[Rcpp-devel] Problems returning fload

[Romain Francois]

I suspect this is about your float[] not being initialized. I would 
recommend you to have a look at containers such as std::vector<> from 
the STL.


But this is a version that only uses Rcpp classes:


#include <Rcpp.h>
using namespace Rcpp ;

// [[Rcpp::export]]
NumericVector test_func(NumericMatrix my_matrix){
     int num_cols = my_matrix.ncol(); // extract matrix dimensions

     NumericVector my_expr(num_cols);  // output array

     for(int j = 0; j < num_cols; j++) {
         double sum_ = sum( my_matrix(_,j) ) ;
         my_expr[j] = sum_ / ( sum_ + j ) ;
     }
     return my_expr;
}

Put this in a file (foo.cpp) and then do:

 > sourceCpp( "foo.cpp" )
 > test_func( my_matrix)
[1] 1.0000000 0.6666667 0.0000000 0.4000000 0.5000000 0.2857143 0.3333333
[8] 0.3000000

This uses the Rcpp sugar sum function to sum a column: my_matrix(_,j) is 
a view on the j^th column.


Romain

Le 19/12/12 19:45, Jeff Stevens a écrit :
> Hi,
>
> I'm trying to use Rcpp to speed up some nasty for loops in my R code
> (Ubuntu 12.04, R 2.15.2, Rcpp 0.10.1, gcc 4.6.3).  However, I'm a newbie
> to C and to Rcpp, and I'm having trouble returning the correct output.
> Basically, I input an R matrix to an inline (using Rcpp plugin) C
> function that runs through each element of the matrix and performs some
> calculations to generate an array (float).  When I run the program in C,
> it calculates the correct values.  However, when I run it in R, it give
> erratic results.  Here is example code that results in the same problem:
> test_src <- '
> Rcpp::NumericMatrix my_matrix(a);   // input matrix
> int num_rows = my_matrix.nrow(), num_cols = my_matrix.ncol(); // extract
> matrix dimensions
> float my_sum[50];
> float my_sum2[50];
> Rcpp::NumericVector my_expr(num_cols);  // output array
>
>      for(int i = 0; i < num_rows; i++) {
>        for(int j = 0; j < num_cols; j++) {
>        my_sum[j] = my_sum[j] + my_matrix(i,j);   // sum columns
>        my_sum2[j] = (my_sum[j] + j);             // sum columns plus the
> column index
>        }
>      }
>      for(int j = 0; j < num_cols; j++) {
>        my_expr[j] = my_sum[j] / my_sum2[j];      // divide sum by sum2
>      }
> return my_expr;
> '
> test_func <- cxxfunction(signature(a = "numeric"), test_src, plugin =
> "Rcpp")
>
> my_matrix <- matrix(c(0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0,
> 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0),
> nrow = 5, ncol = 8)
>
> test_func(my_matrix)
>
> When I run this, I get:
> 1.0000000  0.6666667       NaN       NaN 1.0000000 0.2857143 0.3333333
> 0.3000000
>
> If I run the following in C:
> int num_rows = 5, num_cols = 8;
> int my_matrix[5][8];
> float my_sum[50];
> float my_sum2[50];
> float my_expr[50];
> int my_matrix[5][8] = {
>      {0, 1, 1, 0, 1, 0, 0, 1},
>      {1, 0, 0, 0, 0, 0, 0, 1},
>      {0, 0, 1, 0, 1, 1, 1, 1},
>      {0, 1, 1, 0, 0, 0, 1, 1},
>      {0, 0, 1, 0, 1, 1, 1, 0}
> };
>
> int main () {
>      for(int i = 0; i < num_rows; i++) {
>        for(int j = 0; j < num_cols; j++) {
>        my_sum[j] = my_sum[j] + my_matrix[i][j];  // sum columns
>        my_sum2[j] = (my_sum[j] + j);             // sum columns plus the
> column index
>        }
>      }
>      for(int j = 0; j < num_cols; j++) {
>        my_expr[j] = my_sum[j] / my_sum2[j];      // divide sum by sum2
>        printf("%f\t", my_expr[j]);
>      }
> }
>
> I get
> 1.000000  0.666667   0.666667   0.000000   0.428571   0.285714
> 0.333333   0.363636
>
> I'm sure that there is something simple that I am doing wrong.  Any
> suggestions?
>
> Thanks,
> Jeff
>
>
> _______________________________________________
> Rcpp-devel mailing list
> Rcpp-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
>


-- 
Romain Francois
Professional R Enthusiast
+33(0) 6 28 91 30 30

R Graph Gallery: http://gallery.r-enthusiasts.com

blog:            http://romainfrancois.blog.free.fr
|- http://bit.ly/RE6sYH : OOP with Rcpp modules
`- http://bit.ly/Thw7IK : Rcpp modules more flexible