Using Rcout with Rcpp / RcppArmadillo to coordinate output with R

The new RcppArmadillo release 0.2.35 now supports the Rcpp::Rcout output stream device. Based on a contributed Rcpp patch by Jelper Ypma, the Rcpp::Rcout output stream gets redirected to R’s buffered output. In other words, R’s own output and that eminating from C++ code using Rcpp::Rcout are now both in sync. This avoids a stern warning from Section 5.6 in the Writing R Extensions manual:

Using C++ iostreams, as in this example, is best avoided. There is no guarantee that the output will appear in the R console, and indeed it will not on the R for Windows console. Use R code or the C entry points (*note Printing) for all I/O if at all possible.

and does in fact provide exactly what is recommended: the same entry points R itself uses.

Below is a sample program, once again using the wonderful inline package to compile, load and link C++ code into R from a simple text variable submitted to the cxxfunction. What is shown in R code to load the package, the definition of the C++ code as assigned to a variable src and the creation of the dynamically-loadaded R function called fun which contains the code from we compiled, link and load via a single call to cxxfunction() given src.

library
library(inline)

src <- '

  Rcpp::Rcout << "Armadillo version: " << arma::arma_version::as_string() << std::endl;

  // directly specify the matrix size (elements are uninitialised)
  arma::mat A(2,3);

  // .n_rows = number of rows    (read only)
  // .n_cols = number of columns (read only)
  Rcpp::Rcout << "A.n_rows = " << A.n_rows << std::endl;
  Rcpp::Rcout << "A.n_cols = " << A.n_cols << std::endl;

  // directly access an element (indexing starts at 0)
  A(1,2) = 456.0;

  A.print("A:");

  // scalars are treated as a 1x1 matrix,
  // hence the code below will set A to have a size of 1x1
  A = 5.0;
  A.print("A:");

  // if you want a matrix with all elements set to a particular value
  // the .fill() member function can be used
  A.set_size(3,3);
  A.fill(5.0);
  A.print("A:");


  arma::mat B;

  // endr indicates "end of row"
  B << 0.555950 << 0.274690 << 0.540605 << 0.798938 << arma::endr
    << 0.108929 << 0.830123 << 0.891726 << 0.895283 << arma::endr
    << 0.948014 << 0.973234 << 0.216504 << 0.883152 << arma::endr
    << 0.023787 << 0.675382 << 0.231751 << 0.450332 << arma::endr;

  // print to the cout stream
  // with an optional string before the contents of the matrix
  B.print("B:");

  // the << operator can also be used to print the matrix
  // to an arbitrary stream (cout in this case)
  Rcpp::Rcout << "B:" << std::endl << B << std::endl;

  // save to disk
  B.save("B.txt", arma::raw_ascii);

  // load from disk
  arma::mat C;
  C.load("B.txt");

  C += 2.0 * B;
  C.print("C:");


  // submatrix types:
  //
  // .submat(first_row, first_column, last_row, last_column)
  // .row(row_number)
  // .col(column_number)
  // .cols(first_column, last_column)
  // .rows(first_row, last_row)

  Rcpp::Rcout << "C.submat(0,0,3,1) =" << std::endl;
  Rcpp::Rcout << C.submat(0,0,3,1) << std::endl;

  // generate the identity matrix
  arma::mat D = arma::eye<arma::mat>(4,4);

  D.submat(0,0,3,1) = C.cols(1,2);
  D.print("D:");

  // transpose
  Rcpp::Rcout << "trans(B) =" << std::endl;
  Rcpp::Rcout << trans(B) << std::endl;

  // maximum from each column (traverse along rows)
  Rcpp::Rcout << "max(B) =" << std::endl;
  Rcpp::Rcout << max(B) << std::endl;

  // maximum from each row (traverse along columns)
  Rcpp::Rcout << "max(B,1) =" << std::endl;
  Rcpp::Rcout << max(B,1) << std::endl;

  // maximum value in B
  Rcpp::Rcout << "max(max(B)) = " << max(max(B)) << std::endl;

  // sum of each column (traverse along rows)
  Rcpp::Rcout << "sum(B) =" << std::endl;
  Rcpp::Rcout << sum(B) << std::endl;

  // sum of each row (traverse along columns)
  Rcpp::Rcout << "sum(B,1) =" << std::endl;
  Rcpp::Rcout << sum(B,1) << std::endl;

  // sum of all elements
  Rcpp::Rcout << "sum(sum(B)) = " << sum(sum(B)) << std::endl;
  Rcpp::Rcout << "accu(B)     = " << accu(B) << std::endl;

  // trace = sum along diagonal
  Rcpp::Rcout << "trace(B)    = " << trace(B) << std::endl;

  Rcpp::Rcout << std::endl;
'

fun <- cxxfunction(signature(), body=src, plugin="RcppArmadillo")

setwd("/tmp")                           # adjust on other OSs

fun()                                   # output to stdout

sink("rcpparma.log.txt")                # start 'sink' to output to file
fun()                                   # no output to screen
sink()                                  # stop 'sink'

We then switch to a temporary directory (as the example code, taken from one of the two examples in Conrad’s Armadillo sources, creates a temporary file) and run the new function. To demontrate how it does in fact now mesh perfectly with R, we create an output ‘sink’ (which catches all output) and re-run.

This simple example demonstrated how we can use the new Rcout output stream from Rcpp to have dynamically-loaded C++ code cooperate more cleanly with the (buffered) R output. It also demontrated some of the nice features in Armadillo which we bring to R via RcppArmadillo.