Re: [boost] [safe_numerics] review

Subject: Re: [boost] [safe_numerics] review
From: Peter Dimov (lists_at_[hidden])
Date: 2017-03-11 23:43:35

> - that safe_signed_range and safe_unsigned_range have a default policy of 
> native seems wrong to me.

In fact, the promotion policy can be eliminated entirely if the underlying 
type is always made automatic.

    safe<int> x( 5 ); // safe_base<INT_MIN, INT_MAX, throw_exception>
    safe<signed char> y( 5 ); // safe_base<SCHAR_MIN, SCHAR_MAX, 
throw_exception>

    auto z = x + y; // safe_base<INT_MIN+SCHAR_MIN, INT_MAX+SCHAR_MAX, 
throw_exception>

    safe<int> z = x + y; // converts above into safe<int>, checks

Since the only visible effect of the promotion policy is to determine the 
first template parameter of the result, if there's no such template 
parameter, there'd be no need to determine it.

Having both intmax_t and uintmax_t is a bit of a nuisance though, we still 
have to carry the 'signed' bit around.