An obvious singleton implementation for .NET?

Question

I was thinking about the classic issue of lazy singleton initialization - the whole matter of the inefficiency of:

if (instance == null)
{
    instance = new Foo();
}
return instance;

Anyone who knows what a Singleton is is familiar with the issue(you only need the if once). It's trivial but irritating.

So, I thought of an alternate solution, at least for .NET(although it should work anywhere that has some equivalent to function pointers:

public class Foo
{
    private delegate Foo FooReturner();

    private static Foo innerFoo;

    private static FooReturner fooReturnHandler = new FooReturner(InitialFooReturner);

    public static Foo Instance
    {
        get
        {
            return fooReturnHandler();
        }
    }
    private static Foo InitialFooReturner()
    {
        innerFoo = new Foo();
        fooReturnHandler = new FooReturner(NewFooReturner); 
        return innerFoo;
    }

    private static Foo NewFooReturner()
    {
        return innerFoo;
    }

}

In short - the Instance returns a delegate method. The delegate is initially set to a method that initializes your instance, then changes the delegate to point at a simple Return method.

Now, I like to think I'm not terrible at my job, but I have no pretensions about being awesome. I have not seen an example of this code anywhere.

Ergo, I come to the conclusion that I am missing something. Something major. Either that the whole problem is too trivial to bother thinking that much about or this does something horrible that will destroy the universe. Or I fail at searching and therefore haven't seen the hundreds of developers using this method. Something, anyway.

I was hoping the good folks here at Stack Overflow could clue me in as to what(leaving aside the controversy on whether one should use a Singleton at all).

EDIT for clarification:

This is not performance code(although if the design actively degrades performance beyond the traditional model, that would be interesting to know).

It was written purely as proof-of-concept, and I am further aware that it is not thread-safe as it properly should be. Is there any reason why it could NOT be made thread-safe by it's very nature?

Answer 1

This is the canonical, thread safe, lazy Singleton pattern in C#:

public sealed class Singleton
{
    Singleton(){}
    public static Singleton Instance
    {
        get
        {
            return Nested.instance;
        }
    }        
    class Nested
    {
        // Explicit static constructor to tell C# compiler
        // not to mark type as beforefieldinit
        static Nested() {}    
        internal static readonly Singleton instance = new Singleton();
    }
}

Answer 2

the whole matter of the inefficiency of:...

What inefficiency?

Those instructions will result into an extremely fast fragments of assembly code. I am completely sure that there is nothing to be gained by trying to "optimize" this. Even if you come up with something faster, it will be at a significant complexity cost.

Unless you do have positive evidence that this code is affecting your performance, you should use the simplest approach that solves your problem.

Answer 3

Have you measured the performance?

Do you believe that an additional function call is cheaper than an if?

I agree with the others that using the static ctor works well for this initialization. This will also get rid of the inherent race condition that the you have since .net guarentees that static constructors will only be called once.

Answer 4

To prevent from having to copy the singleton code, you could make the type generic, as such:

public abstract class Singleton<T>
    where T: class, new()
{
    public static T Instance
    {
        get { return Nested.instance; }
    }

    private class Nested
    {
        // Explicit static constructor to tell C# compiler
        // not to mark type as beforefieldinit
        static Nested() { }

        internal static readonly T instance = new T();
    }
}

public sealed class MyType : Singleton<MyType>
{
}

class Program
{
    static void Main()
    {
        // two usage pattterns are possible:
        Console.WriteLine(
            ReferenceEquals(
                Singleton<MyType>.Instance, 
                MyType.Instance
            )
        );
        Console.ReadLine();
    }
}

Answer 5

I always use this regular (non-lazy) one (you could nest it like the other examples): (requires using System.Reflection;)

public class SingletonBase<T> where T : class
{
    static SingletonBase()
    {
    }

    public static readonly T Instance = 
        typeof(T).InvokeMember(typeof(T).Name, 
                                BindingFlags.CreateInstance | 
                                BindingFlags.Instance |
                                BindingFlags.Public |
                                BindingFlags.NonPublic, 
                                null, null, null) as T;
}