specialize a template class constructor

Question

I want to specialize a template class constructor:

If type is int default value is 50 and -50. and if it's float default should be 0.5 and -0.5 .

My code is :

#include <iostream>
#include <limits>
#include <type_traits>

template<typename T>
class Foo{
public:
    template<typename = typename std::enable_if<
        std::is_integral<T>::value&& !std::is_floating_point<T>::value>::type>
        Foo(T value1 = 50, T value2 = -50) :value1_(value1), value2_(value2){}

    template<typename = typename std::enable_if<
        std::is_floating_point<T>::value>::type>
        Foo(T value1 = 0.5, T value2 = -0.5, void* dummy = 0) : value1_(value1), value2_(value2){}
    T value1_, value2_;
};

int main()
{
    Foo<float> test;
    std::cout << test.value1_ << " " << test.value2_ << '\n';

    Foo<int> test2;
    std::cout << test2.value1_ << " " << test2.value2_;
}

It works just fine in visual studio 2013 .

But gcc 4.9.2 rejects it :

main.cpp: In instantiation of 'class Foo<float>':
main.cpp:29:13:   required from here
main.cpp:19:3: error: no type named 'type' in 'struct std::enable_if<false, void>'
   Foo(T value1 = 50, T value2 = -50) :value1_(value1), value2_(value2){}
   ^
main.cpp: In instantiation of 'class Foo<int>':
main.cpp:32:11:   required from here
main.cpp:23:3: error: no type named 'type' in 'struct std::enable_if<false, void>'
   Foo(T value1 = 0.5, T value2 = -0.5, void* dummy = 0) : value1_(value1), value2_(value2){}
   ^

Is my code wrong? if so why visual studio compile it ? or maybe it's a gcc bug ?!

Answer 1

Your code is incorrect. The top-level T cannot be used in a SFINAE context for its methods, which is what you are trying to do. Only a substitution that takes place in an immediate context may result in a deduction failure (§14.8.2/8):

Only invalid types and expressions in the immediate context of the function type and its template parameter types can result in a deduction failure. [ Note: The evaluation of the substituted types and expressions can result in side effects such as the instantiation of class template specializations and/or function template specializations, the generation of implicitly-defined functions, etc. Such side effects are not in the “immediate context” and can result in the program being ill-formed. — end note ]

GCC and Clang are correct in rejecting your code.

A workaround is just to introduce a dummy template type that defaults to the top-level T, and SFINAE on that one. Like so:

template <typename T_ = T, // now the subsequent line *is* in
                           // an immediate context
          typename = typename std::enable_if <
              std::is_integral<T_>::value&& !std::is_floating_point<T_>::value
          >::type>
Foo(T value1 = 50, T value2 = -50) 
:value1_(value1), value2_(value2) { }

Note that is_integral and is_floating_point are mutually exclusive, you should only have to check one or the other.

In this example, it'd probably be a lot simpler just to farm out the default values to another struct, so that you can have just one constructor that looks like:

Foo(T value1 = FooDefaults<T>::value1, T value2 = FooDefaults<T>::value2)
: value1_(value1), value2_(value2)
{ }