Another crazy C++ suggested addition that some other people here actually said made sense, much to my suprise:
Define the built-in type "0". Literally, the keyword is a zero digit character.
Functions/methods/constructors may be defined as taking a "0" argument. This means they can only be called with the literal value zero (any constant expression that can be assigned to a pointer and called the "null pointer" in current C++). You cannot declare variables or return values of the type "0", and you cannot name the "0" argument or use it in the implementation.
The purpose is of course to allow a class to define a conversion from null, without inadvertantly defining a conversion from any integer:
class witty_pointer<obj> {
public:
witty_pointer(obj*);
witty_pointer(0); // null pointer constructor
operator=(witty_pointer &);
operator=(0); // assignment from a null constant
};
I think that existance of a constructor from 0 would also mean that the automatic boolean cast of a!=0 would no longer exist and you would have to provide a cast to bool for "if (a)" to work. It may also be argued that there is an automatic cast to bool of "a != A(0)" but this could lead to unexpected code bloat.
'0' would also be useful for overloading functions that take pointers or integers in cases where the implementation for the value of zero is much simpler than for any other value. The compiler would then produce direct calls to the zero implementation if a constant zero is provided. It is up to the programmer to insure that the non-zero version acts just like the zero version if passed a zero that was calculated.
It may also be a good idea to say that static variables with a zero constructor are initialized by that constructor.
Bill Spitzak