SQL基础--分组查询

函数名	功能	备注
SUM	求和
AVG	求平均值
COUNT	统计数目	COUNT(*)与COUNT(列名) / COUNT(DISTINCT(列名))，后者不包括列值为NULL的行
MAX	分组内最大值	任何数据类型
MIN	分组内最小值	任何数据类型

字段为NULL，对分组函数计算结果的影响

函数名	影响	原因
AVG	求取的平均值不同	总和不变，个数减少
COUNT	统计数目	COUNT(*)与COUNT(列名) / COUNT(DISTINCT(列名))，后者不包括列值为NULL的行

分组函数具体到某个字段时，会跳过该字段为NULL的记录

SELECT AVG(NVL(commission_pct, 0)) FROM employees;
SELECT AVG(commission_pct) FROM employees;

SELECT中包含的非分组函数中的字段必须出现在GROUP BY 中

ROLLUP、CUBE、GROUPING、GROUPING SETS

ROLLUP分组规则：

select group_id,job,name,avg(salary) from group_test group by rollup(group_id, job,name);

PL/SQL Developer Export

	GROUP_ID	JOB	NAME	AVG(SALARY)
1	10	Coding	Bruce	1000
2	10	Coding		1000
3	10	Director	Hill	1000
4	10	Director		1000
5	10	Architect	Gideon	1000
6	10	Architect		1000
7	10	Programmer	Clair	1000
8	10	Programmer		1000
9	10			1000
10	20	Coding	Jason	2000
11	20	Coding		2000
12	20	Director	Michael	2000
13	20	Director		2000
14	20	Architect	Martin	2000
15	20	Architect		2000
16	20	Programmer	Joey	2000
17	20	Programmer		2000
18	20			2000
19	30	Coding	Rebecca	3000
20	30	Coding		3000
21	30	Director	Sabrina	3000
22	30	Director		3000
23	30	Architect	Richard	3000
24	30	Architect		3000
25	30	Programmer	Rex	3000
26	30	Programmer		3000
27	30			3000
28	40	Coding	Samuel	4000
29	40	Coding		4000
30	40	Director	Wendy	4000
31	40	Director		4000
32	40	Architect	Tina	4000
33	40	Architect		4000
34	40	Programmer	Susy	4000
35	40	Programmer		4000
36	40			4000
37				2500

从右往左，先按group_id和job分组，再在组内按group_id分组，然后不分组；

对于GROUP BY ROLLUP(C1，c2, c3,…,cn)：
从右往左，对c1…cn-1，cn, 依次减少一个进行分组，最后不分组统计；

CUBE分组规则：

select group_id,job,name,sum(salary) from group_test group by CUBE(group_id, job,name);

PL/SQL Developer Export

	GROUP_ID	JOB	NAME	SUM(SALARY)
1				40000
2			Rex	3000
3			Hill	1000
4			Joey	2000
5			Susy	4000
6			Tina	4000
7			Bruce	1000
8			Clair	1000
9			Jason	2000
10			Wendy	4000
11			Gideon	1000
12			Martin	2000
13			Samuel	4000
14			Michael	2000
15			Rebecca	3000
16			Richard	3000
17			Sabrina	3000
18		Coding		10000
19		Coding	Bruce	1000
20		Coding	Jason	2000
21		Coding	Samuel	4000
22		Coding	Rebecca	3000
23		Director		10000
24		Director	Hill	1000
25		Director	Wendy	4000
26		Director	Michael	2000
27		Director	Sabrina	3000
28		Architect		10000
29		Architect	Tina	4000
30		Architect	Gideon	1000
31		Architect	Martin	2000
32		Architect	Richard	3000
33		Programmer		10000
34		Programmer	Rex	3000
35		Programmer	Joey	2000
36		Programmer	Susy	4000
37		Programmer	Clair	1000
38	10			4000
39	10		Hill	1000
40	10		Bruce	1000
41	10		Clair	1000
42	10		Gideon	1000
43	10	Coding		1000
44	10	Coding	Bruce	1000
45	10	Director		1000
46	10	Director	Hill	1000
47	10	Architect		1000
48	10	Architect	Gideon	1000
49	10	Programmer		1000
50	10	Programmer	Clair	1000
51	20			8000
52	20		Joey	2000
53	20		Jason	2000
54	20		Martin	2000
55	20		Michael	2000
56	20	Coding		2000
57	20	Coding	Jason	2000
58	20	Director		2000
59	20	Director	Michael	2000
60	20	Architect		2000
61	20	Architect	Martin	2000
62	20	Programmer		2000
63	20	Programmer	Joey	2000
64	30			12000
65	30		Rex	3000
66	30		Rebecca	3000
67	30		Richard	3000
68	30		Sabrina	3000
69	30	Coding		3000
70	30	Coding	Rebecca	3000
71	30	Director		3000
72	30	Director	Sabrina	3000
73	30	Architect		3000
74	30	Architect	Richard	3000
75	30	Programmer		3000
76	30	Programmer	Rex	3000
77	40			16000
78	40		Susy	4000
79	40		Tina	4000
80	40		Wendy	4000
81	40		Samuel	4000
82	40	Coding		4000
83	40	Coding	Samuel	4000
84	40	Director		4000
85	40	Director	Wendy	4000
86	40	Architect		4000
87	40	Architect	Tina	4000
88	40	Programmer		4000
89	40	Programmer	Susy	4000

从右往左，先按name分组，再按job分组汇总，在按job分组内部在从右至左按name分组，最后按group分组时重复以上步骤

对于GROUP BY CUBE(C1，c2, c3,…,cn)：
从右往左，现不分组统计，在对ci分组，完成后按ci-1分组,在组内对ci…cn重复之前步骤

GROUPING函数：

没有被GROUPING返回1，被GROUPING返回0

GROUPING SETS:

按不同SET分组，再合并查询结果，如果SET之间有特殊包含关系，输出结果次序可能不同；

秒客网