I need to perform calculations for each matrix entry in an Rcpp::List. The list contains an arbitrary combination of dense and sparse matrices. This list is passed to C++ from R. My question here concerns both approaches to implementation, and efficiency. My current thought is to loop over each element of the list, check for its type (dense or sparse) and generate an arma::mat or arma::sp_mat for the element, and use it for calculations, preferably without copying. I am confused about the following:

1. How to check whether the type of matrix is dense or sparse from Rcpp::List[d] where d is an index ?
2. I am able to use the advanced constructor for arma::mat via NumericMatrix that avoids copying (as below), but I am not sure how to go about doing the same with arma::sp_mat. These matrices are large, and I want to avoid copying.

Rcpp::NumericMatrix Ad = RcppListObject[d];
arma::mat ad=arma::mat( Ad.begin(), Ad.nrow(), Ad.ncol(), false );
//any needed calculations based on the d^{th} entry uses `ad`.

3. Closely related to 2, there are a variety of calculations I will need to do with the d^{th} matrix in the List, at different points of the algorithm. Do I have to do the two steps noted in point 2 every time before I can do computations with it, or is there a more efficient route? ( For example, can arma::field help, and handle a combination of arma::mat & arma::sp_mat matrices without copying? )

Thank you very much!