Template specialization for multiple types

Question

Title is a little ambiguous.

Lets say I have a template defined as:

template < typename T >
void foo ( int x ) ;
template <>
void foo<char> ( int x ) ;
template <>
void foo<unsigned char> ( int x ) ;
template <>
void foo<short> ( int x ) ;
...

Internally both foo<signed>() and foo<unsigned>() do exactly the same thing. The only requirement is that T be an 8bit type.

I could do this by creating another template to type define a standard type based on size.

template < typename T, size_t N = sizeof( T ) > struct remap ;
template < typename T, size_t > struct remap< 1 >
{
    typedef unsigned char value;
}
...

Note, function templates cannot have default parameters. This solution only relocates the problem to another template and also introduces a problem if somebody tried passing a struct type as a parameter.

What is the most elegant way to solve this without repeating those function declarations?

This is not a C++11 question.

Answer 1

One possibility is to specialize a class template for multiple types at once:

// from: http://en.cppreference.com/w/cpp/types/enable_if
    template<bool B, class T = void>
    struct enable_if {};

    template<class T>
    struct enable_if<true, T> { typedef T type; };

template < typename A, typename B >
struct is_same
{
    static const bool value = false;
};
template < typename A >
struct is_same<A, A>
{
    static const bool value = true;
};


template < typename T, typename dummy = T >
struct remap;

template < typename T >
struct remap
<
    T,
    typename enable_if<    is_same<T, unsigned char>::value
                        || is_same<T, signed char>::value, T >::type
>
{
    void foo(int);
};


int main()
{
    remap<signed char> s;
    s.foo(42);
}

---

Another possibility is to specialize a class template for categories of types (type traits):

#include <cstddef>

template < typename T >
struct is_integer
{
    static const bool value = false;
};
template<> struct is_integer<signed char> { static const bool value = true; };
template<> struct is_integer<unsigned char> { static const bool value = true; };


template < typename T, typename dummy = T, std::size_t S = sizeof(T) >
struct remap;

template < typename T >
struct remap
<
    T
    , typename enable_if<is_integer<T>::value, T>::type
    , 1 // assuming your byte has 8 bits
>
{
    void foo(int);
};


int main()
{
    remap<signed char> s;
    s.foo(42);
}

Answer 2

You need your remap trait to simply map from input types to output types, and have your foo<T>(int) interface function delegate to a foo_implementation<remap<T>::type>(int) implementation. i.e.:

template <typename T>
struct remap {
    // Default: Output type is the same as input type.
    typedef T type;
};

template <>
struct remap<char> {
    typedef unsigned char type;
};

template <>
struct remap<signed char> {
    typedef unsigned char type;
};

template <typename T>
void foo_impl(int x);

template <>
void foo_impl<unsigned char>(int x) {
    std::cout << "foo_impl<unsigned char>(" << x << ") called\n";
}

template <typename T>
void foo(int x) {
    foo_impl<typename remap<T>::type>(x);
}

See it live at ideone.com.

That said, it might be realistically simpler to define foo_char, foo_int and foo_short and just call the correct one from client code. foo<X>() isn't syntactically much different from foo_X().