具有基于单列数据的动态标题的 MySQL 数据透视表

Question

我正在尝试编写一个查询来创建一个数据“表”，如下所示：

SELECT cs.`category_id`, cs.`ProcessDate`, cs.`PercentChange`
  FROM `Category_Statistics` cs
 WHERE cs.`ProcessDate` >= '2011-05-10'
   AND cs.`ProcessDate` <= '2011-05-14'

这会返回如下内容：

CategoryId  |  ProcessDate  | PercentChange
-------------------------------------------
category_4  |  2011-05-10   |      10
category_4  |  2011-05-11   |      18
category_4  |  2011-05012   |      12
...
category_7  |  2011-05-10   |      21
category_7  |  2011-05-11   |      7
...
category_12 |  2011-05-10   |      7
category_12 |  2011-05-11   |      15

现在我希望结果是这样的（来自 MySQL 查询，而不是由应用程序操作）：

CategoryId    | 2011-05-10 | 2011-05-11 | 2011-05-12 | 2011-05-13 | 2011-05-14 |
--------------------------------------------------------------------------------
category_4    |     10     |     18     |     12     |      9     |      14    |
category_7    |     21     |      7     |     16     |      14    |      13    |
categeory_12  |      7     |     15     |     11     |      19    |       8    |
--------------------------------------------------------------------------------

对此有两个警告：

1. 日期范围可以扩大或缩小 （取决于查询）

2. PercentChange 在某些情况下可能为空 案例（比如说 category_7 / 2011-05-12 可能没有设置值）

所以最终我不太确定如何构建查询的选择部分以反映动态列数（我知道它与 CONCAT 有关）。

编辑 --> 部分工作代码 -->

SELECT `CategoryId`,
   MAX(IF(c.`ProcessedOn` = '2011-04-20', c.`PercentChange`, NULL)) AS '2011-04-20',
   MAX(IF(c.`ProcessedOn` = '2011-04-21', c.`PercentChange`, NULL)) AS '2011-04-21',
   MAX(IF(c.`ProcessedOn` = '2011-04-22', c.`PercentChange`, NULL)) AS '2011-04-22',
   MAX(IF(c.`ProcessedOn` = '2011-04-23', c.`PercentChange`, NULL)) AS '2011-04-23',
   MAX(IF(c.`ProcessedOn` = '2011-04-24', c.`PercentChange`, NULL)) AS '2011-04-24'
  FROM `Category_Gravity` c
 WHERE c.`ProcessedOn` >= '2011-04-20'
   AND c.`ProcessedOn` <= '2011-04-24'
 GROUP BY `CategoryId`

我现在需要做的就是转动

MAX(IF(c.`ProcessedOn` = '2011-04-20', c.`PercentChange`, NULL)) AS '2011-04-20',

变成更动态的东西（因为我的日期范围会改变）

Answer 1

看看这个类似的线程，我写了一个 sp 来完成任务

Join two tables (with a 1-M relationship) where the second table needs to be 'flattened' into one row

编辑。更新答案

create table `pivot` (
  `id` int(11) not null auto_increment,
  `categoryid` int(11) default null,
  `processdate` date default null,
  `percentchange` int(11) default null,
  primary key (`id`)
) engine=myisam auto_increment=9 default charset=latin1;

/*Data for the table `pivot` */

insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (1,4,'2011-05-10',1);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (2,4,'2011-05-11',22);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (3,4,'2011-05-12',3);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (4,7,'2011-05-10',4);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (5,7,'2011-05-11',5);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (6,12,'2011-05-10',6);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (7,12,'2011-05-12',7);
insert  into `pivot`(`id`,`categoryid`,`processdate`,`percentchange`) values (8,4,'2011-05-13',12);



delimiter //
drop procedure if exists dynamic_view2//
create procedure dynamic_view2(in sdate date,in edate date)
begin
declare finish int default 0;
declare cdate date;
declare str varchar(10000) default "select categoryid,";
declare curs cursor for select processdate from pivot where processdate between sdate and edate group by processdate;
declare continue handler for not found set finish = 1;
open curs;
my_loop:loop
fetch curs into cdate;
if finish = 1 then
leave my_loop;
end if;
set str = concat(str, "max(case when processdate = '",cdate,"' then percentchange else null end) as `",cdate,"`,");
end loop;
close curs;
set str = substr(str,1,char_length(str)-1);
set @str = concat(str," from pivot
            group by categoryid");

prepare stmt from @str;
execute stmt;
deallocate prepare stmt;
end;//
delimiter ;


mysql> call dynamic_view2('2011-05-10','2011-05-13');
+------------+------------+------------+------------+------------+
| categoryid | 2011-05-10 | 2011-05-11 | 2011-05-12 | 2011-05-13 |
+------------+------------+------------+------------+------------+
|          4 |          1 |         22 |          3 |         12 |
|          7 |          4 |          5 |       NULL |       NULL |
|         12 |          6 |       NULL |          7 |       NULL |
+------------+------------+------------+------------+------------+
3 rows in set (0.00 sec)