get return type of function

[dshawul]

Hi,
Say I have overloaded functions with different arguments AND return types. Each set of argument has only one corresponding return type.

Vector grad(Scalar)
Tensor grad(Vector)

Later I have:

template <class T>
void test(T x) {
    ...  Y = grad(x) 
}

Then how do I automatically declare the type of Y. Do I need to make the grad function a template and specialize each of them ?
Thanks for any help

[superbonzo]

in c++11, you can simply write

"auto Y = grad(x);"
"auto const& Y = grad(x);"
...

or

"typedef decltype(grad(x)) GradType;

GradType Y = grad(x);"

etc... depending on what's the expected type of Y.

with c++2003 compilers, things get more complex ...

[dshawul]

Thanks a lot.
Unfortunately I am not using c++11 so I can not use those solutions.
I get 'error C3861: 'decltype': identifier not found' for the second solution.
Do you have the solution that works for in Visual studio 2008?
I appreciate your help.

[superbonzo]

the easiest way is to write a trait class

template <class T>
struct grad_type {};

template <>
struct grad_type<Scalar> { typedef Vector type; };

template <>
struct grad_type<Vector> { typedef Tensor type; };

template <class T>
void test(T x) {
    typename grad_type<Y>::type  Y = grad(x)
}

of course, you need to specialize the trait for each possible parameter type. It would be easier if Scalar,Vector,Tensor had some high-level info in them ( like a rank grading ), in this way the trait could compute the grad_type automatically at compile time.

Anyway, I think emulating "auto"/"decltype" is possible in this case via some template machinery like SFINAE, but I wouldn't advise it if you don't know what you're doing ...

moreover, another simple solution could be to refactor the code, if possible; something like

template <class T,class G>
void test_impl(T& x,G y) {
	// ...
}

template <class T>
void test(T x) {
    test_impl( x, grad(x) );
}

[dshawul]

You are a life saver. I will use the traits method for now.
Scalar,vector and tensor are 1st,2nd and 3rd rank tensors but they are of completely different types represented by
double, double[], and double[][]. So I don't think I can automatically compute them. But anyway it is important to know
the higher or lower ranks in many places of the code so I will use that.
Thanks again.