SQL之行列转置以及非常规的行列转置

一、准备工作

# 创建数据表
CREATE TABLE ChengJi
(
    Name    varchar(32),
    Subject varchar(32),
    Result  int(10)
);
# 插入数据
insert into ChengJi
values ('张三', '语文', 80),
       ('张三', '数学', 90),
       ('张三', '物理', 85),
       ('李四', '语文', 85),
       ('李四', '数学', 92),
       ('李四', '物理', 82);

二、行转列

整体分两步走

1、先预处理数据，将数据进行初步的行转列，便于后续的分组处理

select Name,
       case
            when Subject = '语文' then Result else 0
       end as 'Chinese',
       case
           when Subject = '数学' then Result else 0
       end as 'Math',
       case
           when Subject = '物理' then Result else 0
       end as 'Pha'
from ChengJi;

2、对预处理完毕的数据进行分组聚合，使多行数据汇聚到一个组内，达到数据集中的结果，这其中要注意的一点是：要明确按照哪个字段进行聚合操作。

with t1 as(select Name,
       case
            when Subject = '语文' then Result else 0
       end as 'Chinese',
       case
           when Subject = '数学' then Result else 0
       end as 'Math',
       case
           when Subject = '物理' then Result else 0
       end as 'Pha'
from ChengJi)
select Name,
       sum(Chinese) as 'Chinese',
       sum(Math) as 'Math',
       sum(Pha) as 'Pha'
from t1
group by Name;

三、列转行

为便于理解，我们将刚才已经转置好的结果插入到一个结果表内

1、创建一个结果表

create table ChengJi_2(
    Name varchar(255),
    Chinese int,
    Math int,
    Pha int
);

2、将行转列结果插入到结果表

insert into chengji_2
with t1 as(select Name,
       case
            when Subject = '语文' then Result else 0
       end as 'Chinese',
       case
           when Subject = '数学' then Result else 0
       end as 'Math',
       case
           when Subject = '物理' then Result else 0
       end as 'Pha'
from ChengJi)
select Name,
       sum(Chinese) as 'Chinese',
       sum(Math) as 'Math',
       sum(Pha) as 'Pha'
from t1
group by Name
;

3、对结果表进行列转行的操作，列转行相对于行转列较为简单，可直接使用 union all 进行操作。

select Name,Chinese from ChengJi_2
union all
select Name,Math from ChengJi_2
union all
select Name,Pha from ChengJi_2;

四、特殊的列转行/行转列

但是对于一些特殊的行列转置，以上方法就不再使用，通常情况下，我们的行列转置是有可以进行分组聚合操作可以完成的，而生产实践中也多数如此，但是有时有一些特殊的操作是以上方法无法完成的，这就需要一些特殊的行列转置来完成，对此，我给出了以下的方案。

1、准备工作，创建数据表并插入数据

CREATE TABLE 2003a
(
    seat   varchar(255) ,
    status varchar(255) ,
    rowid  varchar(255)
)
;
INSERT INTO 2003a
VALUES ('2', '已预订', 'A');
INSERT INTO 2003a
VALUES ('3', '未预订', 'A');
INSERT INTO 2003a
VALUES ('4', '未预订', 'A');
INSERT INTO 2003a
VALUES ('5', '未预订', 'A');
INSERT INTO 2003a
VALUES ('6', '未预订', 'B');
INSERT INTO 2003a
VALUES ('7', '未预订', 'B');
INSERT INTO 2003a
VALUES ('8', '未预订', 'B');
INSERT INTO 2003a
VALUES ('9', '未预订', 'B');
INSERT INTO 2003a
VALUES ('10', '未预订', 'B');
INSERT INTO 2003a
VALUES ('11', '未预订', 'C');
INSERT INTO 2003a
VALUES ('12', '已预订', 'C');
INSERT INTO 2003a
VALUES ('13', '已预订', 'C');
INSERT INTO 2003a
VALUES ('14', '未预订', 'C');
INSERT INTO 2003a
VALUES ('15', '未预订', 'C');
INSERT INTO 2003a
VALUES ('16', '未预订', 'D');
INSERT INTO 2003a
VALUES ('17', '未预订', 'D');
INSERT INTO 2003a
VALUES ('18', '未预订', 'D');
INSERT INTO 2003a
VALUES ('19', '未预订', 'D');
INSERT INTO 2003a
VALUES ('20', '已预订', 'D');

2、明确需求

原有表的结构：

2,已预订,A
3,未预订,A

需要完成的工作：

2,3
已预定，未预定
A,A

在这里我们可以很明显的看出，我们需要做的就是如何进行 行转列/列转行 的操作，在这里的行列转置是整行/整列进行转置，不再是依靠某个字段进行分组处理或者使用 union all 进行整体操作，因此，我是用以下方案来完成。

1、对原表字段进行 group_concat，指定 “，”为字段值之间的分隔符

           SELECT
                GROUP_CONCAT(rowid ORDER BY rowid ASC SEPARATOR ', ') AS rowid,
                GROUP_CONCAT(status ORDER BY rowid ASC SEPARATOR ', ') AS status,
                GROUP_CONCAT(seat ORDER BY rowid ASC SEPARATOR ', ') AS seat
            FROM (
                     SELECT rowid, status, seat from `2003a`
                 ) AS subquery

2、将所有的字段按照值聚合到一个数据表格内之后，我们可以使用 union all 来进行字段拆分

with t1 as (SELECT
                GROUP_CONCAT(rowid ORDER BY rowid ASC SEPARATOR ', ') AS rowid,
                GROUP_CONCAT(status ORDER BY rowid ASC SEPARATOR ', ') AS status,
                GROUP_CONCAT(seat ORDER BY rowid ASC SEPARATOR ', ') AS seat
            FROM (
                     SELECT rowid, status, seat from `2003a`
                 ) AS subquery)
select seat from t1
union all
select status from t1
union all
select rowid from t1

3、使用 SUBSTRING_INDEX来进行拆分，将所有的字段值拆分成单独的值

with t1 as (SELECT
                GROUP_CONCAT(rowid ORDER BY rowid ASC SEPARATOR ', ') AS rowid,
                GROUP_CONCAT(status ORDER BY rowid ASC SEPARATOR ', ') AS status,
                GROUP_CONCAT(seat ORDER BY rowid ASC SEPARATOR ', ') AS seat
            FROM (
                     SELECT rowid, status, seat from `2003a`
                 ) AS subquery)
,t2 as (select seat from t1
union all
select status from t1
union all
select rowid from t1)
select SUBSTRING_INDEX(seat,',',1) as p1,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',2),',',-1) as p2,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',3),',',-1) as p3,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',4),',',-1) as p4,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',5),',',-1) as p5,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',6),',',-1) as p6,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',7),',',-1) as p7,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',8),',',-1) as p8,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',9),',',-1) as p9,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',10),',',-1) as p10,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',11),',',-1) as p11,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',12),',',-1) as p12,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',13),',',-1) as p13,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',14),',',-1) as p14,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',15),',',-1) as p15,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',16),',',-1) as p16,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',17),',',-1) as p17,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',18),',',-1) as p18,
       SUBSTRING_INDEX(SUBSTRING_INDEX(seat,',',19),',',-1) as p19
from t2

在这里需要注意的是：第一个SUBSTRING_INDEX我们取的是源数据的第一个值，第二个SUBSTRING_INDEX以及之后的，我们取得是源数据的倒数第一个值，因此这里需要注意一下我们给到的是“-1”

至此，我们使用group_concat()以及SUBSTRING_INDEX()来达到了特殊的行列转置操作。