连接两个表(具有 1-M 关系)，其中第二个表需要“展平"为一行
给定以下表格:
学生
+----+-------+
| id | Name |
+----+-------+
| 1 | Chris |
| 2 | Joe |
| 3 | Jack |
+----+-------+
报名
+---------------+------------+-----------+----------+
| enrollment_id | student_id | course_id | complete |
+---------------+------------+-----------+----------+
| 1 | 1 | 55 | true |
| 2 | 1 | 66 | true |
| 3 | 1 | 77 | true |
| 4 | 2 | 55 | true |
| 5 | 2 | 66 | false |
| 6 | 3 | 55 | false |
| 7 | 3 | 66 | true |
+---------------+------------+-----------+----------+
我想要以下内容
+----+-------+-----------+-----------+-----------+
| id | Name | Course 55 | Course 66 | Course 77 |
+----+-------+-----------+-----------+-----------+
| 1 | Chris | true | true | true |
| 2 | Joe | true | false | NULL |
| 3 | Jack | false | true | NULL |
+----+-------+-----------+-----------+-----------+
注意 1: 我知道 mysql 不能有动态列(如果我错了，请纠正我！)所以我对查询开始感到满意:
Note 1: I know mysql can't have dynamic columns (correct me if I'm wrong!) so I am happy with the query starting as:
SELECT id, name, course_55, course_66, course_77 etc...
我对此很满意，因为课程数量是固定的(准确地说是 4 门).理想情况下我希望它是动态的；也就是说，不必在 SELECT 子句中手动编写每个课程.
I happy with this because there is a fixed number of courses (4 to be exact). Ideally I would want it to be dynamic; that is, not having to manually write each course in the SELECT clause.
注意 2: 这需要 mysql 纯 - 我不想诉诸 PHP.
Note 2: This needs to mysql pure - I don't want to resort to PHP.
该数据库目前有 10000+ 名学生，注册人数为 10000+ * 4(因为正好有 4 门课程，每个学生都在所有 4 个模块中).
The database currently stands at 10000+ students with 10000+ * 4 enrollments (as there is exactly 4 courses, and every student is in all 4 modules).
注意 3:Student.user_id 已编入索引，enrollment.enrollment_id、enrollment.student_id 和enrollment.course_id 也已编入索引.
Note 3: Student.user_id is indexed and so is enrollment.enrollment_id, enrollment.student_id, and enrollment.course_id.
推荐答案
select s.id,s.name,
max(case when e.course_id = 55 then complete else null end) as c55,
max(case when e.course_id = 66 then complete else null end) as c66,
max(case when e.course_id = 77 then complete else null end) as c77
from student as s
left join enrollment as e
on s.id = e.student_id
group by s.id
@克里斯.使用存储过程，您甚至可以在不知道列数的情况下创建动态数据透视表.这是链接
@Chris. Using stored procedure you could even create dynamic pivot table without knowing before the number of columns. This is the link
http://forum.html.it/forum/showthread.php?s=&threadid=1456236
我在意大利论坛上对类似问题的回答.有一个完整的例子可以帮助你理解背后的逻辑.:)
of an answer of mine on an italian forum to a similar problem. There is a complete example that could help you to understand the logic behind. :)
编辑.使用 MYSQL 动态视图更新
这是我的起始转储:
/*Table structure for table `student` */
drop table if exists `student`;
create table `student` (
`id` int(10) unsigned not null auto_increment,
`name` varchar(50) default null,
primary key (`id`)
) engine=myisam;
/*Data for the table `student` */
insert into `student`(`id`,`name`) values (1,'chris');
insert into `student`(`id`,`name`) values (2,'joe');
insert into `student`(`id`,`name`) values (3,'jack');
drop table if exists enrollment;
create table `enrollment` (
`enrollment_id` int(11) auto_increment primary key,
`student_id` int(11) default null,
`course_id` int(11) default null,
`complete` varchar(50) default null
) engine=myisam auto_increment=8 default charset=latin1;
/*Data for the table `enrollment` */
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (1,1,55,'true');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (2,1,66,'true');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (3,1,77,'true');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (4,2,55,'true');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (5,2,66,'false');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (6,3,55,'false');
insert into `enrollment`(`enrollment_id`,`student_id`,`course_id`,`complete`) values (7,3,66,'true');
这是动态视图的存储过程:
and this is the stored procedure for the dynamic view:
delimiter //
drop procedure if exists dynamic_view//
create procedure dynamic_view()
begin
declare finish int default 0;
declare cid int;
declare str varchar(10000) default "select s.id,s.name,";
declare curs cursor for select course_id from enrollment group by course_id;
declare continue handler for not found set finish = 1;
open curs;
my_loop:loop
fetch curs into cid;
if finish = 1 then
leave my_loop;
end if;
set str = concat(str, "max(case when e.course_id = ",cid," then complete else null end) as course_",cid,",");
end loop;
close curs;
set str = substr(str,1,char_length(str)-1);
set @str = concat(str," from student as s
left join enrollment as e
on s.id = e.student_id
group by s.id");
prepare stmt from @str;
execute stmt;
deallocate prepare stmt;
-- select str;
end;//
delimiter ;
现在让我们称之为
mysql> call dynamic_view();
+----+-------+-----------+-----------+-----------+
| id | name | course_55 | course_66 | course_77 |
+----+-------+-----------+-----------+-----------+
| 1 | chris | true | true | true |
| 2 | joe | true | false | NULL |
| 3 | jack | false | true | NULL |
+----+-------+-----------+-----------+-----------+
3 rows in set (0.00 sec)
Query OK, 0 rows affected (0.05 sec)
现在我们插入另外两条具有两个不同课程的记录:
Now we insert other two records with two different courses:
insert into `enrollment`(`student_id`,`course_id`,`complete`) values (1,88,'true');
insert into `enrollment`(`student_id`,`course_id`,`complete`) values (3,99,'true');
我们回忆一下这个过程.结果如下:
and we recall the procedure. This is the result:
mysql> call dynamic_view();
+----+-------+-----------+-----------+-----------+-----------+-----------+
| id | name | course_55 | course_66 | course_77 | course_88 | course_99 |
+----+-------+-----------+-----------+-----------+-----------+-----------+
| 1 | chris | true | true | true | true | NULL |
| 2 | joe | true | false | NULL | NULL | NULL |
| 3 | jack | false | true | NULL | NULL | true |
+----+-------+-----------+-----------+-----------+-----------+-----------+
3 rows in set (0.00 sec)
Query OK, 0 rows affected (0.02 sec)
仅此而已.:)
相关文章