I have the following code:
// Case #1
float f = 1.0f;
float f2 = sqrt(f * pi);
// Case #2
double d = 1.0;
double d2 = sqrt(d * pi);
Is there any way to define the variable pi so that operator* and sqrt will operate on floats in Case #1, but will operate on doubles in Case #2?
Perhaps it might be possible with C++14 variable templates?
Sort of. You can certainly define such a pi:
template <class T> constexpr double pi = 3.14159...;
template <> constexpr long double pi<long double> = 3.14159...L;
template <> constexpr float pi<float> = 3.14159...F;
But you have to specify which pi you want:
float f2 = sqrt(f * pi<float>);
double d2 = sqrt(d * pi<double>);
More directly, you could define some pi object that just has overloaded operator* depending on the type:
struct Pi {
template <class T>
decltype(pi<T>) operator*(T val) { return val * pi<T>; }
template <class T>
friend decltype(pi<T>) operator*(T val, Pi) { return pi<T> * val; }
};
That lets you get the syntax you want, but that's weird, don't do it.
double value instead of int. template <class T> constexpr double pi = 3.0;