sql - 使用 DELETE 逻辑 NOT IN 嵌套查询的最快性能

Question

我正在尝试为这个 DELETE（和 SELECT）查询获得最快的性能。有没有更好的方法来删除记录，因为这需要 10 多分钟才能运行？我想它必须进行自己的排序和合并，直到它可以找到记录。

SELECT COUNT([VISIT_ID])
FROM [dbo].[I2B2_SRC_VISITS]
WHERE [PATIENT_ID] NOT IN (
    SELECT [PATIENT_ID] FROM [dbo].[I2B2_SRC_PATIENT]
)

DELETE FROM [dbo].[I2B2_SRC_VISITS]
WHERE [PATIENT_ID] NOT IN (
    SELECT [PATIENT_ID] FROM [dbo].[I2B2_SRC_PATIENT]
)

编辑：我不能像使用 SELECT 那样将 DELETE 放在该查询前面。但这是 DELETE 语句的最终结果。

DELETE FROM [dbo].[I2B2_SRC_VISITS]
WHERE [VISIT_ID] IN
(
    SELECT a.[VISIT_ID]
    FROM    [dbo].[I2B2_SRC_VISITS] a
            LEFT JOIN [dbo].[I2B2_SRC_PATIENT] b
                ON a.[PATIENT_ID] = b.[PATIENT_ID]
    WHERE b.[PATIENT_ID]  IS NULL
)

Answer 1

怎么样做JOIN呢？

DELETE  a
FROM    [dbo].[I2B2_SRC_VISITS] a
        LEFT JOIN [dbo].[I2B2_SRC_PATIENT] b
            ON a.[PATIENT_ID] = b.[PATIENT_ID]
WHERE   b.[PATIENT_ID] IS NULL

确保[PATIENT_ID]两个表中的列都有键定义，这使得它更快。

---

正确的。NOT EXIST是更好的一个。

DELETE  a 
FROM    [dbo].[I2B2_SRC_VISITS] a 
WHERE   NOT EXISTS
        ( 
            SELECT  1 
            FROM    [dbo].[I2B2_SRC_PATIENT] b
            WHERE   a.[PATIENT_ID] = b.[PATIENT_ID] 
        )