next_permutation and cartesian product for ranges?

[Timon Gehr]

On 10/09/2012 11:23 PM, bearophile wrote:
> Andrei Alexandrescu:
>
>> Yes, we need next_permutation. It will be up to you to convince the
>> Grand Inquisition Committee of Reviewers that next_even_permutation is
>> necessary and/or sufficient.
>
> I don't like the design of C++ STL here. Instead of a
> next_permutation(), what about ranges that yield all permutations, all
> adjacent permutations, all combinations?
>
> [snip code]

I agree that that's better conceptually, but your implementations have
overflow issues. (And it is not too obvious how to fix that, given the
existing range design.)