【问题标题】:Pivoting 2 columns from 3 Tables and creating pivot-column-names to avoid conflict - SQL-Server 2008R2从 3 个表中旋转 2 列并创建数据透视列名称以避免冲突 - SQL-Server 2008R2
【发布时间】:2013-04-01 21:21:35
【问题描述】:

介绍和问题

在我的示例中,我有教师、学生和课程。我想大致了解哪些课程由谁在哪些房间教授,以及该课程中的所有学生。我有基本的设置运行(带有一些手工编码的语句)。但是直到现在我还没有准备好正确的 STUFF 语句:

  • 准备@colsStudents,这样我就可以将名称放在列标题中,并且无需弄乱 id(添加 100)以避免rooms.id 和students.id 之间的冲突
  • 准备@colsRooms,这样我就不必硬编码房间名了
  • 使用 EXEC sp_executesql @sql; 将 i 放在一起

您可以找到所有用于创建此架构的 sql 语句以及最后的数据。

想要的结果概览课程,

我想旋转列 RoomName 和 StudentName 并将列值用作新列名。所有用于创建表和数据的 SQL 语句都在最后。

Id | Course | Teacher | A3 | E7 | Penny | Cooper | Koothrap. | Amy
---+--------+---------+----+----+-------+--------+-----------+-----+
1  | C# 1   | Marc G. |    | 1  |  1    |        |           |
2  | C# 2   | Sam S.  |    | 1  |  1    |        |      1    |
3  | C# 3   | John S. | 1  |    |       |    1   |           |
4  | C# 3   | Reed C. |    | 1  |       |        |      1    |
5  | SQL 1  | Marc G. | 1  |    |       |        |           |  
6  | SQL 2  | Marc G. | 1  |    |       |        |           |  
7  | SQL 3  | Marc G. |    | 1  |       |    1   |           |  1
8  | SQL 3  | Gbn     | 1  |    |       |        |      1    |   

到目前为止我有什么

With PivotData as (
Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
        ,r.Id as RoomId, r.RoomName as RoomName
        ,100 + s.Id as StudentId, s.StudentName as Student 
    FROM CourseDetails cd 
        Left JOIN Courses c ON cd.CourseId = c.Id
        Left JOIN Teachers t ON cd.TeacherId = t.Id
        Left JOIN CourseMember cm ON cd.Id = cm.CourseDetailsId
        Left JOIN Students s ON cm.StudentId = s.Id 
        Left JOIN Rooms r ON cd.RoomId = r.Id
        )    
Select Course, Teacher
    , [1] as A3, [2] as E7 -- RoomColumns
    , [101] as Koothrappali, [102] as Cooper, [103] as Penny, [104] as Amy -- StudentColumns
    FROM (
        Select Course, Teacher, RoomName, RoomId,Student, StudentId
        From PivotData) src
    PIVOT( Max(RoomName) FOR RoomId IN ([1],[2])) as P1
    PIVOT( Count(Student) FOR StudentId IN ([101],[102],[103],[104]) ) as P2

缺少什么

上面的语句是手工准备的。由于我事先不知道房间或学生,我需要为列房间和学生动态创建枢轴语句。在 SO 上有很多如何做到这一点的例子。正常的做法是使用 STUFF:

DECLARE @colsStudents AS NVARCHAR(MAX);
SET @colsStudents = STUFF(
        (SELECT N',' + QUOTENAME(y) AS [text()] FROM 
            (SELECT DISTINCT 100 + Id AS y FROM dbo.Students) AS Y 
                ORDER BY y
                FOR XML PATH('')
        ),1
        ,1
        ,N'');
Select @colsStudents                    

这将为学生 ID 返回 [101],[102],[103],[104]。我为每个 id 添加了 100 以避免students.id 和 teh rooms.id 列之间的冲突。

正如介绍中提到的,我需要动态创建类似的东西

[1] as RoomName_1, [2] as RoomName_1 -- RoomColumns
[1] as StudentName1, [2] as StudentName2, ... ,[4] as Amy -- StudentColumns

但是我对 stuff 语句的所有尝试都失败了。

创建表和数据的所有 SQL 语句

CREATE TABLE [dbo].[Teachers](
    [Id] [int] IDENTITY(1,1) NOT NULL,
    [TeacherName] [nvarchar](120) NULL,
    CONSTRAINT PK_Teachers PRIMARY KEY CLUSTERED (Id))

