Restriction of access to function [duplicate]

Question

This question already has answers here:

std::enable_if to conditionally compile a member function (8 answers)

Closed 8 years ago.

I have a generic class with a function that I want to restrict to instances of floating point types only, at compile time. As shown in the example below:

template <typename T>
class ClassName
{
    // instance variables, etc..

    void some_method()
    {
        // do stuff, but only for floating point types
    }
}

How do I make the compiler reject the usage of some_method for ClassName of non-floating point types?

I have been looking at SFINAE but I simply can't get it to work, so after several hours of failing I'm asking for your help.

Thanks :)

Answer 1

You can use a combination of std::is_floating_point and std::enable_if to only enable the function for floating point types:

#include <type_traits>

template <typename T>
class ClassName
{
    // instance variables, etc..
 public:
  template<typename T2 = T,
           typename = typename std::enable_if< std::is_floating_point<T2>::value >::type>
  void some_method()
  { 
    // do stuff, but only for floating point types
  } 
};

int main()
{
  ClassName<double> d; // OK
  d.some_method();     // OK
  ClassName<int> i;    // OK
  i.some_method();     // ERROR
}

Answer 2

Use static_assert, if your compiler support c++11

void some_method()
{
    static_assert( std::is_floating_point<T>::value, "Only for floating points" );
    // do stuff, but only for floating point types
}

Then there will be compiler error if you try call this method for non floating points parameter.

And for not-floating point:

static_assert( !std::is_floating_point<T>::value, "and only for non-floating point" );

Answer 3

Something like this:

template< typename Tdummy = T, typename = typename std::enable_if< std::is_floating_point< Tdummy >::value >::type >
 void some_method()
{
}

EDIT to elaborate. This will result in the following. Compiler will generate some_method() only for ClassName with floating-point template parameter. It will not be generated for non-floating types and will result in a compile-time error.

#include <type_traits>

template <typename T>
class ClassName
{
    // instance variables, etc..
    template< typename Tdummy = T, typename = typename std::enable_if< std::is_floating_point< Tdummy >::value >::type >
    void some_method()
    {
        // do stuff, but only for floating point types
    }

 void some_general_method
  {
   // general stuff for all types
  }
};

int main()
{
 ClassName< float > bar;
 ClassName< int > foo;

 bar.some_general_method(); // OK
 foo.some_general_method(); // OK

 bar.some_method(); // OK
 foo.some_method(); // Compile-time ERROR 

 return( 0 );
}

Answer 4

void some_method(){

    if (std::is_floating_point<T>::value)
    {
        // do stuff, but only for floating point types
    }
    else
    {
        return;
    }
}

I tried with boost::is_floating_point too :-

#include <boost/utility/enable_if.hpp>
#include <boost/type_traits/is_floating_point.hpp>

template <typename T>
class ClassName {
typename boost::enable_if<boost::is_floating_point<T> >::type
some_method(const T & t)
{

}
};

int main()
{
ClassName<float> p; //Compiles

/* Following throws error, 
error: no type named 'type' in 
    'struct boost::enable_if<boost::is_floating_point<int>, void>'
ClassName<int> q;
*/
}

Answer 5

As detailed in this answer, you need the member function to be a template for SFINAE to work (Live example at Coliru):

template <typename T>
class ClassName
{
    // instance variables, etc..

public:
    template <typename = typename std::enable_if<std::is_floating_point<T>::value>::type>
    void some_method()
    {
        // do stuff, but only for floating point types
    }
};

Answer 6

Update per R. Martinho Fernandes comment

#include <type_traits>

template <typename T>
struct ClassName
{
    // instance variables, etc..

    template<typename R = T>
    void some_method()
    {
        static_assert(std::is_floating_point<R>::value,
            "ClassName<T>::some_method is implemented only for floating "
                "point T");
        // do stuff, but only for floating point types
    }
};

int main()
{   
    ClassName<float> f;
    f.some_method();
    ClassName<int> i;
    i.some_method(); // <-- static_asserts here
    return 0;
}