PostgreSQL 分组集合新功能（GROUPING SETS,CUBE,ROLLUP）

本文主要是介绍PostgreSQL 分组集合新功能（GROUPING SETS,CUBE,ROLLUP），希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

PostgreSQL 分组集合新功能（GROUPING SETS,CUBE,ROLLUP）

实验环境

操作系统：windows 10 家庭中文版数据库系统： PostgreSQL 9.6.2

说明

postgresql从9.5版本开始新加入了group by的分组集合功能，提供了GROUPING SETS,CUBE,ROLLUP参数，使用方式与oracle完全一致，下面是实际测试说明

构建测试环境

创建表t并插入测试数据：

create table t(id int,name varchar(20),class int,score int);insert into t values(1,'math',1,90);
insert into t values(2,'math',2,80);
insert into t values(3,'math',1,70);
insert into t values(4,'chinese',2,60);
insert into t values(5,'chinese',1,50);
insert into t values(6,'chinese',2,60);
insert into t values(7,'physical',1,70);
insert into t values(8,'physical',2,80);
insert into t values(9,'physical',1,90);

结果：

test=# select * from t;id |   name   | class | score
----+----------+-------+-------1 | math     |     1 |    902 | math     |     2 |    803 | math     |     1 |    704 | chinese  |     2 |    605 | chinese  |     1 |    506 | chinese  |     2 |    607 | physical |     1 |    708 | physical |     2 |    809 | physical |     1 |    90
(9 行记录)

普通的group by

根据name和class字段求和：

test=# select name,class,sum(score)
test-# from t
test-# group by name,class
test-# order by name,class
test-# ;name   | class | sum
----------+-------+-----chinese  |     1 |  50chinese  |     2 | 120math     |     1 | 160math     |     2 |  80physical |     1 | 160physical |     2 |  80
(6 行记录)

grouping set

GROUPING SETS的每个子列表可以指定零个或多个列或表达式，并且与其直接在GROUP BY子句中的解释方式相同。一个空的分组集合意味着所有的行都被聚合到一个组中（即使没有输入行存在，也是输出）。

test=# select name,class,sum(score)
test-# from t
test-# group by grouping sets((name),(class),())
test-# order by name,class
test-# ;name   | class | sum
----------+-------+-----chinese  |       | 170math     |       | 240physical |       | 240|     1 | 370|     2 | 280|       | 650
(6 行记录)

顺带一提，默认的group by语句相当于grouping set在grouping set后的参数填上所有group by的字段。如下：

test=# select name,class,sum(score)
test-# from t
test-# group by grouping sets((name,class))
test-# order by name,class
test-# ;name   | class | sum
----------+-------+-----chinese  |     1 |  50chinese  |     2 | 120math     |     1 | 160math     |     2 |  80physical |     1 | 160physical |     2 |  80
(6 行记录)

与不使用grouping set语句时的结果完全相同

rollup

* rollup((a),(b),(c))等价于grouping sets((a,b,c),(a,b),(a),()) *

test=# select name,class,sum(score)
test-# from t
test-# group by rollup((name),(class))
test-# order by name,class
test-# ;name   | class | sum
----------+-------+-----chinese  |     1 |  50chinese  |     2 | 120chinese  |       | 170math     |     1 | 160math     |     2 |  80math     |       | 240physical |     1 | 160physical |     2 |  80physical |       | 240|       | 650
(10 行记录)

等价于：

grouping sets((name,class),(name),())

cube

* cube((a),(b),(c))等价于grouping sets((a,b,c),(a,b),(a,c),(a),(b,c),(b),(c),()) *

test=# select name,class,sum(score)
test-# from t
test-# group by cube((name),(class))
test-# order by name,class
test-# ;name   | class | sum
----------+-------+-----chinese  |     1 |  50chinese  |     2 | 120chinese  |       | 170math     |     1 | 160math     |     2 |  80math     |       | 240physical |     1 | 160physical |     2 |  80physical |       | 240|     1 | 370|     2 | 280|       | 650
(12 行记录)

等价于：

grouping sets((name,class),(name),(class),())

实际应用

我遇到一个需求，需要在分组统计总和之外附加所有组的总和，命名为total：

test=# select coalesce(name,'total') as name,
test-# coalesce(class,0) as class,
test-# coalesce(sum(score),0) as sum_score,
test-# coalesce(round(avg(score),2),0) as avg_score
test-# from t
test-# group by grouping sets((name,class),())
test-# order by name,class
test-# ;name   | class | sum_score | avg_score
----------+-------+-----------+-----------chinese  |     1 |        50 |     50.00chinese  |     2 |       120 |     60.00math     |     1 |       160 |     80.00math     |     2 |        80 |     80.00physical |     1 |       160 |     80.00physical |     2 |        80 |     80.00total    |     0 |       650 |     72.22
(7 行记录)