Sql server SQL Server将前两行连接为两列

我有两张桌子：

表

用户

：

Id, Name, Email,...

表

图像

：

UserId, Priority, Path,...

我想创建一个用户视图，每个用户都有两个具有最高优先级的图像，结果如下：

User.Id, User.Name, User.Email, Image1, Image2
-------------------------------------------------

我假设，如果用户没有图像，则

Image1

和

Image2

两列都将

NULL

，如果只有一个图像，则将设置

Image1

，并且

Image2

将

NULL

---

我如何才能最有效地实现这一点？我指的是数据库性能。

将这些示例模式/数据用作模型：

CREATE TABLE #User (Id INT, Name VARCHAR(MAX), Email VARCHAR(MAX))
CREATE TABLE #Images (UserId INT, [Priority] INT, [Path] VARCHAR(MAX))
INSERT INTO #User VALUES
(1, 'Bob', 'bob@foo.com'),
(2, 'Jim', 'jim@foo.com')

INSERT INTO #Images VALUES
(1, 10, 'path1'),
(1, 5, 'path2'),
(1, 7, 'path3')

您可以使用

PIVOT

：

SELECT UserId, [1] AS PathOfImage1, [2] AS PathOfImage2
FROM 
(SELECT UserId, [Path], 
        ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY [Priority] DESC) AS rn
 FROM #Images ) AS source
 PIVOT 
 (
   MAX([Path])
   FOR rn IN ([1], [2])
 ) AS pvt

SELECT t.UserId, 
       [PathOfImage1] = MAX(CASE WHEN rn = 1 THEN [Path] ELSE NULL END),
       [PathOfImage2] = MAX(CASE WHEN rn = 2 THEN [Path] ELSE NULL END),
       [PriorityOfImage1] = MAX(CASE WHEN rn = 1 THEN [Priority] ELSE NULL END),
       [PriorityOfImage2] = MAX(CASE WHEN rn = 2 THEN [Priority] ELSE NULL END)
FROM (
  SELECT UserId, [Path], [Priority],
         ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY [Priority] DESC) AS rn
  FROM #Images ) t
GROUP BY t.UserId

要在单个记录中获取每个用户的前两个图像，请执行以下操作：

UserId  PathOfImage1    PathOfImage2
------------------------------------
1       path1           path3

请注意，

Id=2

的用户不会出现在上述结果集中，因为他没有任何相关图像

现在，您可以对上述查询生成的表表达式执行和

OUTER APPLY

：

SELECT u.Id, u.Name, Images.PathOfImage1, Images.PathOfImage2
FROM #User u
OUTER APPLY (
   SELECT UserId, [1] AS PathOfImage1, [2] AS PathOfImage2
   FROM (
      SELECT UserId, [Path], 
             ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY [Priority] DESC) AS rn
      FROM #Images ) AS source
   PIVOT 
   (
      MAX([Path])
      FOR rn IN ([1], [2])
   ) AS pvt
   WHERE UserId = u.Id
) Images

为了获得所需的输出：

Id  Name    PathOfImage1    PathOfImage2
----------------------------------------
1   Bob     path1           path3
2   Jim     NULL            NULL

编辑：

如果应从

Image

表中拉出多个列，则必须将

PIVOT

替换为旧式条件聚合，因为SQL Server不支持

PIVOT

上的多个聚合：

SELECT UserId, [1] AS PathOfImage1, [2] AS PathOfImage2
FROM 
(SELECT UserId, [Path], 
        ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY [Priority] DESC) AS rn
 FROM #Images ) AS source
 PIVOT 
 (
   MAX([Path])
   FOR rn IN ([1], [2])
 ) AS pvt

SELECT t.UserId, 
       [PathOfImage1] = MAX(CASE WHEN rn = 1 THEN [Path] ELSE NULL END),
       [PathOfImage2] = MAX(CASE WHEN rn = 2 THEN [Path] ELSE NULL END),
       [PriorityOfImage1] = MAX(CASE WHEN rn = 1 THEN [Priority] ELSE NULL END),
       [PriorityOfImage2] = MAX(CASE WHEN rn = 2 THEN [Priority] ELSE NULL END)
FROM (
  SELECT UserId, [Path], [Priority],
         ROW_NUMBER() OVER (PARTITION BY UserId ORDER BY [Priority] DESC) AS rn
  FROM #Images ) t
GROUP BY t.UserId

---

谢谢，不用锋利的工具也能用。顺便问一下，为什么表名中有一个sharp。有什么特别的吗？@qub1n

#

符号用于表示临时表。在发布查询之前，我不得不模拟一些示例数据来测试我的查询，临时表是一种快速的方法。好啊谢谢。@qub1n您也可以尝试用

左连接替换
OUTER APPLY
，看看这两种连接在性能上的比较。我正在玩这段代码，试图增加图像的高度和宽度。你能给我一个提示吗？我已经将所有字段放入MAX（ServerPath、Height、Width），但它似乎是错误的。。。。