Question

Consider this class template:

template <typename T1, typename T2, bool B>
class SomeClass { };

Now, I’d like to provide two implementations based on B==true and B==false. That is, I’d like to say something like:

template <ANYTHING, ANYTHING, true> class SomeClass {
// First implementation
};

template <ANYTHING, ANYTHING, false> class SomeClass {
// Second implementation
};

How can this be done in C++(11)?

Answer 1

With partial specialization:

// primary
template<typename X, typename Bool>
struct Foo;

template<typename X>
struct Foo<X, std::true_type> {};

template<typename X>
struct Foo<X, std::false_type> {};

// use
Foo<X, std::true_type> x;

I use a type-wrapper for bool, but you can also do that with
non-type template parameters:

// primary
template<typename, bool>
struct Foo;

template<typename X>
struct Foo<X, true> {};

template<typename X>
struct Foo<X, false> {};

// use
Foo<X, true> x;

Sometimes you can compute the value used for partial specialization
with meta-programming in the default argument:

// primary
template<typename X, typename is_integral_ = std::is_integral<X>::type>
struct Foo;

This makes the configuration variable overridable by user choice.

struct my {};
Foo<my, std::true_type> x;

To prevent that, dispatch through inheritance:

// primary, where Foo_impl is any of the above
template<typename X>
struct Foo : public Foo_impl<X> {};