sql server 行列转换

【1】创建测试数据

CREATE TABLE [dbo].[tbl_Student](
    [ID] [int] IDENTITY(1,1) NOT NULL,
    [学生ID] [varchar](50) COLLATE Chinese_PRC_CI_AS NULL,
    [学生姓名] [nvarchar](50) COLLATE Chinese_PRC_CI_AS NULL,
    [课程ID] [nchar](10) COLLATE Chinese_PRC_CI_AS NULL,
    [课程名称] [nvarchar](50) COLLATE Chinese_PRC_CI_AS NULL,
    [成绩] [int] NULL,
    [教师ID] [varchar](50) COLLATE Chinese_PRC_CI_AS NULL,
    [教师姓名] [nvarchar](50) COLLATE Chinese_PRC_CI_AS NULL
) ON [PRIMARY]

insert into tbl_Student values('S3','王五','K4 ','政治',53,'T4','赵老师');
insert into tbl_Student values('S1','张三 ','K1 ','数学',61,'T1','张老师');
insert into tbl_Student values('S2','李四','K3 ','英语',88,'T3','李老师');
insert into tbl_Student values('S1','张三 ','K4 ','政治',77,'T4','赵老师');
insert into tbl_Student values('S2','李四','K4 ','政治',67,'T5','周老师');
insert into tbl_Student values('S3','王五','K2 ','语文',90,'T2','王老师');
insert into tbl_Student values('S3','王五','K1 ','数学',55,'T1','张老师');
insert into tbl_Student values('S1','张三 ','K2 ','语文',81,'T2','王老师');
insert into tbl_Student values('S4','赵六','K2 ','语文',59,'T1','王老师');
insert into tbl_Student values('S1','张三 ','K3 ','英语',37,'T3','李老师');

【2】行转列

方法1：case when

select 学生姓名,课程ID,成绩,教师ID,教师姓名,
max(case 课程名称 when '政治' then 成绩 else 0 end) 政治, 
max(case 课程名称 when '语文' then 成绩 else 0 end) 语文,
max(case 课程名称 when '数学' then 成绩 else 0 end) 数学,
max(case 课程名称 when '英语' then 成绩 else 0 end) 英语
from tbl_Student group by 学生姓名,课程ID,成绩,教师ID,教师姓名

方法2：pivot

select *
from (select * from tbl_Student) 
a pivot (max(成绩) for 课程名称 in (语文,数学,英语,政治)) b

结果如图：

之所以很多地方为NULL是因为pivot和unpivot会把除 pivot()括号内的字段都作为分组项，所以如果想实现如下图效果。

则需要先做一个子查询或CTE来把相关字段给筛选出来，代码如下（或可以用case when做行转列直接group by 指定字段）

select *
from (select 学生姓名,成绩,课程名称 from tbl_Student) 
a pivot (max(成绩) for 课程名称 in (语文,数学,英语,政治)) b

结果如下：

【3】列转行：（或可以用union all做列传行）

CREATE TABLE [dbo].[tbl_列转行测试](
    [UserID] [int] NULL,
    [UserNo] [int] NULL,
    [A] [int] NULL,
    [B] [int] NULL,
    [C] [int] NULL
) ON [PRIMARY]

insert into [tbl_列转行测试] values(1           ,      1       ,        11   ,       22     ,     33)

select * from [tbl_列转行测试]

SELECT  USERID,USERNO,attribute,value
FROM (select * from tbl_列转行测试)a
  UNPIVOT
  (
    value FOR attribute IN(A, B,C)
  ) AS UPV

【4】行列转换实践

（1）常规列转行

需求：

解决：

--drop table #temp1
--drop table #temp2

create table #temp1([A-1] varchar(100),[A-2] varchar(100),[A-3] varchar(100))
insert into #temp1 values('张三','成都','123')
insert into #temp1 values('李四','北京','456')

create table #temp2([英文字段] varchar(100),[中文字段] varchar(100))
insert into #temp2 values('A-1','姓名')
insert into #temp2 values('A-2','地址')
insert into #temp2 values('A-3','电话')

