Handling Strings with Rcpp

Kevin Ushey — written Jan 8, 2013 — source

This is a quick example of how you might use Rcpp to send and receive R ‘strings’ to and from R. We’ll demonstrate this with a few operations.

Sort a String with R

Note that we can do this in R in a fairly fast way:

my_strings <- c("apples", "and", "cranberries")
R_str_sort <- function(strings) {
  sapply( strings, USE.NAMES=FALSE, function(x) {
    intToUtf8( sort( utf8ToInt( x ) ) )
    })
  }
R_str_sort( my_strings )

[1] "aelpps"      "adn"         "abceeinrrrs"

Sort a String with C++/Rcpp

Let’s see if we can re-create the output with Rcpp.

#include <Rcpp.h>
using namespace Rcpp;
 
// [[Rcpp::export]]
std::vector< std::string > cpp_str_sort( std::vector< std::string > strings ) {
  
  int len = strings.size();
 
  for( int i=0; i < len; i++ ) {
    std::sort( strings[i].begin(), strings[i].end() );
  }
  
  return strings;
}

Note the main things we do here:

Rcpp’s attributes handle any as-ing and wrap-ing of vectors; we even just specify our return type as std::vector< std::string >.
We then call the void method std::sort, which can sort a string in place,
… and we return that vector of strings.

Now, let’s test it, and let’s benchmark it as well.

cpp_str_sort( my_strings )

[1] "aelpps"      "adn"         "abceeinrrrs"

long_strings <- rep( paste( collapse="", sample( letters, 1E5, replace=TRUE ) ),
                     times=100 )
 
rbenchmark::benchmark( cpp_str_sort(long_strings),
                       R_str_sort(long_strings),
                       replications=3
                       )

                        test replications elapsed relative user.self
1 cpp_str_sort(long_strings)            3   0.898    1.000     0.883
2   R_str_sort(long_strings)            3   2.356    2.624     2.350
  sys.self user.child sys.child
1    0.014          0         0
2    0.007          0         0

Note that the C++ implementation is quite a bit faster (on my machine). However, std::sort will not handle UTF-8 encoded vectors.

Now, let’s do something crazy – let’s see if we can use Rcpp to perform an operation that takes a vector of strings, and returns a list of vectors of strings. (Or, in R parlance, a list of vectors of type character).

We’ll do a simple ‘split’, such that each string is split every n indices.

Split a string at consecutive indices n

#include <Rcpp.h>
using namespace Rcpp;
 
// [[Rcpp::export]]
List cpp_str_split( std::vector< std::string > strings, int n ) {
  
  int num_strings = strings.size();
  
  List out(num_strings);
  
  for( int i=0; i < num_strings; i++ ) {
    
    int num_substr = strings[i].length() / n;
    std::vector< std::string > tmp;
    
    for( int j=0; j < num_substr; j++ ) {
      
      tmp.push_back( strings[i].substr( j*n, n ) );
      
    }
    
    out[i] = tmp;
    
  }
  
  return out;
}

Main things to notice:

We declare the output to be a List,
We form a List container of size num_strings,
We construct the split strings one by one, then place them back into our output container (note how with out[i] = tmp, we can assign our vector of strings directly as an element of the list),
We return the list we constructed.

cpp_str_split( c("abcd", "efgh", "ijkl"), 2 )

[[1]]
[1] "ab" "cd"

[[2]]
[1] "ef" "gh"

[[3]]
[1] "ij" "kl"

cpp_str_split( c("abc", "de"), 2 )

[[1]]
[1] "ab"

[[2]]
[1] "de"

My solution is perhaps a bit deficient (bug or feature?) in that it truncates any strings not long enough; ideally, we’d either improve the C++ code or form an appropriate wrapper to the function in R (and warn the user if truncation might occur).

Hopefully this gives you a better idea how you might use Rcpp to perform more extensive string manipulation with R character vectors.

tags: string vector

Extending R with C++ and Fortran — Dirk Eddelbuettel and JBrandon Duck-Mayr
Benchmarking Rcpp code with RcppClock — Zach DeBruine
Simulation Smoother using RcppArmadillo — Tomasz Woźniak