2008/5/29, Scott McMurray <me22.ca+boost@gmail.com>:
On Tue, May 27, 2008 at 5:06 PM, Joachim Faulhaber <afojgo@googlemail.com> wrote:
My generic library code has no problem with the 'value nature' of built in types. Because from the initial default constructed value
unsigned int x = int(); // x = 0 you can reach ++x // x = 1 ... ++x // x = 2^32
and thus all values of the types set of values.
This is not the case for boost::date_time objects like e.g. ptime
Nor is it the case for floating point numbers.
Yes, but I specifically addressed integral numerical types here
Nor, technically, for signed numbers, where overflow is impl-defined (or worse, afaik).
speaking of the finite set of machine represented values and their reachability by any operators (e.g. --) of a signed integral numerical type, yes. I am only emphasizing on default ctor and ++ because these were the minimally required operations on an integral types T in itl::interval<T> to write code that seemed to be fairly generic. Regards Joachim