相信我们经常会遇到这样的场景:想要了解双十一天猫购买化妆品的人员中平均消费额度是多少(这可能有利于对商品价格区间的定位);或者不同年龄段的化妆品消费占比是多少(这可能有助于对商品备货量的预估)。
这个时候就要用到分组查询,分组查询的目的是为了把数据分成多个逻辑组(购买化妆品的人员是一个组,不同年龄段购买化妆品的人员也是组),并对每个组进行聚合计算的过程:。
分组查询的语法格式如下:
select cname, group_fun,... from tname [where condition]
group by group_expression [having group_condition];
说明一下:
1、group_fun 代表聚合函数,是指对分组的数据进行聚合计算的函数。
2、group_expression 代表分组表达式,允许多个,多个之间使用逗号隔开。
3、group_condition 分组之后,再对分组后的数据进行条件过滤的过程。
4、分组语法中,select后面出现的字段 要么是group by后面的字段,要么是聚合函数的列,其他类型会报异常,我们下面的内容中会详细说明。
说分组之前,先来看看聚合函数,聚合函数是分组查询语法格式中重要的一部分。我们经常需要汇总数据而不用把它们实际检索出来,所以MySQL提供了专门的函数。使用这些函数,可用于计算我们需要的数据,以便分析和生成报表。
聚合函数有以下几种。
函数 | 说明 |
AVG() | 返回指定字段的平均值 |
COUNT() | 返回查询结果行数 |
MAX() | 返回指定字段的最大值 |
MIN() | 返回指定字段的最小值 |
SUM() | 返回指定字段的求和值 |
AVG()通过对表中行数计数并计算特定列值之和,求得该列的平均值。 AVG()可用来返回所有列的平均值,也可以用来返回特定列或行的平均值。
下面示例返回用户表中用户的平均年龄:
mysql> select * from user2;
+----+--------+------+----------+-----+
| id | name | age | address | sex |
+----+--------+------+----------+-----+
| 1 | brand | 21 | fuzhou | 1 |
| 2 | helen | 20 | quanzhou | 0 |
| 3 | sol | 21 | xiamen | 0 |
| 4 | weng | 33 | guizhou | 1 |
| 5 | selina | 25 | NULL | 0 |
| 6 | anny | 23 | shanghai | 0 |
| 7 | annd | 24 | shanghai | 1 |
| 8 | sunny | NULL | guizhou | 0 |
+----+--------+------+----------+-----+
8 rows in set
mysql> select avg(age) from user2;
+----------+
| avg(age) |
+----------+
| 23.8571 |
+----------+
1 row in set
注意点:
1、AVG()只能用来确定特定数值列的平均值 。
2、AVG()函数忽略列值为NULL的行,所以上图中age值累加之后是除以7,而不是除以8。
COUNT()函数进行计数。 可以用COUNT()确定表中符合条件的行的数目。
count 有 count(*)、count(具体字段)、count(常量) 三种方式来体现 下面 演示了count(*) 和 count(cname)的用法。
mysql> select * from user2;
+----+--------+------+----------+-----+
| id | name | age | address | sex |
+----+--------+------+----------+-----+
| 1 | brand | 21 | fuzhou | 1 |
| 2 | helen | 20 | quanzhou | 0 |
| 3 | sol | 21 | xiamen | 0 |
| 4 | weng | 33 | guizhou | 1 |
| 5 | selina | 25 | NULL | 0 |
| 6 | anny | 23 | shanghai | 0 |
| 7 | annd | 24 | shanghai | 1 |
| 8 | sunny | NULL | guizhou | 0 |
+----+--------+------+----------+-----+
8 rows in set
mysql> select count(*) from user2 where sex=0;
+----------+
| count(*) |
+----------+
| 5 |
+----------+
1 row in set
mysql> select count(age) from user2 where sex=0;
+------------+
| count(age) |
+------------+
| 4 |
+------------+
1 row in set
可以看到,都是取出女生的用户数量,count(*) 比 count(age) 多一个,那是因为age中包含null值。
所以:如果指定列名,则指定列的值为空的行被COUNT()函数忽略,但如果COUNT()函数中用的是星号( *),则不忽略。
关于count 可以看另一篇,详细分析了几种count的使用和性能比较: SELECT COUNT 小结
MAX()返回指定列中的最大值,MIN()返回指定列中的最小值。
mysql> select * from user2;
+----+--------+------+----------+-----+
| id | name | age | address | sex |
+----+--------+------+----------+-----+
| 1 | brand | 21 | fuzhou | 1 |
| 2 | helen | 20 | quanzhou | 0 |
| 3 | sol | 21 | xiamen | 0 |
