Question

What is a good usage of the is-operator?

The construct below for casting is not the recommended way to go, virtually all documentation prefers the as-operator with a null-check.

if(obj is SomeClass)
{
  SomeClass some = (SomeClass)obj;
  ....
}

And sure this is a (very small) performance increase and some even mention the tread safety.
And yes this is true…

So, why do we have the is-operator?
Where does the “as-operator with a null-check” not work or is not the way to go?
Does is have an advantage to restrict the scope of you declaration you get by using the is-operator?

Answer 1

as doesn’t work with non-nullable structs:

object o = 123;
int i = o as int; // compile error

however:

object o = 123;
if(o is int) {
    int i = (int)o;
    //...
}

of course, from 2.0 onwards you could also use:

int? i = o as int?;

and test for null like usual.

There is also the scenario that you don’t care about the values of the object… you just need to know what it is:

if(obj is Something)
    throw new InvalidOperationException("Seriously, don't do that");
// phew! dodged a bullet; we're ok here...

Note that GetType() is not appropriate for this, as you don’t want to have to consider subclasses, interfaces, etc manually.