;with t1 as (
    select * from #temp1 unpivot
    ( 
        value for attribute in ([A-1],[A-2],[A-3])
    ) t
)
,t2 as (
    select t1.value,t2.[中文字段]
    ,row_number() over(partition by t2.[中文字段] order by t1.value) as rn 
    from t1 
    join #temp2 t2 on t1.attribute=t2.[英文字段]
)
--select * from t2
select 姓名,地址,电话 from t2 pivot( max(value) for [中文字段] in (姓名,地址,电话)) q

（2）实用列传行

该题来自csdn论坛，答案出自sql server技术群小小大神

需求：

　　 =》

思路，

（1）把 item1,item2.....item_name1,item_name2...... 全部值存在新生成列x，所有原本的列名存在新生成列y

（2）然后通过自连接，根据 y 列所存储值（即原列名）的规律，来把 item1,item2..... 和 item_name1,item_name2.... 区分开来，重新划分成2列

CREATE TABLE #A(code VARCHAR(30),name VARCHAR(200),
item1 VARCHAR(30), item_name1 VARCHAR(100) ,
item2 VARCHAR(30), item_name2 VARCHAR(100) ,
item3 VARCHAR(30), item_name3 VARCHAR(100) 
)
INSERT INTO #A(code,name,item1,item_name1,item2,item_name2)
SELECT '1001','A1001','W01','W011001','W03','W011003' UNION ALL
SELECT '1002','D1001','K01','K011001','P09','P011009' UNION ALL
SELECT '1003','G1001','M01','M011001','N03','N011003' 

INSERT INTO #A(code,name,item3,item_name3)
VALUES('1004','xxxx','yy','yy001')

INSERT INTO #A(code,name,item1,item_name1,item2,item_name2,item3,item_name3)
VALUES('1005','11',1,2,3,4,5,6)

;with cet_t1 as (
SELECT *   FROM 
    (  -- 列转行之前，必须把相关列所在值转成同一种数据类型，否则组合生成的新列根本不知道应该是什么数据类型和长度
        SELECT 
        code ,CONVERT(VARCHAR(100),item_name1) AS item_name1,CONVERT(VARCHAR(100),item_name2) AS item_name2,CONVERT(VARCHAR(100),item_name3) AS item_name3,CONVERT(VARCHAR(100),item1) AS item1,CONVERT(VARCHAR(100),item2) AS item2,CONVERT(VARCHAR(100),item3) AS item3
        FROM #A
    ) A
    UNPIVOT
    (
        x FOR y IN (item_name1,item_name2,item_name3,item1,item2,item3)
    )p
)
SELECT a.code,a.x AS item,b.x AS item_name FROM cet_t1 a 
INNER JOIN  cet_t1 b ON b.code = a.code AND STUFF(a.y,1,4,'')=STUFF(b.y,1,9,'') AND SUBSTRING(a.y,5,1)<>'_'

drop table #A

实现结果：

该程序如果有多个item/item_name，需要操作的话，换成动态SQL

--列数不固定,但是item和item_name后面跟的数字是一套一套的

CREATE TABLE #A(code VARCHAR(30),name VARCHAR(200),
item1 VARCHAR(30), item_name1 VARCHAR(100) ,
item2 VARCHAR(30), item_name2 VARCHAR(100) ,
item3 VARCHAR(30), item_name3 VARCHAR(100) 
)
INSERT INTO #A(code,name,item1,item_name1,item2,item_name2)
SELECT '1001','A1001','W01','W011001','W03','W011003' UNION ALL
SELECT '1002','D1001','K01','K011001','P09','P011009' UNION ALL
SELECT '1003','G1001','M01','M011001','N03','N011003' 

INSERT INTO #A(code,name,item3,item_name3)
VALUES('1004','xxxx','yy','yy001')

INSERT INTO #A(code,name,item1,item_name1,item2,item_name2,item3,item_name3)
VALUES('1005','11',1,2,3,4,5,6)

DECLARE @sql VARCHAR(max)=''
DECLARE @where VARCHAR(max)=''
DECLARE @filed VARCHAR(max)=''

