as_cpp<int>(LGLSXP scalar)

[romainfrancois]

This is mostly useful for NA :

cpp11::cpp_function("int echo(int x){ return x; }")
str(echo(1))
#>  int 1
str(echo(2L))
#>  int 2
str(echo(NA))
#>  int NA

^{Created on 2020-07-23 by the reprex package (v0.3.0.9001)}

[jimhester]

Hmm, I am not sure we want this though, I think you should have to pass NA_integer_ in this case. I generally have tried to avoid implicit coercions in the cpp11 interfaces.

[romainfrancois]

I'd say most users regard NA as typeless, and I figured since there already is a branch for REALSXP it would not be too bad. Maybe that could be limited to NA and still abort for TRUE and FALSE.

[jimhester]

Sure, doing the conversion for NA only makes sense to me, and I guess it the coercion is relatively benign for this case.

[romainfrancois]

🤔 now that I've seen that as_cpp<T> can handle various types of integer for T and that NA_real also isn't handled, maybe it's not needed at all/confusing.

cpp11::cpp_function('int echo(int x){
  return x;
}')
echo(NA_real_)
#> Error: Expected single integer value

^{Created on 2020-07-27 by the reprex package (v0.3.0.9001)}

[romainfrancois]

I mean we would have to come up with NA for long long etc ... maybe instead for scalars that might be NA, we could have some wrapper class that would deal with it, i.e. something like:

template <typename T>
scalar {
  scalar(SEXP from) : from_(from){ // assert that this is indeed a scalar of a type compatible to T}

  bool is_na();
  inline T get();

  SEXP from;
};