Question

如何优化以下查询？

DELETE FROM #Data 
WHERE ID NOT IN (SELECT MIN(ID)
                 FROM #Data 
                 GROUP BY SerialNumber, VendorName)

最多需要2分钟才能执行。

这是表结构：

CREATE TABLE #Data
(
    ID                 INT IDENTITY (1, 1),
    ItemSupplierKey    INT          NOT NULL,   
    SerialNumber       VARCHAR(100) NOT NULL,
    VendorName         VARCHAR(50)  NOT NULL
);

它包含257316条记录。

以下是此临时表的索引：

CREATE CLUSTERED INDEX PX_Data ON #Data (SerialNumber, VendorName);
CREATE INDEX IX_Data ON #Data (ID);

Answer 1

您可以重写删除以使用left join加入相同的#Data表并删除已加入记录null的位置：

delete d
from #Data d
    left join (
         SELECT MIN(ID) as ID FROM #Data GROUP BY SerialNumber, VendorName
    ) t1 on d.ID = t1.ID
where t1.ID is null

Answer 2

使用common table expression row_number()时，您应该会看到性能提升：

WITH CTE AS (
    SELECT *, ROW_NUMBER() OVER (PARTITION BY SerialNumber, VendorName ORDER BY ID) RN
    FROM Data
)
DELETE FROM CTE WHERE RN != 1

SQL查询调优

2 个答案: