When I need a break, I often look what is on the horizon for my favorite programming languages to get excited about: Concepts in C++, better value-type support in C#. But sometimes I also get to learn something new about a language I thought I knew inside out. Today was such a day: I learned about partial methods in C#.

Partial Classes

We are all familiar with partial classes in C#:

// in file X_1.cs
partial class X {
    // whatever you want in your class
}

// in file X_2.cs
partial class X {
    // whatever else you want in your class
}

At compile time, the two declarations are merged. All fields, properties, methods, etc. from each of the two declarations are present in the final definition of X. If you have ever worked with WinForms and the designer for that in VisualStudio, then you have probably been using a lot of partial classes: The WinForms editor generated most of the UI code and you merely implemented a few handlers. Less known is that you can actually vary the class declaration for partial classes (in some places). For example:

// in file X.cs
partial class X {
    // whatever you want in your class
}

// in file X_IWhatever.cs
partial class X : IWhatever {
    // define how X implements the IWhatever interface
}

The interfaces that X implements are simply collected from all declarations of a partial class. I have never used it, but I am sure that there are places where you can put this to good use to put the implementation of an interface into another file.

Partial Methods

Within a partial class, you can define a partial method like this:

partial class X {
    // declare the method in on place
    partial void myPartialMethod();
}

partial class X {
    // define it in another
    partial void myPartialMethod() {

    }
}

This seems to only split the declaration of the method from its definition (as is common practice in C++), but there is more to it. The main feature is that you don’t have to provide a definition. You can simply forgo the implementation. In that case, whenever a call to the undefined-but-declared partial method is made, the compiler will silently drop it. Due to this behavior, there are strict restrictions on what methods can be marked partial:

  • its return type must be void,
  • it cannot declare an access modifier and is automatically private,
  • it cannot have out parameters (but ref is fine),
  • it cannot be extern.

For more information, see the C# programming guide. In the context of GUI development and WinForms, this behavior of partial methods makes perfect sense. The programming guide mentions that partial methods are to be treated like events (think OnButtonClicked etc.), and there is no need to respond to every single event.

Is there another good usecase for partial methods? I am not sure whether I would really advise making use of this feature. If you want a quick way to disable a call to a method without changing each callsite, you should rather use the Conditional attribute. The particularly nasty thing about partial methods is that by design the compiler won’t complain if the implementation is missing, and that by itself almost entirely rules out using this feature.