How to pass in functions as arguments in Rcpp / C++?

Question

I'm trying to write a function which can take in functions as its arguments in Rcpp. I have written an example function in R that shows the kind of functionality that I'm aiming for:

simulate_and_evaluate <- function(simulate, evaluate) {
  y <- simulate(1)
  eval <- evaluate(y)
  return(eval)
}

simulate_fun <- function(n) rnorm(n, 0, 1)
evaluate_fun <- function(x) dnorm(x, 0, 1)

simulate_and_evaluate(simulate = simulate_fun,
                      evaluate = evaluate_fun)

In this function simulate_and_evaluate, this takes in two arguments which are both functions, one that simulates a number and one that evaluates a function with this simualted number. So as an example, we can simulate a value from a standard normal and evaluate the density of a standard normal at that point. Does anyone know if there's a way to do this in Rcpp?

Is there a reason you are tying to use Rcpp here? Is it to speed up code? Because if you are trying to call a standard R function in c++ land you probably wont get much of a boost. Rcpp works create if you are running efficient c++ code. But that doesn't seem to be what you are trying to do here. — MrFlick
– MrFlick, Commented Dec 12, 2019 at 21:36
Also see package RcppDE with a worked example of passing compiled user-supplied functions around to an optimizer. — Dirk is no longer here
– Dirk is no longer here, Commented Dec 13, 2019 at 1:34
@MrFlick In the actual functions that I wish to implement, I take in these functions to simulate and evaluate densities, but also perform other operations with these results. I hoped that by being able to turn some of the R code that did this into Rcpp code, this would give me some improved performance. I understand that by just calling the basic R functions in C++ wouldn't improve performance, but hoped the other things the function would do with the results of these functions would increase efficiency. — user-2147482565
– user-2147482565, Commented Dec 13, 2019 at 13:07

Dirk is no longer here · Accepted Answer · 2019-12-13 01:50:32Z

Rcpp aims for seamless interfacing of R and C++ objects. As functions are first class R objects represented internally as a type a SEXP can take, we can of course also ship them with Rcpp. There are numerous examples.

So here we simply rewrite your function as a C++ function:

Rcpp::cppFunction("double simAndEval(Function sim, Function eval) {
  double y = as<double>(sim(1));
  double ev = as<double>(eval(y));
  return(ev);
}")

And we can then set the RNG to the same value, run your R function and this C++ function and get the same value. Which is awesome.

R> set.seed(123)
R> simulate_and_evaluate(simulate = simulate_fun,
+ evaluate = evaluate_fun)
[1] 0.341
R> set.seed(123) # reset RNG
R> simAndEval(simulate_fun, evaluate_fun)
[1] 0.341
R>

But as @MrFlick warned you, this will not run any faster because we added no compiled execution of the actual functions we are merely calling them from C++ rathern than R.

The topic has been discussed before. Please search StackOverflow, maybe with a string [rcpp] Function to get some meaningful hits.

Collectives™ on Stack Overflow

How to pass in functions as arguments in Rcpp / C++?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related