MySQL窗口函数实现榜单排名

来源：jb51　　时间：2023/4/12 11:14:30　　对本文有异议

INSERT INTO praise_record (name, praise_num) VALUES ('John', 5);
INSERT INTO praise_record (name, praise_num) VALUES ('Jane', 3);
INSERT INTO praise_record (name, praise_num) VALUES ('Bob', 10);
INSERT INTO praise_record (name, praise_num) VALUES ('Alice', 3);
INSERT INTO praise_record (name, praise_num) VALUES ('David', 7);
INSERT INTO praise_record (name, praise_num) VALUES ('oct', 7);

然后就可以开始实现我们的需求：返回点赞的榜单，并返回排名

rank()

使用rank()函数返回点赞的榜单， rank() over()

## 注意这里返回的rank字段要用反引号包起来
select name, praise_num, rank() over (order by praise_num desc) as `rank` from praise_record;
+-------+------------+------+
| name  | praise_num | rank |
+-------+------------+------+
| Bob   |         10 |    1 |
| David |          7 |    2 |
| oct   |          7 |    2 |
| John  |          5 |    4 |
| Jane  |          3 |    5 |
| Alice |          3 |    5 |
+-------+------------+------+

可以看到使用rank()函数的时候相同的点赞数会返回相同的排名，排名会产生跳跃，最终的排名不是连续的

dense_rank()

使用dense_rank()函数返回点赞的榜单， dense_rank() over()

 select name, praise_num, dense_rank() over (order by praise_num desc) as `rank` from praise_record;
 
 +-------+------------+------+
| name  | praise_num | rank |
+-------+------------+------+
| Bob   |         10 |    1 |
| David |          7 |    2 |
| oct   |          7 |    2 |
| John  |          5 |    3 |
| Jane  |          3 |    4 |
| Alice |          3 |    4 |
+-------+------------+------+

与rank()函数相同的是，相同点赞数会返回相同的排名，但是dense_rank()返回的最终排名是连续的排名

row_number()

row_number()函数返回点赞的榜单，row_number() over()

 select name, praise_num, row_number() over (order by praise_num desc) as `rank` from praise_record;
 +-------+------------+------+
| name  | praise_num | rank |
+-------+------------+------+
| Bob   |         10 |    1 |
| David |          7 |    2 |
| oct   |          7 |    3 |
| John  |          5 |    4 |
| Jane  |          3 |    5 |
| Alice |          3 |    6 |
+-------+------------+------+

row_number()函数适合当返回的列表只需要序号时使用

以上三个函数都是MySQL8.0新加入的，所以在MySQL5.7这些老版本上我们可以模拟实现一下，顺便学习一下这三个窗口函数的实现原理

rank()函数的模拟实现

select p1.name, p1.praise_num, count(p2.praise_num) + 1 as `rank` from praise_record p1
left join praise_record p2 on p1.praise_num < p2.praise_num
group by p1.name, p1.praise_num
order by `rank`;
+-------+------------+------+
| name  | praise_num | rank |
+-------+------------+------+
| Bob   |         10 |    1 |
| David |          7 |    2 |
| oct   |          7 |    2 |
| John  |          5 |    4 |
| Jane  |          3 |    5 |
| Alice |          3 |    5 |
+-------+------------+------+

我们可以使用自联接的方式将每个分数低于当前行分数的记录计数，最后将计数值加1作为当前行的排名，来模拟实现rank()

dense_rank()的模拟实现

select p1.name, p1.praise_num, count(distinct p2.praise_num) + 1 as `dense_rank` from praise_record p1
left join praise_record p2 on p1.praise_num < p2.praise_num
group by p1.name, p1.praise_num
order by `dense_rank`;
+-------+------------+------------+
| name  | praise_num | dense_rank |
+-------+------------+------------+
| Bob   |         10 |          1 |
| oct   |          7 |          2 |
| David |          7 |          2 |
| John  |          5 |          3 |
| Jane  |          3 |          4 |
| Alice |          3 |          4 |
+-------+------------+------------+

dense_rank的实现与rank差不多，唯一的区别是增加了distinct对点赞数做了去重，这样子对不同的点赞数返回的排名就是连续的

row_number的模拟实现

##使用自定义变量得先初始化
set @rowNum = 0;
select name, praise_num, @rowNum := @rowNum +1 as `row_number`  from praise_record order by praise_num desc ;
+-------+------------+------------+
| name  | praise_num | row_number |
+-------+------------+------------+
| Bob   |         10 |          1 |
| David |          7 |          2 |
| oct   |          7 |          3 |
| John  |          5 |          4 |
| Jane  |          3 |          5 |
| Alice |          3 |          6 |
+-------+------------+------------+