在我看来，性能、名称空间污染等都是次要的。问问自己什么是合乎逻辑的。该方法在逻辑上操作类型的实例，还是与类型本身相关?如果是后者，则将其设置为静态方法。只有当它与不受您控制的类型相关时，才将其移动到实用程序类中。

Sometimes there are methods which logically act on an instance but don't happen to use any of the instance's state yet. For instance, if you were building a file system and you'd got the concept of a directory, but you hadn't implemented it yet, you could write a property returning the kind of the file system object, and it would always be just "file" - but it's logically related to the instance, and so should be an instance method. This is also important if you want to make the method virtual - your particular implementation may need no state, but derived classes might. (For instance, asking a collection whether or not it's read-only - you may not have implemented a read-only form of that collection yet, but it's clearly a property of the collection itself, not the type.)

对于类中的复杂逻辑，我发现私有静态方法在创建隔离逻辑时很有用，其中实例输入在方法签名中明确定义，并且不会发生实例副作用。所有输出必须通过返回值或out/ref参数。将复杂的逻辑分解成无副作用的代码块可以提高代码的可读性和开发团队对它的信心。

另一方面，它可能导致类被大量实用方法所污染。通常，逻辑命名、文档和团队编码约定的一致应用程序可以缓解这种情况。

使方法静态意味着您可以从类外部调用该方法，而无需首先创建该类的实例。这在使用第三方供应商对象或附加组件时非常有用。想象一下，如果你在调用con. writeline()之前必须先创建一个Console对象“con”;

Just to add to @Jason True's answer, it is important to realise that just putting 'static' on a method doesn't guarantee that the method will be 'pure'. It will be stateless with regard to the class in which it is declared, but it may well access other 'static' objects which have state (application configuration etc.), this may not always be a bad thing, but one of the reasons that I personally tend to prefer static methods when I can is that if they are pure, you can test and reason about them in isolation, without having to worry about the surrounding state.

ReSharper不检查逻辑。它只检查方法是否使用实例成员。 如果方法是私有的，并且只被(可能只有一个)实例方法调用，这是一个让它成为实例方法的标志。

在我看来，性能、名称空间污染等都是次要的。问问自己什么是合乎逻辑的。该方法在逻辑上操作类型的实例，还是与类型本身相关?如果是后者，则将其设置为静态方法。只有当它与不受您控制的类型相关时，才将其移动到实用程序类中。

Sometimes there are methods which logically act on an instance but don't happen to use any of the instance's state yet. For instance, if you were building a file system and you'd got the concept of a directory, but you hadn't implemented it yet, you could write a property returning the kind of the file system object, and it would always be just "file" - but it's logically related to the instance, and so should be an instance method. This is also important if you want to make the method virtual - your particular implementation may need no state, but derived classes might. (For instance, asking a collection whether or not it's read-only - you may not have implemented a read-only form of that collection yet, but it's clearly a property of the collection itself, not the type.)