Question

I’m asking because I’m trying to use a mocking framework (Mockito) which does not allow you to mock static methods. Looking into it I’ve found quite a few blog posts saying that you should have as few static methods as possible, but I’m having difficulty wrapping my head around why. Specifically why methods that don’t modify the global state and are basically helper methods. For instance I have a class called ApiCaller that has several static methods. One of the static method’s purpose is to execute an HTTP call, deal with any custom issues our server might have returned (ex. user not logged in) and return the response. To simplify, something like:

public class ApiCaller {
...
   public static String makeHttpCall(Url url) {
        // Performs logic to retrieve response and deal with custom server errors
        ...
        return response;
   }
}

To use this all I have to do is call ApiCaller.makeHttpCall(url)
Now I could easily make this a non static method like:

public class ApiCaller {
...
   public String makeHttpCall(Url url) {
        // Performs logic to retrieve response and deal with custom server errors
        ...
        return response;
   }
}

and then to use this method call new ApiCaller().makeHttpCall() but this just seems like extra overhead. Can anyone explain why this is bad and if there is a better solution to making the methods non static (other than just removing the keyword) so that I can stub out these methods using the mocking framework?

Thanks!

Answer 1

The problem with static methods is they’re very hard to fake when they’re not relevant to the system you’re trying to test. Imagine this code:

public void systemUnderTest() {
    Log.connectToDatabaseForAuditing();
    doLogicYouWantToTest();
}

The connectToDatabaseForAuditing() method is static. You don’t care what this method does for the test you want to write. But, to test this code now you need an available database.

If it were not static the code would look like this:

private Logger log; //instantiate in a setter AKA dependency injection/inversion of control

public void systemUnderTest() {
    log.connectToDatabaseForAuditing();
    doLogicYouWantToTest();
}

And your test would be trivial to write without a database now:

@Before
public void setUp() {
    YourClass yourClass = new YourClass();
    yourClass.setLog(new NoOpLogger());

}

//.. your tests

Imagine trying to do that when the method is static. I can’t really think of a way except for modifying the logger to have a static variable called inTestMode that you set to true in the setUp() to make sure it doesn’t connect to a database.