Question

我正在尝试从非常大的Audits表（数百万行）中检索数据。所以我需要尽可能高效地运行查询。首先我正在使用子查询返回ObjectTypeId并使用它来限制Audit表上的查询

此查询需要4分钟才能运行：

select distinct Audits.ObjectTypeID,  COUNT(*) as Count
from Audits as Audits
where Audits.ObjectTypeID = 
(select distinct ObjectType.ObjectTypeID from ObjectType where ObjectName = 'Data')
group by Audits.ObjectTypeID

如果我在ObjectTypeID中默认，则查询在42秒内运行

select distinct(Audits.ObjectTypeID), COUNT(*) as Count
from Audits 
where Audits.ObjectTypeID = 1
group by Audits.ObjectTypeID

但是，在隔离运行时子查询只需要一秒钟就可以运行。那么为什么第一个查询需要这么长时间？

Answer 1

我可以看到三件可能有用的事情：

将ObjectTypeID拉入变量：因为它应该只有一个值
在两个查询上取出DISTINCT，因为它们应该是不必要的（子查询应该只有一个值，并且您在外部查询中按该值进行分组
取出GROUP BY，因为您只查询一个ObjectTypeID

所以最终的查询是：

DECLARE @ObjectTypeID INT

SELECT @ObjectTypeID = (select ObjectType.ObjectTypeID 
                        from ObjectType 
                        where ObjectName = 'Data')

select Audits.ObjectTypeID,  COUNT(*) as Count
from Audits as Audits
where Audits.ObjectTypeID = @ObjectTypeID

如果您将此作为单个语句执行而不是作为批处理或存储过程（意味着您不能使用变量），则可以保留子查询：

select Audits.ObjectTypeID,  COUNT(*) as Count
from Audits as Audits
where Audits.ObjectTypeID = 
    (select ObjectType.ObjectTypeID 
     from ObjectType 
     where ObjectName = 'Data')

Answer 2

您获得最佳效果的部分可能是这一行：

where Audits.ObjectTypeID = 
(select distinct ObjectType.ObjectTypeID from ObjectType where ObjectName = 'Data')

实际上，您正在对表的每一行调用相同的查询，它将搜索整个ObjectType表并返回该子查询的整个结果。如果您的ObjectType表格很大，这将是一个巨大的性能影响。您可以使用EXISTS加快查询的该部分，以便在找到结果后提前返回。这是一个例子：

SELECT a.ObjectTypeID,  COUNT(*) as Count
FROM Audits a
WHERE EXISTS 
(
    SELECT ot.ObjectTypeID
    FROM ObjectType ot 
    WHERE ot.ObjectName = 'Data' AND ot.ObjectTypeID = a.ObjectTypeID
)
GROUP BY a.ObjectTypeID

Answer 3

你能试试吗

SELECT DISTINCT Audits.ObjectTypeID,  COUNT(*) as Count
FROM Audits as Audits
INNER JOIN 
(SELECT DISTINCT ObjectTypeId, ObjectName FROM ObjectType 
 WHERE ObjectName = 'Data') as ObjectType ON Audits.ObjectTypeID = ObjectType.ObjectTypeID
GROUP BY Audits.ObjectTypeID

SQL慢子查询

3 个答案: