I'm interested in learning a little about template meta-programming. In the code below I'm attempting to find an unsigned integral type large enough to hold N bits, specified at compile time, using some template recursion.
template <typename T>
struct NextIntegralType
{
};
template <>
struct NextIntegralType<void>
{
typedef unsigned char type;
};
template <>
struct NextIntegralType<unsigned char>
{
typedef unsigned short type;
};
...More type 'iteration' here...
template<size_t BITS, typename T>
struct FindIntegralType2
{
typedef std::conditional<BITS <= sizeof(typename T::type)*8, T, FindIntegralType2<BITS, NextIntegralType<typename T::type>>> _type;
typedef typename _type::type type;
};
template<size_t BITS>
struct FindIntegralType
{
typedef typename FindIntegralType2<BITS, NextIntegralType<void>>::type type;
};
When I declare a variable and assign an integral value to it...
FindIntegralType<15>::type test(4000);
I get the following:
error: no matching function for call to ‘FindIntegralType2<15u, NextIntegralType<unsigned char> >::FindIntegralType2(int)’
note: candidates are:
note: constexpr FindIntegralType2<15u, NextIntegralType<unsigned char> >::FindIntegralType2()
note: candidate expects 0 arguments, 1 provided
note: constexpr FindIntegralType2<15u, NextIntegralType<unsigned char> >::FindIntegralType2(const FindIntegralType2<15u, NextIntegralType<unsigned char> >&)
note: no known conversion for argument 1 from ‘int’ to ‘const FindIntegralType2<15u, NextIntegralType<unsigned char> >&’
It seems like my recursion isn't 'unwinding'. Can anyone point me in the right direction?
Note: I'm using GCC 4.6.
EDIT:
I found a post that I missed before:
Automatically pick a variable type big enough to hold a specified number
Which points to an answer in boost (where they always are):
boost_integer
This should solve both my practical need and intellectual curiosity.