Rcpp: Floating point exception by using double type

Question

I tried to write a simple function in cpp which I can use in rcpp, like:

#include <Rcpp.h> 
using namespace Rcpp;

// [[Rcpp::export]]
inline static double calc(double a, double b, double c){ 
    return a - b + c;
}

// [[Rcpp::export]]
void other_function() {
    double res = calc(1.00, 0.00, 3.00);        // no error

    NumericVector resV;
    resV[0] = calc(1.00, 0.00, 3.00);   // Floating point exception

    Rcout << res;                       // Floating point exception
}

But every time if I try to access the result of the function calc() I got an Floating point exception.

If I only source the R File and call the funcion res() from the console everything seems to be fine:

source("myscript.R")
calc(1.00,0.00,3.00)
>> 4.00

What I already tried:

• convert double to NumericVector res and only working with res[0] > not working

• try float > not working

• do the calculation direct in the other_function > same error

UPDATE:

Everytime I try to call other_function() in an other function I get the following message:

Floating point exception

For example:

void another_function() {
    other_function();       // produce Floating point exception
}

If I call the other_function() directly in R everything is fine!

Also this code (thanks to snake_style) didn´t work for me:

NumericVector resV = NumericVector::create(progressionC(1.00, 0.00, 3.00));

---

Solved

I solved the problem by simple reinstall the Rcpp package. I don´t know why I got this error... anyway, now it works.

Answer 1

Not an answer but I want to store my failed attempts at reproducing this issue. I am using R 3.5.1 on Debian stable with gcc 6.3. I used the following code:

#include <Rcpp.h> 
using namespace Rcpp;

// [[Rcpp::export]]
inline static double calc(double a, double b, double c){ 
  return a - b + c;
}

// [[Rcpp::export]]
void other_function() {
  double res = calc(1.00, 0.00, 3.00);        // no error

  NumericVector resV;
  resV[0] = calc(1.00, 0.00, 3.00);   // Floating point exception

  NumericVector resV1(1);
  resV1[0] = calc(1.00, 0.00, 3.00);   // Floating point exception

  NumericVector resV2 = NumericVector::create(calc(1.00, 0.00, 3.00));
  Rcout << res << "/" << resV << "/" << resV1 << "/" << resV2;                       // Floating point exception
}


// [[Rcpp::export]]
void another_function() {
  other_function();       // produce Floating point exception
}

/*** R
calc(1.00, 0.00, 3.00)
other_function()
another_function()
*/

Calling Rcpp::sourceCpp() compiles this code and produces the following output:

> calc(1.00, 0.00, 3.00)
[1] 4

> other_function()
4//4/4
> another_function()
4//4/4

• There are no runtime errors.
• In your original code the result from calc is not assigned (resV outputs nothing).
• Assignment works if either the size of the vector is specified or if calc is used with the static create method as suggested by @snake_style.

Answer 2

Try like this:

NumericVector resV = NumericVector::create(calc(1.00, 0.00, 3.00));