Extract every other row or column of an Eigen matrix as a new matrix

Question

Is there any simple and efficient way to extract the odd/even numbered rows or columns of the below matrix:

int m=999,n = 1000;
MatrixXd mat(m,n);

and extract them into a new matrix?

One possible way is to use an for loop and put the desired rows/ columns into the corresponding rows/columns of the new matrix. But is there any simpler and more efficient way to do so?

Answer 1

No more efficient solution, but for the columns, since you have a column-major matrix with an even number of column, you can reshape the data like such that the even/odd columns forms blocks:

MatrixXd even_cols = MatrixXd::Map(mat.data(), 2*999, 500).topRows(999);
MatrixXd odd_cols  = MatrixXd::Map(mat.data(), 2*999, 500).bottomRows(999);

Another more general approach is to play with strides:

MatrixXd even_cols = MatrixXd::Map(mat.data(),     999, 500, OuterStride<>(2*999));
MatrixXd odd_cols  = MatrixXd::Map(mat.data()+999, 999, 500, OuterStride<>(2*999));

This also work for the even/odd rows with a column-major matrix. In this case we need to define an inner stride of 2:

MatrixXd even_rows = MatrixXd::Map(mat.data(),   500, 1000, Strides<Dynamic,2>(999,2));
MatrixXd odd_rows  = MatrixXd::Map(mat.data()+1, 499, 1000, Strides<Dynamic,2>(999,2));