CODEWITHSUNDEEP
rrcpp
Andrie
Andrie

Reputation: 179558

How to test Rcpp::CharacterVector elements for equality?

I am trying to write some simple Rcpp code examples. This is remarkably easy with the Rcpp and inline packages.

But I am stumped on how to test whether two character elements for equality. The following example compares the first elements of two character vectors. But I can't get it to compile.

What is the trick?

library(Rcpp)
library(inline)

cCode <- '
    Rcpp::CharacterVector cx(x);
    Rcpp::CharacterVector cy(y);
    Rcpp::LogicalVector r(1);
    r[0] = (cx[0] == cy[0]);
    return(r);
    '

cCharCompare <- cxxfunction(signature(x="character", y="character"), 
                            plugin="Rcpp", body=cCode)
cCharCompare("a", "b")

--

The comparison using == works perfectly fine if one of the two elements is a constant. The following code compiles and gives expected results:

cCode <- '
    Rcpp::CharacterVector cx(x);
    Rcpp::LogicalVector r(1);
    r[0] = (cx[0] == "a");
    return(r);
    '

cCharCompareA <- cxxfunction(signature(x="character"), plugin="Rcpp", body=cCode)

cCharCompareA("a")
[1] TRUE

cCharCompareA("b")
[1] FALSE

Upvotes: 17

Views: 5485

Answers (3)

Romain Francois
Romain Francois

Reputation: 17642

The equality operator has been introduced in Rcpp 0.10.4. The implementation looks like this in the string_proxy class:

bool operator==( const string_proxy& other){
    return strcmp( begin(), other.begin() ) == 0 ;
}

So now we can write: 

#include <Rcpp.h>
using namespace Rcpp ;

// [[Rcpp::export]]
LogicalVector test( CharacterVector x, CharacterVector y){
    Rcpp::LogicalVector r(x.size());
    for( int i=0; i<x.size(); i++){
        r[i] = (x[i] == y[i]);
    }
    return(r);
}

And something similar is used on our unit tests:

> test(letters, letters)
 [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[16] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE

Upvotes: 18

Dirk is no longer here
Dirk is no longer here

Reputation: 368519

Very nice (technical) answer by @kohske, but here is something more C++-ish: just compare strings!

library(inline)      ## implies library(Rcpp) when we use the plugin

cCode <- '
    std::string cx = Rcpp::as<std::string>(x);
    std::string cy = Rcpp::as<std::string>(y);
    bool res = (cx == cy);
    return(Rcpp::wrap(res));
    '

cCharCompare <- cxxfunction(signature(x="character", y="character"),
                            plugin="Rcpp", body=cCode)
cCharCompare("a", "b")

If you really want to compare just the first character of the strings, then you can go from x to x.c_str() and either index its initial element, or just dereference the pointer to the first char.

A more R-ish answer could maybe sweep over actual vectors of strings...

Upvotes: 14

kohske
kohske

Reputation: 66902

Try this:

  // r[0] = (cx[0] == cy[0]);
  // r[0] = ((char*)cx[0] == (char*)cy[0]); <- this is wrong
  r[0] = (*(char*)cx[0] == *(char*)cy[0]); // this is correct.

It is not easy to explain, but

  1. CharacterVector is not char[].
  2. operator [] returns StringProxy.
  3. StringProxy is not a type of char.
  4. StringProxy has a member operator function char* that convert StringProxy to char*.

So, maybe (char*)cx[0] is a pointer. Now I forget many things about C++ syntax...

The reason hy the compile fails is the failure of type inference in operator overload == for StringProxy.

Upvotes: 17

Related Questions