在 SQL 中检测到重叠，未在选择中使用子查询且未将重复记录添加到结果集中

Question

我希望创建一个可以在 SQL 中检测到重叠的查询，而无需在选择中使用子查询，也不会向结果集中添加重复记录。

我知道可以这样做，但考虑到两个表中的记录数量，我正在寻找更快的解决方案。

SELECT x.T1Id
    ,CASE 
        WHEN MaxOtherColumn IS NULL
            THEN 0
        ELSE 1
        END HasOverlapInT2
    ,CASE 
        WHEN MaxOtherColumn IS NULL
            THEN OtherColumn
        ELSE MaxOtherColumn
        END MaxOtherColumn
FROM (
    SELECT CASE 
            WHEN EXISTS (
                    SELECT MAX(T2.OtherColumn)
                    FROM T2
                    WHERE T1.ValidFrom <= T2.ValidUntil
                        AND T2.ValidFrom <= T1.ValidUntil
                        AND t1.T1Id = T2.T1Id
                    )
                THEN 1
            ELSE NULL
            END MaxOtherColumn
        ,T1Id
        ,OtherColumn
    FROM T1
    ) x

我不想对每一行都执行存在，而是批量处理。

样本数据

表 1

T1Id (PK)	ValidFrom	ValidUntil	OtherColumn
1	2021-01-01	2021-12-31	1
2	2021-09-23	2021-09-24	2

表 2

T2Id (PK)	T1Id(FK)	ValidFrom	ValidUntil	OtherColumn
1	1	2021-01-01	2021-05-31	9000
2	1	2021-06-01	2021-12-31	9001
3	2	2021-01-01	2021-01-05	7000

预期结果

T1Id	HasOverlapInT2	MaxOtherColumn
1	1	9001
2	0	2

SQL 服务器兼容级别 = 130 (2016)

Answer 1

您可以使用CROSS APPLY 来实现您的预​​期结果（您当前的解决方案不会返回您的预期结果）：

SELECT x.T1Id,
    CASE 
        WHEN MaxOtherColumn IS NULL
        THEN 0
        ELSE 1
    END HasOverlapInT2,
    CASE 
        WHEN MaxOtherColumn IS NULL
        THEN OtherColumn
        ELSE MaxOtherColumn
    END MaxOtherColumn
FROM
(
    SELECT t.MaxOtherColumn,
        T1Id,
        OtherColumn
    FROM T1
    CROSS APPLY
    (
        SELECT MAX(T2.OtherColumn) MaxOtherColumn
        FROM T2
        WHERE T1.ValidFrom <= T2.ValidUntil
        AND T2.ValidFrom <= T1.ValidUntil
        AND t1.T1Id = T2.T1Id
    ) t
) x

Answer 2

一种稍微不同的思考方式，以防APPLY 太令人费解：

;WITH src AS 
(
  SELECT T1.T1Id, 
    T2Other = T2.OtherColumn, 
    T1Other = T1.OtherColumn,
    Overlap = CASE WHEN T2.ValidUntil >= T1.ValidFrom
              AND T2.ValidFrom <= T1.ValidUntil THEN 1 
              ELSE 0 END
  FROM dbo.T1
  INNER JOIN dbo.T2
    ON T1.T1Id = T2.T1Id
)
SELECT T1Id, 
  HasOverlapInT2 = MAX(Overlap), 
  MaxOtherColumn = MAX(CASE Overlap WHEN 1 THEN T2Other ELSE T1Other END)
FROM src
GROUP BY T1Id;

• 示例db<>fiddle

关键是要考虑一种方法，通过一次通过两者来生成结果。虽然相关子查询并不总是意味着性能更差，但它可能会导致扫描一个表中的每一行。