Question

我有以下查询：

DECLARE @Cond_ition int = 2

SELECT T.Id, T.Description, T.A_Id, T.B_Id, T.C_Id FROM Table1 as T
JOIN TableA as A ON A.Id = T.A_Id AND A.GroupId = @Cond_ition
JOIN TableB as B ON B.Id = T.B_Id AND B.GroupId = @Cond_ition
JOIN TableC as C ON C.Id = T.C_Id AND C.GroupId = @Cond_ition

列A_Id，B_Id和C_Id在T的各行中具有少量的不同值。变量@Cond_ition实际上是该查询需要在其中运行的存储过程的参数。

查询将仅从T中返回在A，B和C中具有匹配项的行，除非这些表中的一个或多个表不具有匹配项，否则这很好。如果例如B为空或与A不匹配，则查询应返回T中与A和C匹配的行。如果B和C为空，则我需要T中与A匹配的行。全部为空或没有匹配项，那么我需要来自T的所有行。因此，如果有匹配项，则应充当INNER JOIN，但如果没有匹配项，则应充当LEFT JOIN。

关于如何从联接中获得这种双重“效果”的任何想法？

*编辑：根据要求，我添加了示例数据

Table T  
Id Description A_Id   B_Id    C_Id  
-- ----------- ----   ----    ----  
1  desc1       FANTA  CAN     LIGHT  
2  something2  SPRITE BOTTLE  LIGHT  
3  more3       SPRITE CAN     ZERO  
4  name4       7UP    BOTTLE  ZERO  
5  label5             GLASS   REGULAR  

**Case 1**
Table A
Id     GroupId
--     -------  
FANTA  2
SPRITE 2

Table B
Id   GroupId
--   -------  
CAN  5

*Expected result*
rows 1, 2 and 3

**Case 2**
Table A
Id     GroupId
--     -------  
FANTA  2
SPRITE 2

Table B
Id   GroupId
--   -------  
CAN  2
CAN  5

*Expected result*
rows 1 and 3

**Case 3**
Id     GroupId
--     -------  
FANTA  2
SPRITE 2

Table B
Id   GroupId
--   -------  
CAN  2
CAN  5

Table C
Id   GroupId
--   -------  
ZERO 2

*Expected result*
row 3

**Case 4**
Tables A, B and C empty or no rows with GroupId = 2

*Expected result*
all rows of T

Answer 1

如果我的理解正确，那么您只希望包含匹配项的行（如果有）。如果根本没有匹配的行，那么您需要table1中的所有行。

您可以使用UNION ALL处理此问题：

WITH tabc as (
      SELECT T.Id, T.Description, T.A_Id, T.B_Id, T.C_Id
      FROM Table1 T LEFT JOIN
           TableA A
           ON A.Id = T.A_Id AND A.GroupId = @Cond_ition LEFT JOIN
           TableB B
           ON B.Id = T.B_Id AND B.GroupId = @Cond_ition LEFT JOIN
           TableC C
           ON C.Id = T.C_Id AND C.GroupId = @Cond_ition
      WHERE T.A_Id IS NOT NULL OR T.B_Id IS NOT NULL OR T.C_Id IS NOT NULL
     )
SELECT tabc.*
FROM tabc
UNION ALL
SELECT t.*, NULL, NULL, NULL
FROM table1 t
WHERE NOT EXISTS (SELECT 1 FROM tabc);

Answer 2

感谢@gordonlinoff的投入，它帮助我找到了解决方案。这是符合我要求的查询：

SELECT T.Id, T.Description, T.A_Id, T.B_Id, T.C_Id
FROM Table1 T
JOIN (SELECT Id FROM TableA WHERE GroupId = @Cond_ition
      UNION
      SELECT TOP(1) '' WHERE NOT EXISTS (SELECT TOP(1) Id FROM TableA WHERE GroupId = @Cond_ition)
     )
  AS A ON ISNULL(T.A_Id,'') = CASE
    WHEN NOT EXISTS (SELECT TOP(1) Id FROM TableA WHERE GroupId = @Cond_ition) THEN ISNULL(T.A_Id,'')
    ELSE A.Id
    END
JOIN (SELECT Id FROM TableB WHERE GroupId = @Cond_ition
      UNION
      SELECT TOP(1) '' WHERE NOT EXISTS (SELECT TOP(1) Id FROM TableB WHERE GroupId = @Cond_ition)
     )
  AS B ON ISNULL(T.B_Id,'') = CASE
    WHEN NOT EXISTS (SELECT TOP(1) Id FROM TableB WHERE GroupId = @Cond_ition) THEN ISNULL(T.B_Id,'')
    ELSE B.Id
    END
JOIN (SELECT Id FROM TableC WHERE GroupId = @Cond_ition
      UNION
      SELECT TOP(1) '' WHERE NOT EXISTS (SELECT TOP(1) Id FROM TableC WHERE GroupId = @Cond_ition)
     )
  AS C ON ISNULL(T.C_Id,'') = CASE
    WHEN NOT EXISTS (SELECT TOP(1) Id FROM TableC WHERE GroupId = @Cond_ition) THEN ISNULL(T.C_Id,'')
    ELSE C.Id
    END

我在表A，表B和表C的联接中使用UNION。
首先选择获取与条件匹配的行，然后在没有匹配记录的情况下与常量合并。使用常量可确保UNION的结果永远不会为空。

在JOIN的ON子句中，我用CASE再次检查是否没有匹配的记录，在这种情况下，我将加入T本身的字段（使用ISNULL将返回所有行），否则我将加入子表。

我用想要查询的人更新了dbfiddle的查询。

在INNER JOIN和LEFT JOIN之间切换

2 个答案: