Recently, I've come to prefer mapping 1-1 relationships using Dictionaries instead of Switch statements. I find it to be a little faster to write and easier to mentally process. Unfortunately, when mapping to a new instance of an object, I don't want to define it like this:

var fooDict = new Dictionary<int, IBigObject>()
{
    { 0, new Foo() }, // Creates an instance of Foo
    { 1, new Bar() }, // Creates an instance of Bar
    { 2, new Baz() }  // Creates an instance of Baz
}

var quux = fooDict[0]; // quux references Foo

Given that construct, I've wasted CPU cycles and memory creating 3 objects, doing whatever their constructors might contain, and only ended up using one of them. I also believe that mapping other objects to fooDict[0] in this case will cause them to reference the same thing, rather than creating a new instance of Foo as intended. A solution would be to use a lambda instead:

var fooDict = new Dictionary<int, Func<IBigObject>>()
{
    { 0, () => new Foo() }, // Returns a new instance of Foo when invoked
    { 1, () => new Bar() }, // Ditto Bar
    { 2, () => new Baz() }  // Ditto Baz
}

var quux = fooDict[0](); // equivalent to saying 'var quux = new Foo();'

Is this getting to a point where it's too confusing? It's easy to miss that () on the end. Or is mapping to a function/expression a fairly common practice? The alternative would be to use a switch:

IBigObject quux;
switch(someInt)
{
    case 0: quux = new Foo(); break;
    case 1: quux = new Bar(); break;
    case 2: quux = new Baz(); break;
}

Which invocation is more acceptable?

• Dictionary, for faster lookups and fewer keywords (case and break)
• Switch: More commonly found in code, doesn't require the use of a Func<> object for indirection.