Anonymous Methods - Understanding Scope

Anonymous Methods were introduced in .net 2.0. If you haven't used or seen them before I suggest reading some of the many blogs and MSDN articles about them. I'm not going to introduce them here, I'm going to briefly look into how the scope works when they are invoked from outside the scope they were created in.

Below I have a simple example of a class that has a member variable and a method that returns an anonymous method. The only purpose of this little class is to investigate some scope issues.

public delegate string FooDelegate();

public class Foo
{
    public string Bar { get; set; }

    public FooDelegate GetDelegate()
    {
        string bar = Bar;
        return delegate()
            {
                return string.Format("Member: {0}; Local: {1}", Bar, bar);
            };
    }
}

You can see that in this class the anonymous method we are creating references a local variable and a member variable of the class. This is completely legitimate code in .net 2.0, 3.0 and 3.5. Now lets see how it works, below is a simple example.

Foo foo = new Foo();
foo.Bar = "test";

FooDelegate fooDelegate = foo.GetDelegate();

string result = fooDelegate.Invoke();

The result is "Member: test; Local: test"

You can see that when we invoked the anonymous method both the local and member variables evaluated to "test". Once you accept its ok to reference variables from the scope the anonymous method was created in, this behavior makes sense.

But what happens if we change the property after we create the anonymous method? What would we expect the values to be when we invoke the method? If you've used anonymous methods in other languages it might be quite obvious. If you don't know, or want to validate that you do, the example below will do just that.

Foo foo = new Foo();
foo.Bar = "test";

FooDelegate fooDelegate = foo.GetDelegate();
foo.Bar = "update";

string result = fooDelegate.Invoke();

The result is "Member: update; Local: test"

As you can the see the method did use the updated property, but only for the class property. The value of the local variable it was referring to when the anonymous method was created has not been changed.

Normally a local variable would be discarded and later collected by the GC once execution of the block is complete. In this case it's still in the scope of the anonymous method and is therefore not discarded while the anonymous method is still referenced.

The local variable is not updated with the new value as we set the local variable when we created the delegate, and we changed the class property after that. In this case we can't change the value of that variable as we no longer have a reference to it once we leave the GetDelegate() method. Its only reference by the delegate itself.

This example works because strings create a new instance of a string when they are assigned. The example would behave differently if the local variable was a reference to the same object as the class property.

Using the local variable in an anonymous method that is called from outside scope it was created might not be something you ever do in .Net, but its a concept .Net developers should understand. I've haven't been able to think of a simple, relevant, real world .net example for this, but I do use the same concept in some Javascript I write about in another post.

EDIT: This is known as closure. This article I found on the VS2008 Start Page talks about it and discusses the MSIL generated : Lambdas - Know Your Closures