CREATE TABLE [dbo].[Students](
    [Id] [int] IDENTITY(1,1) NOT NULL,
    [StudentName] [nvarchar](120) NULL,
    CONSTRAINT PK_Students PRIMARY KEY CLUSTERED (Id))

CREATE TABLE [dbo].[Courses](
    [Id] [int] IDENTITY(1,1) NOT NULL,
    [CourseName] [nvarchar](120) NULL,
    CONSTRAINT PK_Courses PRIMARY KEY CLUSTERED (Id))

CREATE TABLE [dbo].[Rooms](
    [Id] [int] IDENTITY(1,1) NOT NULL,
    [RoomName] [nchar](120) NULL,
    CONSTRAINT PK_Rooms PRIMARY KEY CLUSTERED (Id))

CREATE TABLE [dbo].[CourseDetails](
  [Id] [int] IDENTITY(1,1) NOT NULL,
  [CourseId] [int] NOT NULL,
  [TeacherId] [int] NOT NULL,
  [RoomId] [int] NOT NULL,  
  CONSTRAINT PK_CourseDetails PRIMARY KEY CLUSTERED (Id),
  CONSTRAINT FK_CourseDetails_Teachers_Id FOREIGN Key (TeacherId)
    REFERENCES dbo.Teachers (Id),   
  CONSTRAINT FK_CourseDetails_Courses_Id FOREIGN Key (CourseId)
    REFERENCES dbo.Courses (Id),    
  CONSTRAINT FK_CourseDetails_Rooms_Id FOREIGN Key (RoomId)
    REFERENCES dbo.Rooms (Id)       
)       


CREATE TABLE [dbo].[CourseMember](
  [Id] [int] IDENTITY(1,1) NOT NULL,
  [CourseDetailsId] [int] NOT NULL,
  [StudentId] [int] NOT NULL,
  CONSTRAINT PK_CourseMember PRIMARY KEY CLUSTERED (Id),
  CONSTRAINT FK_CourseMember_CourseDetails_Id FOREIGN Key (CourseDetailsId)
    REFERENCES dbo.CourseDetails (Id),   
  CONSTRAINT FK_CourseMember_Students_Id FOREIGN Key (StudentId)
    REFERENCES dbo.Students (Id)     
)



INSERT INTO dbo.Courses (CourseName)
VALUES ('SQL 1 - Basics'),
    ('SQL 2 - Intermediate'),
    ('SQL 3 - Advanced'),   
    ('C# 1 - Basics'),
    ('C# 2 - Intermediate'),    
    ('C# 3 - Advanced')     

INSERT INTO dbo.Students (StudentName)
VALUES
   ('Koothrappali'),   
   ('Cooper'),
   ('Penny'),   
   ('Amy') 

INSERT INTO dbo.Teachers (TeacherName)
VALUES
   ('gbn '),
   ('Sam S.'),
   ('Marc G.'),   
   ('Reed C.'),
   ('John S.')

INSERT INTO dbo.Rooms (RoomName)
VALUES ('A3'), ('E7')


INSERT [dbo].[CourseDetails] (CourseId, TeacherId, RoomId) 
    VALUES (4, 3, 2),(5, 2, 2),
        (6, 5, 1),(6, 4, 2),
        (1,3,1),(2,3,1),(3,3,2),
        (3,1,1)

INSERT [dbo].[CourseMember] (CourseDetailsId, StudentId) 
    VALUES (1,3),(2,3),(2,1),(3,2),(4,1),(7,2),(7,4),(8,1)

【问题讨论】:

  • 您需要将第二个聚合作为计数吗?或者你只是想要一个学生在课堂上的标志?一个学生实际上可以在课堂上不止一次吗?
  • 目前不需要计数;一面旗帜就足够了。在我的脑海中,我没有任何限制可以阻止学生在一门课程中不止一次。但事实并非如此。
  • 我找到了我遇到问题的一个原因。 StudentName、TeacherName 和 RoomName 的类型为 nchar(120),因此每个字段中添加了 120 个符号。我把它改成了 nvarchar(120)。

标签: sql sql-server-2008 pivot


【解决方案1】:

