In what situations is static method a good practice?

Question

I have read the following discussions:

Should private helper methods be static if they can be static , and
Should all methods be static if their class has no mem

Answer 1

It's much easier to refactor methods that are static. It discourages unnecessary references to field members that make the coupling tight. It's also easier to understand calling code because it explicitly passes any objects it interacts with.

Answer 2

Util classes can be OK to be static. As in someone elses example above, with escaping strings. The problem lies when those utils classes perform functions that we would want to mock. For example, the FileUtils class in apache commons - it is often the case that I want to mock file interactions without having to play with real files. If this was an instance class, it would be easy.

Answer 3

Static methods aren't hard to test in and of themselves. The problem is that other code calling the static method is hard to test, because you can't replace the static methods.

I think static methods are fine either when they're private or when they're "utility" methods - e.g. to do string escaping. The problem comes when you use static methods for things that you want to be able to mock out or otherwise replace within tests. Factory methods can be useful too, although dependency injection is generally a better approach - again, it partly depends on whether you want to be able to replace the functionality in tests.

As for not being "OO" - not everything you write in a generally OO language has to be "pure" OO. Sometimes the non-OO route is simply more pragmatic and leads to simpler code. Eric Lippert has a great blog post about this, which unfortunately I can't find right now. However, there's a comment in this post which is relevant. It talks about extension methods rather than static methods, but the principle is the same.

Extension methods are often criticized as being "not OOP enough". This seems to me to be putting the cart in front of the horse. The purpose of OOP is to provide guidelines for the structuring of large software projects written by teams of people who do not need to know the internal details of each other's work in order to be productive. The purpose of C# is to be a useful programming language that enables our customers to be productive on our platforms. Clearly OOP is both useful and popular, and we've therefore tried to make it easy to program in an OOP style in C#. But the purpose of C# is not "to be an OOP language". We evaluate features based on whether they are useful to our customers, not based on whether they conform strictly to some abstract academic ideal of what makes a language object-oriented. We'll happily take ideas from oo, functional, procedural, imperative, declarative, whatever, so long as we can make a consistent, useful product that benefits our customers.

Answer 4

I often use static factory methods instead of or in conjunction with public constructors.

I do this, when i need a constructor that does something you would not expect a constructor to do. I.e. load settings from a file or database.

This approach also provides the possibility of naming the "constructors", based on what they do. This is especially useful, when the parameters themselves are not enough to figure out what happens in a constructor.

Sun uses this approach in

Class.forName("java.lang.Integer");

and

Boolean.valueOf("true");

Answer 5

Generally I avoid static methods when a single instance will work fine. That single instance can implement an interface and can be mocked easily. I'm not saying never but I rarely use statics. I tell my team if they want to use a static it should be cleared by the team. Almost never is my answer.

Answer 6

Static methods are fine in most situations where the singleton pattern gives too much flexibility.

For example, take a simple utility such as raising a primitive to a power - obviously you never need to have any polymorphism in that. Primitive values are of static type and mathematical operations are well defined and don't change. It's not like you'll ever get the situation of having two different implementations an no way of switching between them without rewriting all your client code.

---

(irony off )

Modern JVMs are pretty good at inlining small calls if only one implementation of an interface is loaded. Unless you have profiled your code and know dispatching your utilities to an interface is an overhead, you've no excuse for not making your utility methods into an interface which can be varied if required.