sort_stl.cpp

//' @title Sort using STL
//' @description if compiler flags and standard library support is available
//'   (only tested on glibc), OpenMP, then this will use a parallel sort
//'   algorithm which is significantly faster. It doesn't however deal with NA
//'   values. TODO: handle NA values.
//' @param x vector of strings. In the current implementation, NA is completely
//'   ignored, so is probably converted to "NA"
//' @examples
//' \dontrun{
//' pts <- icd9:::randomPatients(1e7)
//' microbenchmark::microbenchmark(sort_std(pts$icd9), sort(pts$icd9), times = 5)
//' # four times faster on 4 real core (8 with HT) machine.
//' }
// [[Rcpp::export]]
std::vector<std::string> sort_std(std::vector<std::string> x) {
  // will use parallel/algorithm if included
  std::sort(x.begin(), x.end());
  return(x);

}

include <parallel/algorithm>

_GLIBCXX_PARALLEL_PARALLEL_H
and also need -fopenmp

Better still, for factor creation in R, use boost multi_index to have insertion an order container with hash lookup (for dedupe during filling).