MySQL分类排名和分组TOP N实例详解

表结构

学生表如下：

CREATE TABLE `t_student` (
 `id` int NOT NULL AUTO_INCREMENT,
 `t_id` int DEFAULT NULL COMMENT '学科id',
 `score` int DEFAULT NULL COMMENT '分数',
 PRIMARY KEY (`id`)
);

数据如下： 

题目一：获取每个科目下前五成绩排名（允许并列）

允许并列情况可能存在如4、5名成绩并列情况，会导致取前4名得出5条数据，取前5名也是5条数据。

SELECT
s1.*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score < s2.score
GROUP BY
s1.id
HAVING
COUNT( s2.id ) < 5
ORDER BY
s1.t_id,
s1.score DESC

  ps：取前4名时

 分析：

1.自身左外连接，得到所有的左边值小于右边值的集合。以t_id=1时举例，24有5个成绩大于他的（74、64、54、44、34），是第6名，34只有4个成绩大于他的，是第5名......74没有大于他的，是第一名。

SELECT
*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score < s2.score

  2. 把总结的规律转换成SQL表示出来，就是group by 每个student 的 id(s1.id)，Having统计这个id下面有多少个比他大的值(s2.id)

SELECT
s1.*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score < s2.score
GROUP BY
s1.id
HAVING
COUNT( s2.id ) < 5

 3. 最后根据 t_id 分类，score 倒序排序即可。

题目二：获取每个科目下最后两名学生的成绩平均值

取最后两名成绩

SELECT
s1.*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score
GROUP BY
s1.id
HAVING
COUNT( s1.id )< 2
ORDER BY
s1.t_id,
s1.score

并列存在情况下可能导致筛选出的同一t_id 下结果条数大于2条，但题目要求是取最后两名的平均值，多条平均后还是本身，故不必再对其处理，可以满足题目要求。 

 分组求平均值：

SELECT
t_id,AVG(score)
FROM
(
SELECT
s1.*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score
GROUP BY
s1.id
HAVING
COUNT( s1.id )< 2
ORDER BY
s1.t_id,
s1.score
) tt
GROUP BY
t_id

结果： 

分析：

1. 查询出所有t1.score>t2.score 的记录

SELECT
s1.*,s2.*
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score

2. group by s.id 去重，having 计数取2条

3. group by t_id 分别取各自学科的然后avg取均值

题目三：获取每个科目下前五成绩排名（不允许并列）

SELECT
*
FROM
(
SELECT
s1.*,
@rownum := @rownum + 1 AS num_tmp,
@incrnum :=
CASE

WHEN @rowtotal = s1.score THEN
@incrnum
WHEN @rowtotal := s1.score THEN
@rownum
END AS rownum
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score,
( SELECT @rownum := 0, @rowtotal := NULL, @incrnum := 0 ) AS it
GROUP BY
s1.id
ORDER BY
s1.t_id,
s1.score DESC
) tt
GROUP BY
t_id,
score,
rownum
HAVING
COUNT( rownum )< 5

 分析：

1.引入辅助参数

SELECT
s1.*,
@rownum := @rownum + 1 AS num_tmp,
@incrnum :=
CASE

WHEN @rowtotal = s1.score THEN
@incrnum
WHEN @rowtotal := s1.score THEN
@rownum
END AS rownum
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score,
( SELECT @rownum := 0, @rowtotal := NULL, @incrnum := 0 ) AS it

2.去除重复s1.id，分组排序

SELECT
s1.*,
@rownum := @rownum + 1 AS num_tmp,
@incrnum :=
CASE

WHEN @rowtotal = s1.score THEN
@incrnum
WHEN @rowtotal := s1.score THEN
@rownum
END AS rownum
FROM
student s1
LEFT JOIN student s2 ON s1.t_id = s2.t_id
AND s1.score > s2.score,
( SELECT @rownum := 0, @rowtotal := NULL, @incrnum := 0 ) AS it
GROUP BY
s1.id
ORDER BY
s1.t_id,
s1.score DESC

 3.GROUP BY    t_id, score, rownum   然后 HAVING 取前5条不重复的

来源：https://blog.csdn.net/Dlgdlgd/article/details/122129327