Generalizing input arguments for Eigen-types-CodePudding

The template guide provided by Eigen recommends using Eigen::MatrixBase to implicitly deduce template parameters to functions. In this example I want to pass an expression:

#include <iostream>
#include "Eigen/Dense"

template <typename D>
void eval(Eigen::MatrixBase<D>& in, Eigen::MatrixBase<D>& out)
{
    out.array() = in.array()   1;
}

int main()
{
    Eigen::MatrixXd A(2,2);
    A(0,0) = 2;
    A(1,1) = 1;
    Eigen::VectorXd B(2);
    B(1) = 1;
    std::cout << A << ", " << B << std::endl;

    Eigen::VectorXd res(2);

    eval(A*B, res);
    std::cout << res << std::endl;
}

Which outputs an error:

EigenTest.cpp: In function ‘int main()’:
EigenTest.cpp:21:18: error: no matching function for call to ‘eval(const Eigen::Product<Eigen::Matrix<double, -1, -1>, Eigen::Matrix<double, -1, 1>, 0>, Eigen::VectorXd&)’
   21 |     eval(A*B, res);

Is there a more general form of Eigen::MatrixBase that extends to accepting expressions?

CodePudding user response：

You have to use two template parameters and make the input reference refer to a const type. The issue is that the product operator will return a const.

Like this

template <typename U, typename V>
void eval(const Eigen::MatrixBase<U>& in, Eigen::MatrixBase<V>& out)
{
    out.array() = in.array()   1;
}