我个人会做一些不同的事情。由于您试图旋转两个单独的列,它们尖叫着使用UNPIVOT 函数。

unpivot 会将您的多列转换为行,然后再进行旋转。

由于您拥有 SQL Server 2008,因此您可以使用 CROSS APPLY 和值:

  select id, course, teacher, col, flag
  from
  (
    Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
      ,cast(r.Id as varchar(10))as RoomId
      , r.RoomName as RoomName
      ,cast(100 + s.Id as varchar(10)) as StudentId
      , s.StudentName as Student
      , '1' flag
    FROM CourseDetails cd 
    Left JOIN Courses c 
      ON cd.CourseId = c.Id
    Left JOIN Teachers t 
      ON cd.TeacherId = t.Id
    Left JOIN CourseMember cm 
      ON cd.Id = cm.CourseDetailsId
    Left JOIN Students s 
      ON cm.StudentId = s.Id 
    Left JOIN Rooms r 
      ON cd.RoomId = r.Id
  ) d
  cross apply
  (
    values ('roomname', roomname),('student',student)
  ) c (value, col)

见Demo。 unpivot 生成类似于以下的结果:

| ID |               COURSE | TEACHER |          COL | FLAG |
-------------------------------------------------------------
|  1 |        C# 1 - Basics | Marc G. |           E7 |    1 |
|  1 |        C# 1 - Basics | Marc G. |        Penny |    1 |
|  2 |  C# 2 - Intermediate |  Sam S. |           E7 |    1 |
|  2 |  C# 2 - Intermediate |  Sam S. |        Penny |    1 |
|  2 |  C# 2 - Intermediate |  Sam S. |           E7 |    1 |
|  2 |  C# 2 - Intermediate |  Sam S. | Koothrappali |    1 |
|  3 |      C# 3 - Advanced | John S. |           A3 |    1 |
|  3 |      C# 3 - Advanced | John S. |       Cooper |    1 |

您将看到col 数据包含您要转换的所有值。一旦数据在行中,如果将很容易应用一个枢轴:

select id, course, teacher, 
  coalesce(A3, '') A3, 
  coalesce(E7, '') E7, 
  coalesce(Koothrappali, '') Koothrappali, 
  coalesce(Cooper, '') Cooper, 
  coalesce(Penny, '') Penny, 
  coalesce(Amy, '') Amy
from
(
  select id, course, teacher, col, flag
  from
  (
    Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
      ,cast(r.Id as varchar(10))as RoomId
      , r.RoomName as RoomName
      ,cast(100 + s.Id as varchar(10)) as StudentId
      , s.StudentName as Student
      , '1' flag
    FROM CourseDetails cd 
    Left JOIN Courses c 
      ON cd.CourseId = c.Id
    Left JOIN Teachers t 
      ON cd.TeacherId = t.Id
    Left JOIN CourseMember cm 
      ON cd.Id = cm.CourseDetailsId
    Left JOIN Students s 
      ON cm.StudentId = s.Id 
    Left JOIN Rooms r 
      ON cd.RoomId = r.Id
  ) d
  cross apply
  (
    values ('roomname', roomname),('student',student)
  ) c (value, col)
) d
pivot
(
  max(flag)
  for col in (A3, E7, Koothrappali, Cooper, Penny, Amy)
) piv

见SQL Fiddle with Demo。

然后要将其转换为动态 SQL,您只需旋转一列,因此您将使用以下内容获取列列表:

select @cols = STUFF((SELECT  ',' + QUOTENAME(col) 
                    from 
                    (
                      select id, roomname col, 1 SortOrder
                      from rooms
                      union all
                      select id, StudentName, 2
                      from Students
                    ) d
                    group by id, col, sortorder
                    order by sortorder, id
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

这将获得不同房间和学生的列表,然后在枢轴中使用。所以最终的代码是:

DECLARE @cols AS NVARCHAR(MAX),
    @colsNull AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT  ',' + QUOTENAME(col) 
                    from 
                    (
                      select id, roomname col, 1 SortOrder
                      from rooms
                      union all
                      select id, StudentName, 2
                      from Students
                    ) d
                    group by id, col, sortorder
                    order by sortorder, id
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

