[boost] numeric_cast

Subject: [boost] numeric_cast
From: Brandon Kohn (blkohn_at_[hidden])
Date: 2011-06-16 09:44:06

Hello,

Recently I was creating a new numeric type which I wanted to use along 
with boost::numeric_cast. I read over the docs and examples and found 
that it seemed to support (at least partially) the notion of policy 
driven casting operations. I was unable to figure out how one might use 
a custom conversion or overflow policy along with the actual 
boost::numeric_cast<N>(x) function. In the end I created a workaround 
which seemed to work nicely:

template <typename ConversionTraits, typename EnableIf = void>
struct numeric_cast_traits
{
   typedef def_overflow_handler overflow_handler;
   typedef UseInternalRangeChecker range_checker;
   typedef Trunc
     <
         typename ConversionTraits::source_type
     > rounding_policy;
};

template <typename Target, typename Source>
inline Target numeric_cast( Source arg )
{
   typedef boost::numeric::conversion_traits<Target, Source> conv_traits;
   typedef numeric_cast_traits< conv_traits > cast_traits;
   typedef boost::numeric::converter
   <
       Target
     , Source
     , conv_traits
     , typename cast_traits::overflow_handler
     , typename cast_traits::rounding_policy
     , boost::numeric::raw_converter< conv_traits >
     , typename cast_traits::range_checker
   > converter;
   return converter()(arg);
}

This change allowed me to easily customize the numeric_cast function for 
individual conversion types.

Did I miss something in the official library? If not, would this change 
be something good to go in?

Regards,

Brandon Kohn