--动态拼接,先统一类型,然后列转行,在关联得到结果 
SELECT 
@filed=@filed+',CONVERT(VARCHAR(100),'+name+') AS '+name ,
@where=@where+','+name
FROM tempdb.sys.columns WHERE object_id=OBJECT_ID('tempdb..#A') AND name LIKE 'item_%'

print @filed
print @where

SET @sql='
WITH ct
AS
(
    SELECT *   FROM 
    (
    SELECT 
    code '+@filed+'
    FROM #A
    ) A
    UNPIVOT
    (
        x FOR y IN ('+STUFF(@where,1,1,'')+')
    )p
)
SELECT a.code,a.x AS item,b.x AS item_name FROM ct a 
INNER JOIN ct b ON b.code = a.code AND STUFF(a.y,1,4,'''')=STUFF(b.y,1,9,'''') AND SUBSTRING(a.y,5,1)<>''_'''

print(@sql)
EXEC (@sql)

DROP TABLE #A

【5】分隔符行列转换

1、行转列，以','号为分隔符

好的 FOR XML PATH就基本介绍到这里吧，更多关于FOR XML的知识请查阅帮助文档！

接下来我们来看一个FOR XML PATH的应用场景吧！那么开始吧。。。。。。

一个应用场景与FOR XML PATH应用

首先呢！我们在增加一张学生表，列分别为（stuID,sName,hobby）,stuID代表学生编号，sName代表学生姓名，hobby列存学生的爱好！那么现在表结构如下：

这时，我们的要求是查询学生表，显示所有学生的爱好的结果集，代码如下：

SELECT B.sName,LEFT(StuList,LEN(StuList)-1) as hobby FROM (
SELECT sName,
(SELECT hobby+',' FROM student 
  WHERE sName=A.sName 
  FOR XML PATH('')) AS StuList
FROM student A 
GROUP BY sName
) B

结果如下:

分析： 好的，那么我们来分析一下，首先看这句：

SELECT hobby+',' FROM student 
WHERE sName=A.sName 
FOR XML PATH('')

这句是通过FOR XML PATH 将某一姓名如张三的爱好，显示成格式为：“ 爱好1，爱好2，爱好3，”的格式！

那么接着看：

SELECT B.sName,LEFT(StuList,LEN(StuList)-1) as hobby FROM (
SELECT sName,
(SELECT hobby+',' FROM student 
  WHERE sName=A.sName 
  FOR XML PATH('')) AS StuList
FROM student A 
GROUP BY sName
) B

剩下的代码首先是将表分组，在执行FOR XML PATH 格式化，这时当还没有执行最外层的SELECT时查询出的结构为:

可以看到StuList列里面的数据都会多出一个逗号，这时随外层的语句:SELECT B.sName,LEFT(StuList,LEN(StuList)-1) as hobby 就是来去掉逗号，并赋予有意义的列明！

2.列转行

----------------------------------------------------------------

--> 测试数据[huang]

if object_id('[huang]') is not null drop table [huang]
go 
create table [huang]([a] nvarchar(4),[b] nvarchar(10))
insert [huang]
select 'X1','1,4,8' union all
select 'X2','2' union all
select 'X3','3,6' union all
select 'X4','7' union all
select 'X5','5'

--------------SQL查询生成数据--------------------------

select
    a.[a], 
    SUBSTRING([b],number,CHARINDEX(',',[b]+',',number)-number) as [b] 
from
    [huang] a,master..spt_values 
where
    number >=1 and number<=len([b])  
    and type='p' 
    and substring(','+[b],number,1)=','

----------------结果----------------------------

/*

a b

---- ----------

X1 1

X1 4

X1 8

X2 2

X3 3

X3 6

X4 7

X5 5

*/

经典参考文章：https://blog.csdn.net/ml1990s/article/details/16953999

参考文章：https://www.cnblogs.com/doubleliang/archive/2011/07/06/2098775.html

posted @ 2018-04-02 19:27 郭大侠1 阅读(625) 评论(0) 编辑收藏举报

刷新页面返回顶部