CODEWITHSUNDEEP
c++rrcpp
Siqi.Xiang
Siqi.Xiang

Reputation: 23

Calling R's optim function from within C++ using Rcpp

I'm new for Rcpp, and in my code, I must call the R function "optim" from C++. I referred to many examples, but there is still a mistake: "static assertion failed: cannot convert type to SEXP". Here is my code, and the problem is the last function:

#include <RcppArmadillo.h>

using namespace Rcpp;
using namespace RcppArmadillo;
using namespace arma;
using namespace std;
// [[Rcpp::depends(RcppArmadillo)]]

// [[Rcpp::plugins(cpp11)]]

// [[Rcpp::export]]
double fr(arma::colvec x){
  double result = 100 * (x(2) - x(1) * x(1)) * (x(2) - x(1) * x(1)) + (1 -     x(1)) * (1 - x(1));
  return result;  
} 


typedef double (*funcPtr)(arma::colvec x);

// [[Rcpp::export]]
XPtr<funcPtr> putFunPtrInXPtr(){
  return(XPtr<funcPtr>(new funcPtr(&fr)));
}


// [[Rcpp::export]]
arma::colvec callOptimFun(SEXP x) {

  RNGScope scope;

  Rcpp::Environment stats("package:stats");
  Rcpp::Function optim = stats["optim"];
  XPtr<funcPtr> xpfun = putFunPtrInXPtr();
  funcPtr fun = *xpfun;
  Rcpp::List y = optim(x, fun);
  arma::colvec r = y["par"];
  return r; 
}

Unfortunately, I have tried lots of methods for my last function, and all methods have the same error. These are my tries: 1.

// [[Rcpp::export]]
Rcpp::List callOptimFun(arma::colvec x) {

  \\....
  Rcpp::List y = optim(x, fun);
  return y; 
}

2.

// [[Rcpp::export]]
Rcpp::List callOptimFun(arma::colvec x) {

  \\....
  Rcpp::List y = optim(x, fun);
  return y; 
}

3.

// [[Rcpp::export]]
Rcpp::List callOptimFun(SEXP x) {

  \\....
  Rcpp::List y = optim(x, fun);
  return y; 
}

I am not familiar with C++. What may be the problem? Thanks!

Upvotes: 2

Views: 1079

Answers (1)

coatless
coatless

Reputation: 20746

In this case, the use of a functional pointer is problematic. Rcpp has a special type of wrapper that requires the C++ function to not be exported into R called Rcpp::InternalFunction. Under this implementation, you're able to easily incorporate R's optim with your C++ function from C++!

Also, I modified the element accessed indexes in fr function. The reason for this modification is you are hitting an out-of-bounds error because the indexes in C++ are on a zero-based starting system not a one-based system starting system like R. (e.g. x(0) is the same as x[1] in R)

#include<RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]

double fr(arma::vec x){
  double result = 100 * (x(1) - x(0) * x(0)) * (x(1) - x(0) * x(0)) + (1 - x(0)) * (1 - x(0));
  return result;  
} 

// [[Rcpp::export]]
arma::vec optim_rcpp(const arma::vec& init_val){

  Rcpp::Environment stats("package:stats"); 
  Rcpp::Function optim = stats["optim"];    

  Rcpp::List opt_results = optim(Rcpp::_["par"]    = init_val,
                                 Rcpp::_["fn"]     = Rcpp::InternalFunction(&fr),
                                 Rcpp::_["method"] = "BFGS");

  // Extract and coerce from list.
  arma::vec out = Rcpp::as<arma::vec>(opt_results[0]);

  return out;
}

Upvotes: 4

Related Questions