best way to convert DataFrame to Matrix in RCpp

Question

I have an RCpp code, where in a part of code I am trying to convert a DataFrame to a Matrix. The DataFrame only has numbers (no strings or dates).

The following code works:

//[[Rcpp::export]]
NumericMatrix testDFtoNM1(DataFrame x) {
  int nRows=x.nrows();  
  NumericMatrix y(nRows,x.size());
  for (int i=0; i<x.size();i++) {
    y(_,i)=NumericVector(x[i]);
  }  
  return y;
}

I was wondering if there is alternative way (i.e. equivalent of as.matrix in R) in RCpp to do the same, something similar to the following code below (which does NOT work):

//[[Rcpp::export]]
NumericMatrix testDFtoNM(DataFrame x) {
  NumericMatrix y(x);  
  return y;
}

* EDIT *

Thanks for the answers. As Dirk suggested, the C++ code is around 24x faster than either of the two answers and Function version is 2% faster than the internal::convert_using_rfunction version.

I was originally looking for an answer within RCpp without calling R. Should have made that clear when I posted my question.

I'd go with the solution you have in your question. We may need a better way to get `n` as well as `k` from the data.frame (ie as length of names vector). Otherwise it is a good solution in my mind which I like as well as some of the suggested answers. — Dirk Eddelbuettel, Jun 22 '14 at 15:03
And I was of course wrong about n and k retrievers as we do have `.size()`, I was mainly thinking of functions with similar names to those for matrices. May add that at some point... — Dirk Eddelbuettel, Jun 22 '14 at 18:18

score 9 · Accepted Answer · answered Jun 22 '14 at 14:55

9

Similar to Gabor's version, you can do something like this:

#include <Rcpp.h>
using namespace Rcpp ;

//[[Rcpp::export]]
NumericMatrix testDFtoNM(DataFrame x) {
  NumericMatrix y = internal::convert_using_rfunction(x, "as.matrix");  
  return y;
}

answered Jun 22 '14 at 14:55

Romain Francois

17,432
3
51
77

G. Grothendieck · Answer 2 · 2014-06-22T15:05:57.033

7

If you don't mind calling back to R it can be done compactly like this:

#include <Rcpp.h>
using namespace Rcpp;

// [[Rcpp::export]]
NumericMatrix DF2mat(DataFrame x) {
    Function asMatrix("as.matrix");
    return asMatrix(x);
}

UPDATE Incorporated Romain's comment.

edited Jun 22 '14 at 15:05

answered Jun 22 '14 at 14:54

G. Grothendieck

254,981
17
203
341

I would not trust to `static` a `Function` though. – Romain Francois Jun 22 '14 at 15:00
Calling to R has cost. I'd do it in C++. – Dirk Eddelbuettel Jun 22 '14 at 15:02
1

@Dirk, Yes, if this were time critical but its not necessarily true that every part of a C++ program must be fast. – G. Grothendieck Jun 22 '14 at 15:04
@G.Grothendieck just get the function at runtime. So have the `Function asMatrix("as.matrix")` inside `DF2mat`. – Romain Francois Jun 22 '14 at 15:04

best way to convert DataFrame to Matrix in RCpp

2 Answers2

Linked