SQL cartesian加入问题

时间:2009-05-08 11:56:02

标签: sql join hierarchical-data cartesian-product

我有三张桌子

  • 答:A.pID主键,A.Name nvarchar(250)
  • B:B.pID主键,B.Name nvarchar(250)
  • C:C.pID主键,C.Name nvarchar(250)

A和B之间存在n关系(表lA_B与主键lA_B.pID.pInstanceA外键到表A和.pInstanceB外键到表B )

A和C之间存在n关系(table lA_C与主键lA_C.pID.pInstanceA外键到表A和.pInstanceB外键到表C)

  • A1与B1,B2和C1
  • 有关
  • A2与B3和C2,C3
  • 有关
  • A3与B4相关
  • A4与C4相关
  • A5没有关系

这是我的SQL:

CREATE TABLE [dbo].[A]( [pID] [bigint] NOT NULL, [Name] [nvarchar](250) NULL )
CREATE TABLE [dbo].[B]( [pID] [bigint] NOT NULL, [Name] [nvarchar](250) NULL )
CREATE TABLE [dbo].[C]( [pID] [bigint] NOT NULL, [Name] [nvarchar](250) NULL)
CREATE TABLE [dbo].[lA_B]( [pID] [bigint] NOT NULL, [pInstanceA] [bigint] NULL, [pInstanceB] [bigint] NULL )
CREATE TABLE [dbo].[lA_C]( [pID] [bigint] NOT NULL, [pInstanceA] [bigint] NULL, [pInstanceB] [bigint] NULL )

INSERT INTO [dbo].[A] ([pID] ,[Name]) VALUES (1,'A1')
INSERT INTO [dbo].[A] ([pID] ,[Name]) VALUES (2,'A2')
INSERT INTO [dbo].[A] ([pID] ,[Name]) VALUES (3,'A3')
INSERT INTO [dbo].[A] ([pID] ,[Name]) VALUES (4,'A4')
INSERT INTO [dbo].[A] ([pID] ,[Name]) VALUES (5,'A5')

INSERT INTO [dbo].[B] ([pID] ,[Name]) VALUES (1,'B1')
INSERT INTO [dbo].[B] ([pID] ,[Name]) VALUES (2,'B2')
INSERT INTO [dbo].[B] ([pID] ,[Name]) VALUES (3,'B3')
INSERT INTO [dbo].[B] ([pID] ,[Name]) VALUES (4,'B4')

INSERT INTO [dbo].[C] ([pID] ,[Name]) VALUES (1,'C1')
INSERT INTO [dbo].[C] ([pID] ,[Name]) VALUES (2,'C2')
INSERT INTO [dbo].[C] ([pID] ,[Name]) VALUES (3,'C3')
INSERT INTO [dbo].[C] ([pID] ,[Name]) VALUES (4,'C4')

INSERT INTO [dbo].[lA_B] ([pID],[pInstanceA],[pInstanceB])   VALUES   (1,1,1)
INSERT INTO [dbo].[lA_B] ([pID],[pInstanceA],[pInstanceB])   VALUES   (2,1,2)
INSERT INTO [dbo].[lA_B] ([pID],[pInstanceA],[pInstanceB])   VALUES   (3,2,3)
INSERT INTO [dbo].[lA_B] ([pID],[pInstanceA],[pInstanceB])   VALUES   (4,3,4)

INSERT INTO [dbo].[lA_C] ([pID],[pInstanceA],[pInstanceB])   VALUES   (1,1,1)
INSERT INTO [dbo].[lA_C] ([pID],[pInstanceA],[pInstanceB])   VALUES   (2,2,2)
INSERT INTO [dbo].[lA_C] ([pID],[pInstanceA],[pInstanceB])   VALUES   (3,2,3)
INSERT INTO [dbo].[lA_C] ([pID],[pInstanceA],[pInstanceB])   VALUES   (4,4,4)

此查询:

SELECT
  A.Name AS A, 
  B.Name AS B, 
  C.Name AS C
FROM
  A  
  left JOIN lA_B  ON (A.pID = lA_B.pInstanceA)
  left JOIN B     ON (B.pID = lA_B.pInstanceB)
  left JOIN lA_C  ON (A.pID = lA_C.pInstanceA)
  left JOIN C     ON (C.pID = lA_C.pInstanceB) 

返回

A1      B1      C1
A1      B2      C1
A2      B3      C2
A2      B3      C3
A3      B4      NULL
A4      NULL    C4
A5      NULL    NULL 

现在的问题是:-) 如何查询接收

A1      B1      NULL
A1      B2      NULL
A1      NULL    C1
A2      B3      NULL
A2      NULL    C2
A2      NULL    C3
A3      B4      NULL
A4      NULL    C4
A5      NULL    NULL 

问题在于,当我使用B和C进行连接时,结果具有B C的所有组合。如何消除这种情况?

2 个答案:

答案 0 :(得分:4)

您可以使用UNION执行此操作:

SELECT A.Name AS A, B.Name AS B, NULL AS C 
FROM A
left JOIN lA_B ON (A.pID=lA_B.pInstanceA) 
left JOIN B ON (lA_B.pInstanceB=B.pID) 
UNION
SELECT A.Name AS A, NULL AS B, C.Name AS C 
FROM A
left JOIN lA_C ON (A.pID=lA_C.pInstanceA) 
left JOIN C ON (lA_C.pInstanceB=C.pID)

第一部分选择A和B的所有组合,第二部分选择A和C的所有组合。

如果您希望过滤掉(A4,NULL,NULL)等行,因为已经存在一行(A4,NULL,C4),请尝试以下查询:

SELECT A.Name AS A, B.Name AS B, NULL AS C 
FROM A
LEFT JOIN lA_B ON (A.pID=lA_B.pInstanceA) 
LEFT JOIN B ON (lA_B.pInstanceB=B.pID) 
WHERE b.name is not null 
    or not exists(select * from lA_C where A.pID=lA_C.pInstanceA) 
UNION
SELECT A.Name AS A, NULL AS B, C.Name AS C 
FROM A
LEFT JOIN lA_C ON (A.pID=lA_C.pInstanceA) 
LEFT JOIN C ON (lA_C.pInstanceB=C.pID)
WHERE c.name is not null 
ORDER BY A,B,C

对于B上的连接,这表示包含在B中具有匹配的行,或者在C中没有匹配的行.C上的连接仅包括在C中匹配的行。不匹配的行将是从B的加入中获得。

请注意,UNION会过滤掉重复的行,例如DISTINCT。要包含每一行,您可以使用UNION ALL。

答案 1 :(得分:0)

我认为这样做:

select a,
       case when b='zzz' then null else b end as b,
       case when c='zzz' then null else c end as c
from (SELECT  A.Name AS A
             ,b.Name as b
             ,'zzz' as c
      FROM  A    
      JOIN lA_B  ON (A.pID = lA_B.pInstanceA)  
      JOIN B     ON (B.pID = lA_B.pInstanceB)  
      union
      select  a.Name
             ,'zzz'
            ,c.NAme
      from A
      left JOIN lA_C  ON (A.pID = lA_C.pInstanceA)  
      left JOIN C     ON (C.pID = lA_C.pInstanceB)) as a
相关问题