一列保存多個ID（將多個用逗號、特殊符號 隔開的ID轉換成用逗號隔開的名稱）

http://www.cnblogs.com/kevin2013/archive/2012/10/19/2731009.html

背景：在做項目時，經常會遇到這樣的表結構在主表的中有一列保存的是用逗號隔開ID。如，當一個員工從屬多個部門時、當一個項目從屬多個城市時、當一個設備從屬多個項目時，很多人都會在員工表中加入一個deptIds VARCHAR(1000)列（本文以員工從屬多個部門爲例），用以保存部門編號列表（很明顯這不符合第一範式，但很多人這樣設計了，在這篇文章中我們暫不討論在這種應用場景下，如此設計的對與錯，有興趣的可以在回覆中聊聊），然後我們在查詢列表中需要看到這個員工從屬哪些部門。

初始化數據：

部門表、員工表數據：

IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Department]'))
DROP TABLE [dbo].Department
GO
--部門表
CREATE TABLE Department
(
    id int,
    name nvarchar(50)
)
INSERT INTO Department(id,name)
SELECT 1,'人事部'
UNION 
SELECT 2,'工程部'
UNION 
SELECT 3,'管理部'

SELECT * FROM Department


IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Employee]'))
DROP TABLE [dbo].Employee
GO
--員工表
CREATE TABLE Employee
(
    id int, 
    name nvarchar(20),
    deptIds varchar(1000)
)
INSERT INTO Employee(id,name,deptIds)
SELECT 1,'蔣大華','1,2,3'
UNION
SELECT 2,'小明','1'
UNION 
SELECT 3,'小華',''

SELECT * FROM Employee

 

希望得到的結果：

解決方法：

第一步，是得到如下的數據。即將員工表集合與相關的部門集合做交叉連接，其中使用了fun_SplitIds函數（作用是將ids分割成id列表），然後員工集合與這個得到的集合做交叉連接

SELECT E.*,ISNULL(D.name,'') AS deptName
FROM Employee AS E 
   OUTER APPLY dbo.fun_SplitIds(E.deptIds) AS DID
   LEFT JOIN Department AS D ON DID.ID=D.id;

 

第二步，已經得到了如上的數據，然後要做的就是根據ID分組，並對deptName列做聚合操作，但可惜的是SQL SERVER還沒有提供對字符串做聚合的操作。但想到，我們處理樹形結構數據時，用CTE來做關係數據，做成有樹形格式的數據，如此我們也可以將這個問題轉換成做樹形格式的問題，代碼如下：

;WITH EmployeT AS(
--員工的基本信息(使用OUTER APPLY將多個ID拆分開來,然後與部門表相關聯)
--此時已將員工表所存的IDS分別與部門相關聯,下面需要將此集合中的deptName聚合成一個記錄
SELECT E.*,ISNULL(D.name,'') AS deptName
FROM Employee AS E 
   OUTER APPLY dbo.fun_SplitIds(E.deptIds) AS DID
   LEFT JOIN Department AS D ON DID.ID=D.id
),mike AS(
   SELECT id,name,deptIds,deptName
      ,ROW_NUMBER()OVER(PARTITION BY id ORDER BY id) AS level_num
   FROM EmployeT
),mike2 AS(
   SELECT id,name,deptIds,CAST(deptName AS NVARCHAR(100)) AS deptName,level_num 
   FROM mike
   WHERE level_num=1
   UNION ALL
   SELECT m.id,m.name,m.deptIds,CAST(m2.deptName+','+m.deptName AS NVARCHAR(100)) AS deptName,m.level_num 
   FROM mike AS m
      INNER JOIN mike2 AS m2 ON m.ID=m2.id AND m.level_num=m2.level_num+1
),maxMikeByIDT AS(
       SELECT id,MAX(level_num) AS level_num
       FROM mike2 
       GROUP BY ID
)


SELECT A.id,A.name,A.deptIds,A.deptName
FROM mike2 AS A 
   INNER JOIN maxMikeByIDT AS B ON A.id=B.ID AND A.level_num=B.level_num
ORDER BY A.id OPTION (MAXRECURSION 0)

結果如下：

全部SQL：

IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Department]'))
DROP TABLE [dbo].Department
GO
--部門表
CREATE TABLE Department
(
    id int,
    name nvarchar(50)
)
INSERT INTO Department(id,name)
SELECT 1,'人事部'
UNION 
SELECT 2,'工程部'
UNION 
SELECT 3,'管理部'

SELECT * FROM Department


IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Employee]'))
DROP TABLE [dbo].Employee
GO
--員工表
CREATE TABLE Employee
(
    id int, 
    name nvarchar(20),
    deptIds varchar(1000)
)
INSERT INTO Employee(id,name,deptIds)
SELECT 1,'蔣大華','1,2,3'
UNION
SELECT 2,'小明','1'
UNION 
SELECT 3,'小華',''

SELECT * FROM Employee

--創建一個表值函數，用來拆分用逗號分割的數字串，返回只有一列數字的表
IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[fun_SplitIds]'))
DROP FUNCTION [dbo].fun_SplitIds
GO
CREATE FUNCTION dbo.fun_SplitIds(
    @Ids nvarchar(1000)
)
RETURNS @t_id TABLE (id VARCHAR(36))
AS
BEGIN
    DECLARE @i INT,@j INT,@l INT,@v VARCHAR(36);
    SET @i = 0;
    SET @j = 0;
    SET @l = len(@Ids);
    while(@j < @l)
    begin
       SET @j = charindex(',',@Ids,@i+1);
       IF(@j = 0) set @j = @l+1;
       SET @v = cast(SUBSTRING(@Ids,@i+1,@j-@i-1) as VARCHAR(36));
       INSERT INTO @t_id VALUES(@v)
       SET @i = @j;
    END
    RETURN;
END
GO


;WITH EmployeT AS(
--員工的基本信息(使用OUTER APPLY將多個ID拆分開來,然後與部門表相關聯)
--此時已將員工表所存的IDS分別與部門相關聯,下面需要將此集合中的deptName聚合成一個記錄
SELECT E.*,ISNULL(D.name,'') AS deptName
FROM Employee AS E 
   OUTER APPLY dbo.fun_SplitIds(E.deptIds) AS DID
   LEFT JOIN Department AS D ON DID.ID=D.id
),mike AS(
   SELECT id,name,deptIds,deptName
      ,ROW_NUMBER()OVER(PARTITION BY id ORDER BY id) AS level_num
   FROM EmployeT
),mike2 AS(
   SELECT id,name,deptIds,CAST(deptName AS NVARCHAR(100)) AS deptName,level_num 
   FROM mike
   WHERE level_num=1
   UNION ALL
   SELECT m.id,m.name,m.deptIds,CAST(m2.deptName+','+m.deptName AS NVARCHAR(100)) AS deptName,m.level_num 
   FROM mike AS m
      INNER JOIN mike2 AS m2 ON m.ID=m2.id AND m.level_num=m2.level_num+1
),maxMikeByIDT AS(
       SELECT id,MAX(level_num) AS level_num
       FROM mike2 
       GROUP BY ID
)


SELECT A.id,A.name,A.deptIds,A.deptName
FROM mike2 AS A 
   INNER JOIN maxMikeByIDT AS B ON A.id=B.ID AND A.level_num=B.level_num
ORDER BY A.id OPTION (MAXRECURSION 0)