查询优化：子查询中的JOIN

Question

给定以下查询：

SELECT [CustomerID], [CustomerName],[CustomerAddress],[CustomerPhone]
FROM Customers
WHERE CustomerID IN
    (SELECT CustomerID FROM Buys b 
    JOIN CarService cs ON cs.SoldCarNum=b.SoldCarNum        
    JOIN GarageWorkers gw ON cs.WorkerNum=gw.WorkerNum 
    WHERE YEAR(cs.ServiceDate) BETWEEN 2015 AND 2016
    GROUP BY CustomerID
    HAVING COUNT(DISTINCT gw.GarageID)>= 7)

有什么可能的方法使其更有效？ 我不喜欢JOIN，但看不到如何摆脱它们。

编辑：

我使用Microsoft SQL Server。

Answer 1

我猜这是您的查询：

SELECT [CustomerID], [CustomerName], [CustomerAddress], [CustomerPhone]
FROM Customers c
WHERE c.CustomerID IN (SELECT b.CustomerID
                       FROM Buys b JOIN
                            CarService cs
                            ON cs.SoldCarNum = b.SoldCarNum JOIN    
                            GarageWorkers gw
                            ON cs.WorkerNum = gw.WorkerNum 
                       WHERE YEAR(cs.ServiceDate) BETWEEN 2015 AND 2016
                       GROUP BY CustomerID
                       HAVING COUNT(DISTINCT gw.GarageID) >=  7
                      );

这似乎是在2015年和2016年拥有七辆或更多汽车服务的客户。首先，我将日期比较更改为使用实际日期而不是YEAR() （因此可以使用索引）。 看起来像：

SELECT [CustomerID], [CustomerName], [CustomerAddress], [CustomerPhone]
FROM Customers c
WHERE c.CustomerID IN (SELECT b.CustomerID
                       FROM Buys b JOIN
                            CarService cs
                            ON cs.SoldCarNum = b.SoldCarNum JOIN    
                            GarageWorkers gw
                            ON cs.WorkerNum = gw.WorkerNum 
                       WHERE cs.ServiceDate >= '2015-01-01' AND
                             cs.ServiceDate < '2017-01-01'
                       GROUP BY b.CustomerID
                       HAVING COUNT(DISTINCT gw.GarageID) >=  7
                      );

接下来，您需要索引。 我建议：

• CarService(ServiceDate, SoldCarNum, WorkerNum)
• Buys(SoldCarNum, CustomerId)
• GarageWorkers(WorkerNum, GarageID)

这些索引“覆盖”子查询，这意味着它们在子查询中具有所有列。

您没有提及数据库。 在某些数据库中，用JOIN (<subquery>)替换IN (<subquery>)通常也会提高性能。