mysql获取分组后每组的最大值实例详解

1. 测试数据库表如下：

create table test 
( 
  `id` int not null auto_increment, 
  `name` varchar(20) not null default '', 
  `score` int not null default 0, 
  primary key(`id`) 
)engine=InnoDB CHARSET=UTF8; 

2. 插入如下数据：

mysql> select * from test; 
+----+----------+-------+ 
| id | name   | score | 
+----+----------+-------+ 
| 1 | jason  |   1 | 
| 2 | jason  |   2 | 
| 3 | jason  |   3 | 
| 4 | linjie  |   1 | 
| 5 | linjie  |   2 | 
| 6 | linjie  |   3 | 
| 7 | xiaodeng |   1 | 
| 8 | xiaodeng |   2 | 
| 9 | xiaodeng |   3 | 
| 10 | hust   |   2 | 
| 11 | hust   |   3 | 
| 12 | hust   |   1 | 
| 13 | haha   |   1 | 
| 14 | haha   |   2 | 
| 15 | dengzi  |   3 | 
| 16 | dengzi  |   4 | 
| 17 | dengzi  |   5 | 
| 18 | shazi  |   3 | 
| 19 | shazi  |   4 | 
| 20 | shazi  |   2 | 
+----+----------+-------+ 

3. 下面是重点，目的是要按照name分组，然后分组后，获取每组中score分数最多的，sql如下

select a.* from test a inner join (select name,max(score) score from test group by name)b on a.
name=b.name and a.score=b.score order by a.name; 

当然，上面的最后的order by a.name可以去掉

4. 测试结果如下:

+----+----------+-------+ 
| id | name   | score | 
+----+----------+-------+ 
| 3 | jason  |   3 | 
| 6 | linjie  |   3 | 
| 9 | xiaodeng |   3 | 
| 11 | hust   |   3 | 
| 14 | haha   |   2 | 
| 17 | dengzi  |   5 | 
| 19 | shazi  |   4 | 
+----+----------+-------+ 

5. 网上很多方法都是错误的，比如如下一些，亲测是不行的

select * from (select * from test order by score desc) t group by name order by score desc limit 4; 
select score,max(score) from test group by name; 
select * from test where score in (select max(score) from test group by name); 
select * from test where score in (select substring_index(group_concat(score order by score desc separator ','),',',1) from test group by name); 
 
select * from (select name,score,ROW_NUMBER() over(group by name order by score desc) as rowNum from test) rank where rank.rowNum <=1 order by rank.score desc; 
 
select * from( select StoresNo,[CustomerCaseNo],[PaymentsTime], ROW_NUMBER() over(partition by CustomerCaseNo order by [PaymentsTime] desc) as rowNum 
from BAL_paymentsSwiftInfo where StoresNo='zq00000034') ranked where ranked.rowNum <= 1 order by ranked.CustomerCaseNo, ranked.PaymentsTime desc 
 
select * from (select * from test order by score desc) as a group by a.name;