“type* = nullptr”是什么意思

Question

I don't undestand.

template<class T>
T foo2(T t, typename std::enable_if<std::is_integral<T>::value >::type* = nullptr) 
{
    return t;
}

type* = 0 ? What is this.

Answer 1

This is a way of achieving SFINAE : The function can only be selected if all the types can be substituted in properly.

If std::is_integral<T>::value is false (ie, T is not an integral type), std::enable_if<...> will have no member type , thus a substitution failure happens and this function won't be called (and a different overload might).

If T is an integral type, then typename std::enable_if<std::is_integral<T>::value >::type will be void , so the second parameter will be of type void* . It is an unnamed parameter with default value nullptr , so you don't have to specify it.

So you would call it like this:

foo2(0);  // T is `int`, which is integral, so the function can be called
foo2(0, nullptr);  // Same as above, but explicitly passing the parameter

// Can't call the function, because `double` is not integral,
// so second type is a substitution failure
// foo2(0.0);

Note that this would normally be achieved either with a default template parameter:

// Same `void*` if integral, else substitution failure
template<class T, typename std::enable_if<std::is_integral<T>::value >::type* = nullptr>
T foo2(T t)
{
    return t;
}

// Or alternatively with an `int` if integral, else substitution failure
template<class T, typename std::enable_if<std::is_integral<T>::value, int>::type = 0>
T foo2(T t)
{
    return t;
}

Or directly in the return type:

template<class T>
// Returns `T` if it's integral, else substitution failure
typename std::enable_if<std::is_integral<T>::value, T>::type foo2(T t)
{
    return t;
}

And in C++20, you can use requires (or the std::integral concept in this case)

template<class T> requires std::is_integral_v<T>
T foo2(T t)
{
    return t;
}

template<std::integral T>
T foo2(T t)
{
    return t;
}

Answer 2

In C/C++ you can, in a declaration of a function, omit the name of (a) parameter(s). This is a preferred style sometimes when separating an interface from an implementation to avoid having confusion between the names of the parameters of, say, a function prototype & an implementation. I have never quite seen someone do this before. But what they are then doing as @IgorTandetnik pointed out, is initialize that "dummy parameter" with a default value.