Question

I ran with strangeness:

var k = (decimal?)(int?)1; //valid cast
var p = (decimal?)(int?)(object)(int?)1; //valid cast
var l = (decimal?)(object)(int?)1; //Specified cast is not valid

Can someone explain why this happens?

Answer 1

The first cast is:

• Convert int to int?
• Convert int? to decimal?

Both conversions are valid.

The second case is:

• Convert int to int?
• Box the int? (which ends up as a boxed int)
• Unbox the object to int?
• Convert int? to decimal?

All of these conversions are valid.

The third case is:

• Convert int to int?
• Box the int? (which ends up as a boxed int)
• Unbox the object to decimal?

The last conversion here is invalid – you can only unbox to the same value type or its nullable equivalent. (Actually the CLR is somewhat more forgiving than this, but that’s not relevant in this case.)

The conversion to int? in each case is actually irrelevant. The unboxing to a nullable type is somewhat irrelevant too, in that unboxing to a nullable type is like unboxing to a non-nullable type, except that a null reference is unboxed to a null value. Given that there are no null values here, your final example is equivalent to:

object o = 1; // Boxing
decimal d = (decimal) o; // Unboxing, but to the wrong type