select @colsNull = STUFF((SELECT  ', coalesce(' + QUOTENAME(col)+', '''') as '+QUOTENAME(col)
                    from 
                    (
                      select id, roomname col, 1 SortOrder
                      from rooms
                      union all
                      select id, StudentName, 2
                      from Students
                    ) d
                    group by id, col, sortorder
                    order by sortorder, id
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query 
  = 'SELECT 
      id, course, teacher,' + @colsNull + ' 
     from
    (
      select id, course, teacher, col, flag
      from
      (
        Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
          ,cast(r.Id as varchar(10))as RoomId
          , r.RoomName as RoomName
          ,cast(100 + s.Id as varchar(10)) as StudentId
          , s.StudentName as Student
          , ''1'' flag
        FROM CourseDetails cd 
        Left JOIN Courses c 
          ON cd.CourseId = c.Id
        Left JOIN Teachers t 
          ON cd.TeacherId = t.Id
        Left JOIN CourseMember cm 
          ON cd.Id = cm.CourseDetailsId
        Left JOIN Students s 
          ON cm.StudentId = s.Id 
        Left JOIN Rooms r 
          ON cd.RoomId = r.Id
      ) d
      cross apply
      (
        values (''roomname'', roomname),(''student'',student)
      ) c (value, col)
    ) d
    pivot
    (
      max(flag)
      for col in (' + @cols + ')
    ) p '

execute(@query)

见SQL Fiddle with Demo。

注意,我实现了一个要在枢轴中使用的标志,如果房间或学生有值,这基本上会生成一个 Y/N。

这给出了最终结果:

| ID |               COURSE | TEACHER | A3 | E7 | KOOTHRAPPALI | COOPER | PENNY | AMY |
---------------------------------------------------------------------------------------
|  1 |        C# 1 - Basics | Marc G. |    |  1 |              |        |     1 |     |
|  2 |  C# 2 - Intermediate |  Sam S. |    |  1 |            1 |        |     1 |     |
|  3 |      C# 3 - Advanced | John S. |  1 |    |              |      1 |       |     |
|  4 |      C# 3 - Advanced | Reed C. |    |  1 |            1 |        |       |     |
|  5 |       SQL 1 - Basics | Marc G. |  1 |    |              |        |       |     |
|  6 | SQL 2 - Intermediate | Marc G. |  1 |    |              |        |       |     |
|  7 |     SQL 3 - Advanced | Marc G. |    |  1 |              |      1 |       |   1 |
|  8 |     SQL 3 - Advanced |    gbn  |  1 |    |            1 |        |       |     |

附带说明,也可以使用 sql server 中的 unpivot 函数取消透视此数据。 (见Demo with unpivot)

【讨论】:

  • AFAICT 这看起来很有希望;我仍在努力弄清楚这一切。您是否知道取消透视然后交叉应用和仅使用一列进行透视是否会带来性能损失/好处?
  • @threeFourOneSixOneThree CROSS APPLY 和 UNPIVOT 正在做同样的事情,请查看带有 unpivot 的演示 -- sqlfiddle.com/#!3/d95a2/89
  • 我做了一个快速的初步性能差比较:Declare at_sttime datetime 和 Set at_sttime=getdate() 在此处添加您的 sql SELECT RTRIM(CAST(DATEDIFF(MS, at_sttime, GETDATE()) AS CHAR(10))) AS 'TimeTaken' 在第一次冷运行中,我的解决方案花费了 10 毫秒,您的 43 毫秒,第二次运行您的 50 毫秒,我的 0ms(可能仍在缓存中)。由于我有其他进程正在运行(视频播放)并且缓存没有清空,这只是一个弱指标。我将在更大的样本数据(40.000 行)上比较这两种解决方案。但这需要时间来准备。
  • @threeFourOneSixOneThree 嗯,这很不寻常。我从未遇到过使用 pivot/unpivot 的任何性能问题
【解决方案2】:

您可以使用动态 sql 查询为两个数据透视列创建别名字符串, 例如,对于学生列:

DECLARE @colsStudents AS NVARCHAR(MAX),
@colsstudentalias AS NVARCHAR(MAX),
@colsRooms AS NVARCHAR(MAX),
@colsRoomsalias AS NVARCHAR(MAX)

SELECT @colsStudents = STUFF
(
  (
    SELECT DISTINCT ',' + QUOTENAME(100 + Id)
    FROM dbo.Students
    FOR XML PATH('')
  ), 1, 1, ''
)


SELECT @colsstudentalias = STUFF
(
  (
    SELECT DISTINCT ',' + QUOTENAME(100 + Id) 
                + ' as ' + QUOTENAME(ltrim(rtrim(StudentName)))
    FROM dbo.Students
    FOR XML PATH('')
  ), 1, 1, ''
)

SELECT @colsRooms = STUFF
(
  (
    SELECT DISTINCT ',' + QUOTENAME(Id)
    FROM dbo.Rooms
    FOR XML PATH('')
  ), 1, 1, ''
)


