Best practice to implement this template class

[ar115]

I have a class which currently contains N vectors of int
and M vectors of double like below;

template< int N, int M >
class VectorCollection {
std::vector<int> ivec[N];
std::vector<double> dvec[M];
};

probably in future vectors of other types (e.g. bool) will be added
to class. I want to have a simple interface which give access to
interior vectors based on the types of int,double,...
like below

template< int N, int M >
class VectorCollection {
// vectors as above
public:
	template< class T, int K >
	T& data(i);
};

data<int,0>(i) should return ivec[0][i];
data<double,1>(i) should return dvec[1][i] and so

But I do not know how to implement this. It seems very difficult.
Could anyone help me on this?

Thanks a lot

[Regel]

well, you can't do a template partial specialization on a member function, but the question is why would you want to do so? why can't you make two data functions - one for vector of doubles and the other for vectors of integers? that's simpler and there is no code bloat.

[ar115]

Regarding to my needs I am sure that the above interface is good for me.
And it is hard for me ( and maybe boring for you ) to tell why I need that
I need just that to work, is there any good solution?
Thanks

[Regel]

Regarding to my needs I am sure that the above interface is good for me.
And it is hard for me ( and maybe boring for you ) to tell why I need that
I need just that to work, is there any good solution?
Thanks

It's not so difficult to implement, but I think you should consider redesign your application because this can easily lead to code bloat, since for every different combination of <T, K> you actually create a new function(that is the way templates work as far as I know)!

anyway, if that's really what you want you can use function overloading passing a dummy paramter:

template< int N, int M >
class VectorCollection {
private:
	std::vector<int> ivec[N];
	std::vector<double> dvec[M];
public:
	double& data(int i, int k, double dummy)
	{
		return dvec[k][i];
	}

	int& data(int i, int k, int dummy)
	{
		return ivec[k][i];
	}

	template<class T, int K>
	T& data(int i)
	{
		return data(i, K, T());
	}
};

[ar115]

Many Thanks! it is a good solution
But is there better ones?
I was wondering if we can use a nested class and delegate things to it. since we can
use partial specialization for classes, something like this

template< int N, int M >
class VectorCollection {
    std::vector<int> ivec[N];
    std::vector<double> dvec[M];

   template< class T, int N >
   struct nested{
       // implementation stuff
    };
public:
    template< class T, int N >
    T& data( int i ){
          return nested<T,n>::getdata(i);
    }

};

Then the question is that what the "// implementation stuff" of struct nested must be?
Thanks

[Regel]

Many Thanks! it is a good solution
But is there better ones?
I was wondering if we can use a nested class and delegate things to it. since we can
use partial specialization for classes, something like this

template< int N, int M >
class VectorCollection {
    std::vector<int> ivec[N];
    std::vector<double> dvec[M];

   template< class T, int N >
   struct nested{
       // implementation stuff
    };
public:
    template< class T, int N >
    T& data( int i ){
          return nested<T,n>::getdata(i);
    }

};

Then the question is that what the "// implementation stuff" of struct nested must be?
Thanks

Yes you can do this:

   template< class T, int N >
   struct nested{ // that's for double
       T& get() { .. }
    };
   template<int N > // thats for int
   struct nested<int, N>{
       int& get() { .. }
    };