I'm trying to replace a matrix library currently used in my code using Eigen. I've got several classes like this one adding custom methods to base matrix class. In this example I replaced the father class with Eigen:
#include <iostream>
#include <eigen3/Eigen/Dense>
class MyVectorType: public Eigen::Matrix<double, 3, 1> {
public:
MyVectorType(void) :
Eigen::Matrix<double, 3, 1>() {
}
typedef Eigen::Matrix<double, 3, 1> Base;
// This constructor allows you to construct MyVectorType from Eigen expressions
template<typename OtherDerived>
MyVectorType(const Eigen::MatrixBase<OtherDerived>& other) :
Eigen::Matrix<double, 3, 1>(other) {
}
// This method allows you to assign Eigen expressions to MyVectorType
template<typename OtherDerived>
MyVectorType & operator=(const Eigen::MatrixBase<OtherDerived>& other) {
this->Base::operator=(other);
return *this;
}
void customMethod() {
//bla bla....
}
};
The big problem is that is not really easy to manage custom classes in methods. Example:
void foo(MyVectorType& a) {
....
a.customMethod();
}
void foo(Eigen::Ref<MyVectorType::Base> a) {
....
a.customMethod(); <---can't call customMethod here
}
Eigen::Matrix<double, -1, -1, 0, 15, 15> m(3,1);
foo(m); <---can't call it with m;
Eigen::Map<Matrix<double, 3, 1> > map(m.data(), 3, 1);
Eigen::Ref<Matrix<double, 3, 1> > ref(map);
foo(ref); <---it works but I can't call custom method
Usually Eigen provides Ref template class but I cannot use it with custom classes because if I use Ref, I won't be able to call customMethod inside foo in this example, I should Eigen::Ref in my example. Avoiding the use of Ref is a big problem because using Map and Ref Eigen objects is quite important to cast dynamic matrix to fixed one and perform other cast operations. Final question: what's the best strategy to use Eigen in this case?
