MySQL8.0窗口函数实践及小结
MySQL8.0之前，做数据排名统计等相当痛苦，因为没有像Oracle、SQL SERVER 、PostgreSQL等其他数据库那样的窗口函数。但随着MySQL8.0中新增了窗口函数之后，针对这类统计就再也不是事了，本文就以常用的排序实例介绍MySQL的窗口函数。
1、准备工作
创建表及测试数据
mysql> use testdb; Database changed /* 创建表 */ mysql> create table tb_score(id int primary key auto_increment,stu_no varchar(10),course varchar(50),score decimal(4,1)，key idx_stuNo_course(stu_no,course)); Query OK, 0 rows affected (0.03 sec) mysql> show tables; +------------------+ | Tables_in_testdb | +------------------+ | tb_score | +------------------+ /* 新增一批测试数据 */ mysql> insert into tb_score(stu_no,course,score)values('2020001','mysql',90),('2020001','C++',85),('2020003','English',100),('2020002','mysql',50),('2020002','C++',70),('2020002','English',99); Query OK, 6 rows affected (0.00 sec) Records: 6 Duplicates: 0 Warnings: 0 mysql> insert into tb_score(stu_no,course,score)values('2020003','mysql',78),('2020003','C++',81),('2020003','English',80),('2020004','mysql',80),('2020004','C++',60),('2020004','English',100); Query OK, 6 rows affected (0.01 sec) Records: 6 Duplicates: 0 Warnings: 0 mysql> insert into tb_score(stu_no,course,score)values('2020005','mysql',98),('2020005','C++',96),('2020005','English',70),('2020006','mysql',60),('2020006','C++',90),('2020006','English',70); Query OK, 6 rows affected (0.01 sec) Records: 6 Duplicates: 0 Warnings: 0 mysql> insert into tb_score(stu_no,course,score)values('2020007','mysql',50),('2020007','C++',66),('2020007','English',76),('2020008','mysql',90),('2020008','C++',69),('2020008','English',86); Query OK, 6 rows affected (0.01 sec) Records: 6 Duplicates: 0 Warnings: 0 mysql> insert into tb_score(stu_no,course,score)values('2020009','mysql',70),('2020009','C++',66),('2020009','English',86),('2020010','mysql',75),('2020010','C++',76),('2020010','English',81); Query OK, 6 rows affected (0.01 sec) Records: 6 Duplicates: 0 Warnings: 0 mysql> insert into tb_score(stu_no,course,score)values('2020011','mysql',90),('2020012','C++',85),('2020011','English',84),('2020012','English',75),('2020013','C++',96),('2020013','English',88); Query OK, 6 rows affected (0.01 sec) Records: 6 Duplicates: 0 Warnings: 0
2、统计每门课程分数的排名
根据每门课程的分数从高到低进行排名，此时，会出现分数相同时怎么处理的问题，下面就根据不同的窗口函数来处理不同场景的需求
ROW_NUMBER
由结果可以看出，分数相同时按照学号顺序进行排名
mysql> select stu_no,course,score, row_number()over(partition by course order by score desc ) rn -> from tb_score; +---------+---------+-------+----+ | stu_no | course | score | rn | +---------+---------+-------+----+ | 2020005 | C++ | 96.0 | 1 | | 2020013 | C++ | 96.0 | 2 | | 2020006 | C++ | 90.0 | 3 | | 2020001 | C++ | 85.0 | 4 | | 2020012 | C++ | 85.0 | 5 | | 2020003 | C++ | 81.0 | 6 | | 2020010 | C++ | 76.0 | 7 | | 2020002 | C++ | 70.0 | 8 | | 2020008 | C++ | 69.0 | 9 | | 2020007 | C++ | 66.0 | 10 | | 2020009 | C++ | 66.0 | 11 | | 2020004 | C++ | 60.0 | 12 | | 2020003 | English | 100.0 | 1 | | 2020004 | English | 100.0 | 2 | | 2020002 | English | 99.0 | 3 | | 2020013 | English | 88.0 | 4 | | 2020008 | English | 86.0 | 5 | | 2020009 | English | 86.0 | 6 | | 2020011 | English | 84.0 | 7 | | 2020010 | English | 81.0 | 8 | | 2020003 | English | 80.0 | 9 | | 2020007 | English | 76.0 | 10 | | 2020012 | English | 75.0 | 11 | | 2020005 | English | 70.0 | 12 | | 2020006 | English | 70.0 | 13 | | 2020005 | mysql | 98.0 | 1 | | 2020001 | mysql | 90.0 | 2 | | 2020008 | mysql | 90.0 | 3 | | 2020011 | mysql | 90.0 | 4 | | 2020004 | mysql | 80.0 | 5 | | 2020003 | mysql | 78.0 | 6 | | 2020010 | mysql | 75.0 | 7 | | 2020009 | mysql | 70.0 | 8 | | 2020006 | mysql | 60.0 | 9 | | 2020002 | mysql | 50.0 | 10 | | 2020007 | mysql | 50.0 | 11 | +---------+---------+-------+----+ 36 rows in set (0.00 sec)
DENSE_RANK
为了让分数相同时排名也相同，则可以使用DENSE_RANK函数，结果如下：
mysql> select stu_no,course,score, DENSE_RANK()over(partition by course order by score desc ) rn -> from tb_score ; +---------+---------+-------+----+ | stu_no | course | score | rn | +---------+---------+-------+----+ | 2020005 | C++ | 96.0 | 1 | | 2020013 | C++ | 96.0 | 1 | | 2020006 | C++ | 90.0 | 2 | | 2020001 | C++ | 85.0 | 3 | | 2020012 | C++ | 85.0 | 3 | | 2020003 | C++ | 81.0 | 4 | | 2020010 | C++ | 76.0 | 5 | | 2020002 | C++ | 70.0 | 6 | | 2020008 | C++ | 69.0 | 7 | | 2020007 | C++ | 66.0 | 8 | | 2020009 | C++ | 66.0 | 8 | | 2020004 | C++ | 60.0 | 9 | | 2020003 | English | 100.0 | 1 | | 2020004 | English | 100.0 | 1 | | 2020002 | English | 99.0 | 2 | | 2020013 | English | 88.0 | 3 | | 2020008 | English | 86.0 | 4 | | 2020009 | English | 86.0 | 4 | | 2020011 | English | 84.0 | 5 | | 2020010 | English | 81.0 | 6 | | 2020003 | English | 80.0 | 7 | | 2020007 | English | 76.0 | 8 | | 2020012 | English | 75.0 | 9 | | 2020005 | English | 70.0 | 10 | | 2020006 | English | 70.0 | 10 | | 2020005 | mysql | 98.0 | 1 | | 2020001 | mysql | 90.0 | 2 | | 2020008 | mysql | 90.0 | 2 | | 2020011 | mysql | 90.0 | 2 | | 2020004 | mysql | 80.0 | 3 | | 2020003 | mysql | 78.0 | 4 | | 2020010 | mysql | 75.0 | 5 | | 2020009 | mysql | 70.0 | 6 | | 2020006 | mysql | 60.0 | 7 | | 2020002 | mysql | 50.0 | 8 | | 2020007 | mysql | 50.0 | 8 | +---------+---------+-------+----+ 36 rows in set (0.00 sec)
RANK
DENSE_RANK的结果是分数相同时排名相同了，但是下一个名次是紧接着上一个名次的，如果2个并列的第1之后，下一个我想是第3名，则可以使用RANK函数实现
mysql> select stu_no,course,score, rank()over(partition by course order by score desc ) rn -> from tb_score; +---------+---------+-------+----+ | stu_no | course | score | rn | +---------+---------+-------+----+ | 2020005 | C++ | 96.0 | 1 | | 2020013 | C++ | 96.0 | 1 | | 2020006 | C++ | 90.0 | 3 | | 2020001 | C++ | 85.0 | 4 | | 2020012 | C++ | 85.0 | 4 | | 2020003 | C++ | 81.0 | 6 | | 2020010 | C++ | 76.0 | 7 | | 2020002 | C++ | 70.0 | 8 | | 2020008 | C++ | 69.0 | 9 | | 2020007 | C++ | 66.0 | 10 | | 2020009 | C++ | 66.0 | 10 | | 2020004 | C++ | 60.0 | 12 | | 2020003 | English | 100.0 | 1 | | 2020004 | English | 100.0 | 1 | | 2020002 | English | 99.0 | 3 | | 2020013 | English | 88.0 | 4 | | 2020008 | English | 86.0 | 5 | | 2020009 | English | 86.0 | 5 | | 2020011 | English | 84.0 | 7 | | 2020010 | English | 81.0 | 8 | | 2020003 | English | 80.0 | 9 | | 2020007 | English | 76.0 | 10 | | 2020012 | English | 75.0 | 11 | | 2020005 | English | 70.0 | 12 | | 2020006 | English | 70.0 | 12 | | 2020005 | mysql | 98.0 | 1 | | 2020001 | mysql | 90.0 | 2 | | 2020008 | mysql | 90.0 | 2 | | 2020011 | mysql | 90.0 | 2 | | 2020004 | mysql | 80.0 | 5 | | 2020003 | mysql | 78.0 | 6 | | 2020010 | mysql | 75.0 | 7 | | 2020009 | mysql | 70.0 | 8 | | 2020006 | mysql | 60.0 | 9 | | 2020002 | mysql | 50.0 | 10 | | 2020007 | mysql | 50.0 | 10 | +---------+---------+-------+----+ 36 rows in set (0.01 sec)
这样就实现了各种排序需求。
NTILE
NTILE函数的作用是对每个分组排名后，再将对应分组分成N个小组，例如
mysql> select stu_no,course,score, rank()over(partition by course order by score desc )rn,NTILE(2)over(partition by course order by score desc ) rn_group from tb_score; +---------+---------+-------+----+----------+ | stu_no | course | score | rn | rn_group | +---------+---------+-------+----+----------+ | 2020005 | C++ | 96.0 | 1 | 1 | | 2020013 | C++ | 96.0 | 1 | 1 | | 2020006 | C++ | 90.0 | 3 | 1 | | 2020001 | C++ | 85.0 | 4 | 1 | | 2020012 | C++ | 85.0 | 4 | 1 | | 2020003 | C++ | 81.0 | 6 | 1 | | 2020010 | C++ | 76.0 | 7 | 2 | | 2020002 | C++ | 70.0 | 8 | 2 | | 2020008 | C++ | 69.0 | 9 | 2 | | 2020007 | C++ | 66.0 | 10 | 2 | | 2020009 | C++ | 66.0 | 10 | 2 | | 2020004 | C++ | 60.0 | 12 | 2 | | 2020003 | English | 100.0 | 1 | 1 | | 2020004 | English | 100.0 | 1 | 1 | | 2020002 | English | 99.0 | 3 | 1 | | 2020013 | English | 88.0 | 4 | 1 | | 2020008 | English | 86.0 | 5 | 1 | | 2020009 | English | 86.0 | 5 | 1 | | 2020011 | English | 84.0 | 7 | 1 | | 2020010 | English | 81.0 | 8 | 2 | | 2020003 | English | 80.0 | 9 | 2 | | 2020007 | English | 76.0 | 10 | 2 | | 2020012 | English | 75.0 | 11 | 2 | | 2020005 | English | 70.0 | 12 | 2 | | 2020006 | English | 70.0 | 12 | 2 | | 2020005 | mysql | 98.0 | 1 | 1 | | 2020001 | mysql | 90.0 | 2 | 1 | | 2020008 | mysql | 90.0 | 2 | 1 | | 2020011 | mysql | 90.0 | 2 | 1 | | 2020004 | mysql | 80.0 | 5 | 1 | | 2020003 | mysql | 78.0 | 6 | 1 | | 2020010 | mysql | 75.0 | 7 | 2 | | 2020009 | mysql | 70.0 | 8 | 2 | | 2020006 | mysql | 60.0 | 9 | 2 | | 2020002 | mysql | 50.0 | 10 | 2 | | 2020007 | mysql | 50.0 | 10 | 2 | +---------+---------+-------+----+----------+ 36 rows in set (0.01 sec)
3、窗口函数小结
MySQL中还有许多其他的窗口函数，本文列举一些，大家可以自行测试
| 类别 | 函数 | 说明 |
| 排序 | ROW_NUMBER | 为表中的每一行分配一个序号，可以指定分组（也可以不指定）及排序字段 |
| DENSE_RANK | 根据排序字段为每个分组中的每一行分配一个序号。 排名值相同时，序号相同，序号中没有间隙（1,1,2,3这种） | |
| RANK | 根据排序字段为每个分组中的每一行分配一个序号。 排名值相同时，序号相同，但序号中存在间隙（1,1,3,4这种） | |
| NTILE | 根据排序字段为每个分组中根据指定字段的排序再分成对应的组 | |
| 分布 | PERCENT_RANK | 计算各分组或结果集中行的百分数等级 |
| CUME_DIST | 计算某个值在一组有序的数据中累计的分布 | |
| 前后 | LEAD | 返回分组中当前行之后的第N行的值。如果不存在对应行，则返回NULL。比如N=1时，第一名对应的值是第二名的，最后一名结果是NULL |
| LAG | 返回分组中当前行之前的第N行的值。如果不存在对应行，则返回NULL。比如N=1时，第一名对应的值是是NUL，最后一名结果是倒数第2的值 | |
| 首尾中 | FIRST_VALUE | 返回每个分组中第一名对应的字段（或表达式）的值，例如本文中可以是第一名的分数、学号等任意字段的值 |
| LAST_VALUE | 返回每个分组中最后一名对应的字段（或表达式）的值，例如本文中可以是最后一名的分数、学号等任意字段的值 | |
| NTH_VALUE |
返回每个分组中排名第N的对应字段（或表达式）的值，但小于N的行对应的值是NULL |
MySQL中主要的窗口函数先总结这么多，建议还是得动手实践一番。另外，MySQL5.7及之前版本的排序方式的实现很多人已总结，也建议实操一番。
想了解更多内容或参与技术交流可以关注微信公众号【数据库干货铺】或进技术交流群沟通。
