Re: [boost] [xint] Boost.XInt formal review

Subject: Re: [boost] [xint] Boost.XInt formal review
From: Scott McMurray (me22.ca+boost_at_[hidden])
Date: 2011-03-15 15:18:03

On Tue, Mar 15, 2011 at 12:16, Scott McMurray <me22.ca+boost_at_[hidden]> wrote:
> On Tue, Mar 15, 2011 at 11:48, Thomas Klimpel
> <Thomas.Klimpel_at_[hidden]> wrote:
>>
>> [...] Two of these occurrences had the form "(m%std::ptrdiff_t(n)+n)%n", which surprised me a bit. When I replaced them with "modulo(m, n)", the compiler refused to compile it. So I wrote "modulo(m, std::ptrdiff_t(n))" instead, and this compiled fine. Thinking a bit about this, I realized that the template does the right thing by refusing to compile this, whereas the "(m%n+n)%n" construct can lead to surprising results in case n is of type std::size_t.
>>
>
> Perhaps the real thing it's telling you is that you want something like this:
>
>    [snip wrong code]
>
> Because if you really want modulo 4000000000, casting to ptr_diff_t
> will give very strange results on LP32...
>

Uh, that second enable_if_c should of course be disable_if_c:

   template <class T>
   enable_if_c<numeric_limits<T>::is_signed, T>::type
   modulo(T m, const T& n)
   {
       m %= n;
       if (m < 0) m += n;
       return m;
   }

   template <class T>
   disable_if_c<numeric_limits<T>::is_signed, T>::type
   modulo(T m, const T& n)
   {
       m %= n;
       return m;
   }

Oops,
~ Scott