| 4 | weng | 33 | guizhou | 1 |
| 5 | selina | 25 | NULL | 0 |
| 6 | anny | 23 | shanghai | 0 |
| 7 | annd | 24 | shanghai | 1 |
| 8 | sunny | NULL | guizhou | 0 |
+----+--------+------+----------+-----+
8 rows in set
mysql> select max(age),min(age) from user2;
+----------+----------+
| max(age) | min(age) |
+----------+----------+
| 33 | 20 |
+----------+----------+
1 row in set
注意:同样的,MAX()、MIN()函数忽略列值为NULL的行。
SUM()用来返回指定列值的和(总计) ,下面返回了所有年龄的总和,同样的,忽略了null的值
mysql> select * from user2;
+----+--------+------+----------+-----+
| id | name | age | address | sex |
+----+--------+------+----------+-----+
| 1 | brand | 21 | fuzhou | 1 |
| 2 | helen | 20 | quanzhou | 0 |
| 3 | sol | 21 | xiamen | 0 |
| 4 | weng | 33 | guizhou | 1 |
| 5 | selina | 25 | NULL | 0 |
| 6 | anny | 23 | shanghai | 0 |
| 7 | annd | 24 | shanghai | 1 |
| 8 | sunny | NULL | guizhou | 0 |
+----+--------+------+----------+-----+
8 rows in set
mysql> select sum(age) from user2;
+----------+
| sum(age) |
+----------+
| 167 |
+----------+
1 row in set
数据准备,假设我们有一个订货单表如下(记载用户的订单金额和下单时间):
mysql> select * from t_order;
+---------+-----+-------+--------+---------------------+------+
| orderid | uid | uname | amount | time | year |
+---------+-----+-------+--------+---------------------+------+
| 20 | 1 | brand | 91.23 | 2018-08-20 17:22:21 | 2018 |
| 21 | 1 | brand | 87.54 | 2019-07-16 09:21:30 | 2019 |
| 22 | 1 | brand | 166.88 | 2019-04-04 12:23:55 | 2019 |
| 23 | 2 | helyn | 93.73 | 2019-09-15 10:11:11 | 2019 |
| 24 | 2 | helyn | 102.32 | 2019-01-08 17:33:25 | 2019 |
| 25 | 2 | helyn | 106.06 | 2019-12-24 12:25:25 | 2019 |
| 26 | 2 | helyn | 73.42 | 2020-04-03 17:16:23 | 2020 |
| 27 | 3 | sol | 55.55 | 2019-08-05 19:16:23 | 2019 |
| 28 | 3 | sol | 69.96 | 2020-09-16 19:23:16 | 2020 |
| 29 | 4 | weng | 199.99 | 2020-06-08 19:55:06 | 2020 |
+---------+-----+-------+--------+---------------------+------+
10 rows in set
即对于某个字段进行分组,比如针对用户进行分组,输出他们的用户Id,订单数量和总额:
mysql> select uid,count(uid),sum(amount) from t_order group by uid;
+-----+------------+-------------+
| uid | count(uid) | sum(amount) |
+-----+------------+-------------+
| 1 | 3 | 345.65 |
| 2 | 4 | 375.53 |
| 3 | 2 | 125.51 |
| 4 | 1 | 199.99 |
+-----+------------+-------------+
4 rows in set
即对于多个字段进行分组,比如针对用户进行分组,再对他们不同年份的订单数据进行分组,输出订单数量和消费总额:
mysql> select uid,count(uid) as nums,sum(amount) as totalamount,year from t_order group by uid,year;
+-----+------+-------------+------+
| uid | nums | totalamount | year |
+-----+------+-------------+------+
| 1 | 1 | 91.23 | 2018 |
| 1 | 2 | 254.42 | 2019 |
| 2 | 3 | 302.11 | 2019 |
| 2 | 1 | 73.42 | 2020 |
| 3 | 1 | 55.55 | 2019 |
| 3 | 1 | 69.96 | 2020 |
| 4 | 1 | 199.99 | 2020 |
+-----+------+-------------+------+
7 rows in set
