv
  FROM cookies GROUP BY cookie_date
UNION ALL
SELECT cookie_month,cookie_date,COUNT(cookieid) AS uv
  FROM cookies GROUP BY cookie_month,cookie_date

（3）cube

group by A,B,C with cube

• （A、B、C）进行group by
• （A） 进行group by
• （B） 进行group by
• （C） 进行group by
• （A）（B）进行group by
• （A）（C）进行group by
• （B）（C）进行group by
• 全表sum

 select cookie_month,cookie_date,count(cookie_id) as total
  from cookies
 where ds = '20210401'
 group by cookie_month,cookie_date
 with CUBE ;

（4）rollup

rollup 是根据维度在数据结果集中进行的聚合操作。

group by A,B,C with rollup

首先会对(A、B、C)进行group by，然后对(A、B)进行group by，然后是(A)进行group by，最后对全表进行group by操作。