Question

A colleague asked me an interesting question today – is the C# keyword/operator “is” considered reflection?

object tmp = "a string";
if(tmp is String)
{
}

How is this operator implemented behind the scenes? Does it require reflection or introspection? Or because of the strongly typed nature of the language, is the Type of the object immediately accessable as a top-level attribute of the object in memory?

MSDN states that:

Note that the is operator only considers reference conversions, boxing conversions, and unboxing conversions. Other conversions, such as user-defined conversions, are not considered by the is operator.

The ability to consider boxed and unboxed conversions seems to imply to me some sort of introspection.

Answer 1

Referencing ECMA-335, the is operator generates the isinst object model IL instruction (Partition III §4.6), which is part of the base instruction set as opposed to being part of the Reflection library (Partition IV §5.5).

Edit: The is operator is extremely efficient compared to the reflection library. You could perform basically the same test much more slowly via reflection:

typeof(T).IsAssignableFrom(obj.GetType())

Edit 2: You are not correct about the efficiency of the castclass and isinst instructions (which you’ve now edited out of the post). They are highly optimized in any practical VM implementation. The only real performance issue involved is the potential for castclass to throw an exception, which you avoid by using the C# as operator and a test for null (for reference types) or the is operator followed by a cast (for value types).