Thank you for your quick response. I was only thinking about the
element-wise matrix/matrix equivalents that already have vector/vector
implementations. My first look was telling me that this shouldn't be
difficult. That said reading and using templated C++ code is a different
story than modifying it!
Jason
On Sat, Jan 20, 2018 at 5:53 PM, Dirk Eddelbuettel <e...@debian.org> wrote:
>
> On 20 January 2018 at 17:45, Jason Nielsen wrote:
> | Hi all,
> |
> | I was wondering if there is a technical reason that element-wise matrix
> | operations between Rcpp's Matrix type aren't implemented. There is
> | vector/vector, vector/scalar and matrix/scalar but no matrix/matrix. The
> | following naive code:
> |
> | #include <Rcpp.h>
> | using namespace Rcpp;
> |
> | NumericMatrix operator+(NumericMatrix A, NumericMatrix B) {
> | if ((A.nrow() != B.nrow()) | (A.ncol() != B.ncol()))
> | stop("Dimensions of the two matrices do not conform!");
> | NumericVector vA = as<NumericVector>(A);
> | NumericVector vB = as<NumericVector>(B);
> | NumericVector vC = vA + vB;
> | NumericMatrix C = NumericMatrix(A.nrow(), A.ncol(), vC.begin());
> | return C;
> | }
> |
> | // [[Rcpp::export]]
> | NumericMatrix MatAdd(const NumericMatrix A, const NumericMatrix B) {
> | return A+B;
> | }
> |
> | works. From my brief fiddling it seems there is no copying going on
> above
> | so I don't think there is too much extra overhead in the abstraction. My
> | guess is that it just hasn't been a priority and "patches happily
> accepted"
> | is the answer but I'm curious if there is some other less obvious reason
> as
> | it seems to me all the heavy lifting was done to implement vector/vector.
>
> R code:
>
> sample(c("History", "Lazyness"), 1) ## just kidding. or maybe not.
>
> Years back I usually pointed at RcppArmadillo if you wanted "math with
> matrices" as we just didn't have all the operators. Now, we have been
> filling
> them in, so this operator+() makes some sense. Doing them "well" in the
> template code is a bit more involved but mayne you want to poke around?
>
> I am sure some of the regulars here would be happy to help you along.
>
> Dirk
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
>
_______________________________________________
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