SELECT @colsRoomsalias = STUFF
(
  (
    SELECT DISTINCT ',' + QUOTENAME(Id) 
                + ' as ' + QUOTENAME(ltrim(rtrim(RoomName)))
    FROM dbo.Rooms
    FOR XML PATH('')
  ), 1, 1, ''
)

--SELECT @colsStudents, @colsstudentalias, @colsRooms, @colsRoomsalias

DECLARE @sql varchar(max)
set @sql = ';With PivotData as (
Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
        ,r.Id as RoomId, r.RoomName as RoomName
        ,100 + s.Id as StudentId, s.StudentName as Student 
    FROM CourseDetails cd 
        Left JOIN Courses c ON cd.CourseId = c.Id
        Left JOIN Teachers t ON cd.TeacherId = t.Id
        Left JOIN CourseMember cm ON cd.Id = cm.CourseDetailsId
        Left JOIN Students s ON cm.StudentId = s.Id 
        Left JOIN Rooms r ON cd.RoomId = r.Id
        )    
Select Course, Teacher
    , ' + @colsRoomsalias + '
    , ' + @colsstudentalias + '
    FROM (
        Select Course, Teacher, RoomName, RoomId,Student, StudentId
        From PivotData) src
    PIVOT( Max(RoomName) FOR RoomId IN (' + @colsRooms + ')) as P1
    PIVOT( Count(Student) FOR StudentId IN (' + @colsStudents + ') ) as P2'

exec (@sql)

SQL DEMO

【讨论】:

    【解决方案3】:

    我将深入研究上面的两个答案,并将它们与下面的答案进行比较。

    • 我的问题是用Stuff() 函数填充局部变量@RoomNames 和@StudentNames。一个原因是我选择了数据类型 nchar(120) 而不是 nvarchar(120) 用于列 StudentName、RoomName。
    • 我遇到的另一个问题是无法识别新的 columnNames(Student 而不是 StudentName);因此我在此声明中将它们替换为*:Select * From (' + @PivotSrc + N') src

    Philip Kelley 建议使用SELECT @RoomIds = isnull(@RoomIds + ',', '') + '[' + Cast(Id as nvarchar(20))+ ']' FROM Rooms 而不是STUFF(),因为我发现它更短且更易于阅读,所以我现在正在使用它。

    工作解决方案

    DECLARE @StudentNames NVARCHAR(2000),    
        @RoomIds NVARCHAR(2000),
        @RoomNames NVARCHAR(2000),
        @PivotSrc NVARCHAR(MAX),
        @PivotBase NVARCHAR(MAX);
    SELECT @StudentNames = isnull(@StudentNames + ',', '') + '[' + StudentName + ']' FROM Students
    SELECT @RoomIds = isnull(@RoomIds + ',', '') + '[' + Cast(Id as nvarchar(20))+ ']' FROM Rooms
    SELECT @RoomNames = isnull(@RoomNames + ',', '') + '[' + RoomName + ']' FROM Rooms
    
    SET @PivotSrc = N'Select cd.Id, c.CourseName as Course, t.TeacherName as Teacher
            ,r.Id as RoomId, r.RoomName as RoomName
            ,100 + s.Id as StudentId, s.StudentName as Student 
        FROM CourseDetails cd 
            Left JOIN Courses c ON cd.CourseId = c.Id
            Left JOIN Teachers t ON cd.TeacherId = t.Id
            Left JOIN CourseMember cm ON cd.Id = cm.CourseDetailsId
            Left JOIN Students s ON cm.StudentId = s.Id 
            Left JOIN Rooms r ON cd.RoomId = r.Id'
    
    SET @PivotBase = N' Select Course, Teacher, ' 
            + @RoomNames + N', ' 
        + @StudentNames + N' FROM (
           Select * From (' + @PivotSrc + N') src
           PIVOT( Max(RoomName) FOR RoomName IN ('+@RoomNames+ N')) as P1
               PIVOT( Count(Student) FOR Student IN ('+@StudentNames+N') ) as P2) as T'
    
    execute(@PivotBase)
    

    【讨论】:

    • 您使用nchar 而不是varchar 或nvarchar 的原因是什么?
    • 复制粘贴错误,我很久没注意到了。
    猜你喜欢
    • 1970-01-01
    • 2019-08-02
    • 2011-07-02
    • 1970-01-01
    • 1970-01-01
    • 2011-11-25
    • 2014-03-12
    • 2015-03-21
    • 1970-01-01
    相关资源
    最近更新 更多