Determining the “optimal” common numeric type in a template parameter pack

Question

What is the best way to determine a common numeric type in a template parameter pack with:

1. the smallest size,
2. no loss of precision, a

Answer 1

I am a little bit late to the party, here is my solution without Boost:

#include 
#include 
  
template struct mk_signed { typedef I       type; };
template<>   struct mk_signed   { typedef int16_t type; };
template<>   struct mk_signed   { typedef int32_t type; };
template<>   struct mk_signed   { typedef int64_t type; };
template<>   struct mk_signed   { typedef int64_t type; }; 
  
template  struct best_common_numeric_type;
template      struct best_common_numeric_type { typedef T type; };
  
template 
struct best_common_numeric_type {
   typedef typename best_common_numeric_type::type TS;     
   typedef typename std::conditional < (sizeof(T) > sizeof(TS)), T, TS>::type  bigger_integral;
   constexpr static bool fp = std::is_floating_point::value || std::is_floating_point::value;
   constexpr static bool have_signed = !fp && (std::is_signed::value || std::is_signed::value);
  
   typedef typename std::conditional <
     fp,
     typename std::common_type::type,
     typename mk_signed::type
   >::type type;
};