PostgreSQL ROLLUP 是group by 的子句，是生成多个分组集合的快捷功能。与Cube子句的差异是，rollup 不生成基于特定列所有可能的分组集合，生成分组集合为其子集。

ROLLUP假设输入列之间存在层次结构，从而生成有意义的所有分组集合。这就是为什么ROLLUP经常用于生成报表的小计和总计。

cube (c1,c2,c3)生成所有8中可能：

(c1, c2, c3)
(c1, c2)
(c2, c3)
(c1,c3)
(c1)
(c2)
(c3)
()

而ROLLUP(c1,c2,c3)仅生成4种分组集合，假设层级c1 > c2 > c3：

(c1, c2, c3)
(c1, c2)
(c1)
()

CREATE TABLE geeksforgeeks_courses(course_name VARCHAR NOT NULL,segment VARCHAR NOT NULL,quantity INT NOT NULL,PRIMARY KEY (course_name, segment)
);INSERT INTO geeksforgeeks_courses(course_name, segment, quantity)
VALUES('Data Structure in Python', 'Premium', 100),('Algorithm Design in Python', 'Basic', 200),('Data Structure in Java', 'Premium', 100),('Algorithm Design in Java', 'Basic', 300);

   select  course_name,segment,SUM (quantity)from  geeksforgeeks_coursesgroup by (course_name, segment) order by course_name,segment

 select  course_name,segment,SUM (quantity)from  geeksforgeeks_coursesgroup by (course_name, segment) union all select  course_name,NULL,SUM (quantity)from  geeksforgeeks_coursesgroup by (course_name) union all select  NULL,NULL,SUM (quantity)from  geeksforgeeks_coursesorder by course_name,segment

用如下的sql可以达到相同的效果

SELECTcourse_name,segment,SUM (quantity)
FROMgeeksforgeeks_courses
GROUP BYROLLUP (course_name, segment)
ORDER BYcourse_name,segment;

SELECTcoalesce(course_name,'所有课程'),coalesce(segment,'所有'),SUM (quantity)
FROMgeeksforgeeks_courses
GROUP BYROLLUP (course_name, segment)
ORDER BYcourse_name,segment;

SELECTcoalesce(course_name,'所有课程'),coalesce(segment,'所有'),SUM (quantity)
FROMgeeksforgeeks_courses
GROUP BYcube (course_name, segment)

cube 分成了(), course_name,segment,course_name+segment
而rollup分成了(),course_name,course_name+segment

按照自定义进行分组

SELECTcoalesce(course_name,'所有课程'),coalesce(segment,'所有'),SUM (quantity)
FROMgeeksforgeeks_courses
GROUP by GROUPING SETS ((course_name),(segment),(course_name,segment),())