模板类，功能专业化

Question

I want to have a template class that looks something like what I have down below. Then, I want a function in it with a template specialization depending on a CLASS template parameter. How do I make this work? I realize the code I provided is wrong on many levels, but it's just to illustrate the concept.

template <typename _T, size_t num>
class Foo
{
    // If num == 1, I want to call this function...
    void Func<_T, 1>()
    {
        printf("Hi!");
    }

    // Otherwise, I want to call this version.
    void Func<_T, num>()
    {
        printf("Hello world!");
    }
};

Answer 1

struct Otherwise { };
template<size_t> struct C : Otherwise { };

// don't use _Uppercase - those names are reserved for the implementation
// (i removed the '_' char)
template <typename T, size_t num>
class Foo
{
public:
    void Func() { Func(C<num>()); }

private:
    // If num == 1, I want to call this function...
    void Func(C<1>)
    {
        printf("Hi 1!");
    }

    // If num == 2, I want to call this function...
    void Func(C<2>)
    {
        printf("Hi 2!");
    }

    // Otherwise, I want to call this version.
    void Func(Otherwise)
    {
        printf("Hello world!");
    }

    //// alternative otherwise solution:
    // template<size_t I>
    // void Func(C<I>) { .. }
};

Answer 2

There are no partial specialisations of function templates, and to partially specialise a member you need to first partially specialise the class template.

template< typename _T, size_t num >
struct Foo {
    void Func() {
        printf("Hello world!");
    }
};

template< typename _T >
struct Foo< _T, 1 > {
    void Func() {
        printf("Hi!");
    }
};

Now, if Foo also contains methods other than Func whose implementation is independent of the value for num , and you don't want to duplicate their implementation in the specialization of Foo , you can apply the following pattern:

template< typename _T, size_t num >
struct FooFuncBase {
    void Func() {
        printf("Hello world!");
    }
};

template< typename _T >
struct FooFuncBase< _T, 1 > {
    void Func() {
        printf("Hi!");
    }
};

template< typename _T, size_t num >
struct Foo : public FooFuncBase< _T, num > {
  void OtherFuncWhoseImplementationDoesNotDependOnNum() {
    ...
  }
};

Or, using CRTP :

template< typename _Derived, typename _T, size_t num >
struct FooFuncBase {
    void Func() {
        static_cast< _Derived* >(this)->OtherFuncWhoseImplementationDoesNotDependOnNum();
        printf("Hello world!");
    }
};

template< typename _Derived, typename _T >
struct FooFuncBase< _Derived, _T, 1 > {
    void Func() {
        static_cast< _Derived* >(this)->OtherFuncWhoseImplementationDoesNotDependOnNum();
        printf("Hi!");
    }
};

template< typename _T, size_t num >
struct Foo : public FooFuncBase< Foo< _T, num >, _T, num > {
  void OtherFuncWhoseImplementationDoesNotDependOnNum() {
    printf("Other");
  }
};

Answer 3

This is called partial template specialisation. It looks like this:

template<typename _T, size_t num>
class FooBase
{
};

template <typename _T, size_t num>
class Foo : public FooBase<_T,num>
{
    void Func()
    {
        printf("Hello world!");
    }
};


template <typename _T>
class Foo<_T,1> : public FooBase<_T,num>
{
    void Func()
    {
        printf("Hi!");
    }
}