Rcpp: transform matrix to vector

Question

Generally I want to transform a 2-D matrix from Rcpp to a vector in R, it should be very simple by using "as(m)", however, I still get a matrix from R, I wonder why? SHould I remove the attr manually in Rcpp?

#include <Rcpp.h>
#include <string>
using namespace Rcpp;
// [[Rcpp::export]]
NumericVector matrix2vector(NumericMatrix m, const bool byrow=false){
  NumericVector x;
  if(byrow){
    Rcout<< "warning: default by column\n";
    m = transpose(m);
    x = as<NumericVector>(m);
  }else{
    x = as<NumericVector>(m);
  }
  return(x);
}

/*** R
m=matrix(1:15,5,3)
matrix2vector(m,byrow=T)
*/

Answer 1

Thanks to @user2957945 and @Dirk Eddelbuettel.

What I want is as below:

#include <Rcpp.h>
#include <string>
using namespace Rcpp;
// [[Rcpp::export]]
NumericVector matrix2vector(NumericMatrix m, const bool byrow=false){
  NumericVector x;
  if(byrow){
    Rcout<< "warning: default by column\n";
    m = transpose(m);
  }
  NumericVector x(m);
  x.attr("dim") = R_NilValue;
  return(x);
}

/*** R
m=matrix(1:15,5,3)
matrix2vector(m,byrow=T)
# [1]  1  6 11  2  7 12  3  8 13  4  9 14  5 10 15
*/

Answer 2

I am not sure I understand the question, or the attempted answer. You are still returning matrices here. And Rcpp has limited support for reshaping and more advanced Matrix operations.

A slighly simplified version of your code:

Code

#include <Rcpp.h>
using namespace Rcpp;
// [[Rcpp::export]]
NumericVector matrix2vector(NumericMatrix m, const bool byrow=false){
  if (byrow){
    Rcout << "warning: default by column\n";
    m = transpose(m);
  }
  return NumericVector(m);
}

/*** R
m <- matrix(1:15,5,3)
print(matrix2vector(m, byrow = TRUE))
print(matrix2vector(m, byrow = FALSE))
*/

Output

R> Rcpp::sourceCpp("~/git/stackoverflow/61036707/question.cpp")

R> m <- matrix(1:15,5,3)

R> print(matrix2vector(m, byrow = TRUE))
warning: default by column
     [,1] [,2] [,3] [,4] [,5]
[1,]    1    2    3    4    5
[2,]    6    7    8    9   10
[3,]   11   12   13   14   15

R> print(matrix2vector(m, byrow = FALSE))
     [,1] [,2] [,3]
[1,]    1    6   11
[2,]    2    7   12
[3,]    3    8   13
[4,]    4    9   14
[5,]    5   10   15
R> 

All you are executing here, really, is a transpose.

I would try RcppArmadillo which does have explicit rowvec and colvec types, with the latter being aliased to vec. I would have to check what it suggest in terms of reshaping "rows by columns" elements into a rows by columns vector.

Edit: Indeed (as so often) easier in RcppArmadillo.

Code

#include <RcppArmadillo.h>

// [[Rcpp::depends(RcppArmadillo)]]

// [[Rcpp::export]]
arma::mat matrix2vector(arma::mat m, const bool byrow=false){
  if (byrow) {
    return m.as_row();
  } else {
    return m.as_col();
  }
}

/*** R
m <- matrix(1:15,5,3)
print(matrix2vector(m, byrow = TRUE))
print(matrix2vector(m, byrow = FALSE))
*/

Output

R> Rcpp::sourceCpp("~/git/stackoverflow/61036707/answer.cpp")

R> m <- matrix(1:15,5,3)

R> print(matrix2vector(m, byrow = TRUE))
     [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13] [,14] [,15]
[1,]    1    6   11    2    7   12    3    8   13     4     9    14     5    10    15

R> print(matrix2vector(m, byrow = FALSE))
      [,1]
 [1,]    1
 [2,]    2
 [3,]    3
 [4,]    4
 [5,]    5
 [6,]    6
 [7,]    7
 [8,]    8
 [9,]    9
[10,]   10
[11,]   11
[12,]   12
[13,]   13
[14,]   14
[15,]   15
R> 

Note that I am using arma::mat as the return type to not conflict with either colvec or rowvec. You could pick either but then you could not return the answer.

Edit 2: And per the comment below: if you really just want to reshape content you already have there is no need for Rcpp. Just nuke the dim attribute. Works at the R (and C++ if you want) level:

R> m
     [,1] [,2] [,3]
[1,]    1    6   11
[2,]    2    7   12
[3,]    3    8   13
[4,]    4    9   14
[5,]    5   10   15
R> dim(m)
[1] 5 3
R> dim(m) <- NULL
R> m
 [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
R>