新构造的对象作为默认模板函数参数

Question

Test::load should load data (described either by name or index) into passed reference. As third parameter there is a default value in case of loading fail (data are loaded from file).

struct Test
{
    template <typename ValueType, typename DefaultValueType>
    void load(const char * c, ValueType & a, DefaultValueType b)
    {
        std::cout << "1";
    }

    template <typename ValueType, typename DefaultValueType>
    void load(int i, ValueType & a, DefaultValueType b)
    {
        std::cout << "2";
    }
};

I need both ValueType and DefaultValueType because of my special need for strings and custom types. Problem occurs when I want to use default constructed type as default third argument (in const char * version) deduced from ValueType.

template <typename ValueType, typename DefaultValueType>
void load(const char * c, ValueType & a, DefaultValueType b = ValueType())
{
    std::cout << "1";
}

I get expects 3 arguments error (vs2010).

Why it doesn't work? I need to do workaroud to make this work.

---

Here is full "working" code if anyone is interested to play around

#include <iostream>

struct Test
{
    template <typename ValueType, typename DefaultValueType>
    void load(const char * c, ValueType & a, DefaultValueType b/* = ValueType()*/)
    {
        std::cout << "1";
    }

    template <typename ValueType, typename DefaultValueType>
    void load(int i, ValueType & a, DefaultValueType b)
    {
        std::cout << "2";
    }

    //Workaround
    template<typename ValueType>
    void load(const char * c, ValueType & v)
    {
        load(c,v,ValueType());
    }
};

int main()
{
   Test t;
   float f;

   //This is standard behavior
   t.load("bar", f, 0.0f);

   //I want this to be possible call
   t.load("bar", f);
}

Answer 1

You are looking for partial template specialization . However partial template specialization is possible only for classes and not for functions. For functions, you solve the issue through overloading, which is exactly what you have done in your workaround.

Here is a writeup on why template functions have overloading rather than specialization.