CODEWITHSUNDEEP
c++rrcppintegrate
Abc123
Abc123

Reputation: 11

Error when calling C++ function from R and integrate it

I want to numerically integrate a 1-dimensional function (that was written in C++) with the R function integrate. As a short example, I coded the function myfunc in C++. 

    #include <cmath>
    #include <Rcpp.h>

    using namespace std;

    // [[Rcpp::export]]

    double myfunc (double x){
        double result;
        result = exp( -0.5*pow(x,2) + 2*x );
        return result;
    }

After loading myfunc in R and integrating it, I obtain the following error:

    library(Rcpp)
    sourceCpp("myfunc.cpp")
    integrate(myfunc,lower=0,upper=10)

Error in f(x, ...) : Expecting a single value: [extent=21].

Can anyone explain what this error means and how I can solve this problem?

Upvotes: 1

Views: 159

Answers (1)

duckmayr
duckmayr

Reputation: 16940

From help("integrate"):

f must accept a vector of inputs and produce a vector of function evaluations at those points. The Vectorize function may be helpful to convert f to this form.

You have created your function to accept a single value, a double, so when integrate() tries to pass it a vector, it rightfully complains. So, try

#include <Rcpp.h>

// [[Rcpp::export]]
Rcpp::NumericVector myfunc(Rcpp::NumericVector x){
    return exp(-0.5 * pow(x, 2) + 2 * x);
}

/*** R
integrate(myfunc, lower = 0, upper = 10)
*/

Resulting in

integrate(myfunc, lower = 0, upper = 10)
# 18.10025 with absolute error < 5.1e-08

Or, using myfunc() compiled from your C++ code from above,

f <- Vectorize(myfunc)
integrate(f, lower = 0, upper = 10)
# 18.10025 with absolute error < 5.1e-08

Upvotes: 4

Related Questions