这个很简单,就是再分组(group by)之前通过where关键字进行条件过滤,取出我们需要的数据,假设我们只要列出2019年8月之后的数据,源数据只有6条合格的,有两条年份一样被分组的:
mysql> select uid,count(uid) as nums,sum(amount) as totalamount,year from t_order where time > '2019-08-01' group by uid,year;
+-----+------+-------------+------+
| uid | nums | totalamount | year |
+-----+------+-------------+------+
| 2 | 2 | 199.79 | 2019 |
| 2 | 1 | 73.42 | 2020 |
| 3 | 1 | 55.55 | 2019 |
| 3 | 1 | 69.96 | 2020 |
| 4 | 1 | 199.99 | 2020 |
+-----+------+-------------+------+
5 rows in set
有时候我们需要再分组之后再对数据进行过滤,这时候就需要使用having关键字进行数据过滤,再上述条件下,我们需要取出消费次数超过一次的数据:
mysql> select uid,count(uid) as nums,sum(amount) as totalamount,year from t_order where time > '2019-08-01' group by uid,year having nums>1;
+-----+------+-------------+------+
| uid | nums | totalamount | year |
+-----+------+-------------+------+
| 2 | 2 | 199.79 | 2019 |
+-----+------+-------------+------+
1 row in set
这边需要注意区分where和having:
where是在分组(聚合)前对记录进行筛选,而having是在分组结束后的结果里筛选,最后返回过滤后的结果。
可以把having理解为两级查询,即含having的查询操作先获得不含having子句时的sql查询结果表,然后在这个结果表上使用having条件筛选出符合的记录,最后返回这些记录,因此,having后是可以跟聚合函数的,并且这个聚集函数不必与select后面的聚集函数相同。
order条件接在group by后面,也就是统计出每个用户的消费总额和消费次数后,对用户的消费总额进行降序排序的过程。
mysql> select uid,count(uid) as nums,sum(amount) as totalamount from t_order group by uid;
+-----+------+-------------+
| uid | nums | totalamount |
+-----+------+-------------+
| 1 | 3 | 345.65 |
| 2 | 4 | 375.53 |
| 3 | 2 | 125.51 |
| 4 | 1 | 199.99 |
+-----+------+-------------+
4 rows in set
mysql> select uid,count(uid) as nums,sum(amount) as totalamount from t_order group by uid order by totalamount desc;
+-----+------+-------------+
| uid | nums | totalamount |
+-----+------+-------------+
| 2 | 4 | 375.53 |
| 1 | 3 | 345.65 |
| 4 | 1 | 199.99 |
| 3 | 2 | 125.51 |
+-----+------+-------------+
4 rows in set
limit限制关键字一般放在语句的最末尾,比如基于我们上面的搜索,我们再limit 1,只取出消费额最高的那条,其他跳过。
mysql> select uid,count(uid) as nums,sum(amount) as totalamount from t_order group by uid order by totalamount desc limit 1;
+-----+------+-------------+
| uid | nums | totalamount |
+-----+------+-------------+
| 2 | 4 | 375.53 |
+-----+------+-------------+
1 row in set
我们看到上面那我们用了 where、group by、having、order by、limit这些关键字,如果一起使用,他们是有先后顺序,顺序错了会导致异常,语法格式如下:
select cname from tname
where [原表查询条件]
group by [分组表达式]
having [分组过滤条件]
order by [排序条件]
limit [offset,] count;
mysql> select uid,count(uid) as nums,sum(amount) as totalamount from t_order where time > '2019-08-01' group by uid having totalamount>100 order by totalamount desc limit 1;
+-----+------+-------------+
| uid | nums | totalamount |
+-----+------+-------------+
| 2 | 3 | 273.21 |
+-----+------+-------------+
1 row in set
1、分组语法中,select后面出现的字段 要么是group by后面的字段,要么是聚合函数的列,其他类型会报异常:可以自己试试。
2、分组关键字的执行顺序:where、group by、having、order by、limit,顺序不能调换,否则会报异常:可以自己试试。