Question

In the documentation for the .C function in R here, it says that the C function itself must be of return type void. I’m working on some inherited code that uses the .C function (rather than .Call or Rcpp), but I’d like to write a function that “returns” a vector of unknown size. If I knew the size of this vector, I could allocate it in R and pass the pointer to C, whose contents can then be modified. Given that I don’t know its size, what is the way to declare it in R? I imagine it will have this shape, but I don’t know the details:

output <- ?????
storage.mode(a) <- "integer"
tmp <- .C("myfunc",a,output)

Answer 1

I suggest you do use .Call() instead of .C() — see recent discussions on the r-devel list. I conjecture that you will find close to no experienced R developers suggesting to use .C() for new projects.

Here is a simple illustration. You are by no means forced to use Rcpp, the same code can be written in the plain C API as well, albeit not as compactly:

R> fx <- rcpp(signature(sn="integer"), body='   // rcpp() is from inline 0.3.10
+    int size = as<int>(sn);
+    IntegerVector v(size); 
+    for (int i=0; i<size; i++) v[i] = i;
+    return(v); 
+ ') 
R> fx(3) 
[1] 0 1 2   
R> fx(7) 
[1] 0 1 2 3 4 5 6  
R>

In the old .C() interface, you can only return atomistic C types like *int or *double and you cannot allocate space at the C level as you can only “mark” SEXP object for management by R when you can actually return SEXP objects — and only .Call() can.