R - C++ code reproduction in Rcpp Math.Min issue

Question

I am trying to reproduce a C++ code that I found here about the LevenshteinDistance.

More precisely, I am trying to reproduce the part starting by

static int LevenshteinDistance(string s, string t)

until

  return d[n, m];
 }

However, I am experiencing an error with Math.Min(Math.Min(. I am not sure what is the translation in rcpp. I tried with min but it doesn't seem to work. Any clue ?

// [[Rcpp::export]]
NumericVector LevenshteinDistance(NumericVector s, NumericVector t) {
    int n = s.size();
    int m = t.size();
    NumericMatrix d(n+1, m+1);

    if (n == 0)
    {
      return m;
    }

    if (m == 0){
      return n;
    }

    for (int i = 0; i < n; i++)
      d(i, 0) = i;
    for (int j = 0; j < m; j++)
      d(0, j) = j;

    for (int j = 1; j < m; j++)
      for (int i = 1; i < n; i++)
        if (s(i - 1) == t(j - 1))
          d(i, j) = d(i - 1, j - 1);  //no operation
        else
            d(i, j) = min(
            d(i - 1, j) + 1,    //a deletion
            d(i, j - 1) + 1),   //an insertion
            d(i - 1, j - 1) + 1 //a substitution
        );
        return d(n, m);
  }

So with

d(i, j) = min(

I get this error message : no matching function.

Answer 1

Three issues really:

1. You have a typo

   d(i, j) = min(
   d(i - 1, j) + 1,    
   d(i, j - 1) + 1),  //typo here with )
     d(i - 1, j - 1) + 1 
   )
   

2. The use of std::min() is defined to be between two values e.g. std::min(obj1,obj2) whereas you are aiming use min() in an R way.

So,

min(c(1,2,3,4))

But, C++ definition of std::min requires it to be written as:

std::min(1, std::min(2, std::min(3,4)));

3. If you want to be true to the string notion of LevenshteinDistance, then you need to switch how the values are entering your function away from the NumericVector type to any one of the following: std::string, std::vector<std::string> (see Gallery Post: Strings with Rcpp), Rcpp::StringVector (see Gallery Post: Working with Rcpp StringVector), or Rcpp::CharacterVector.

The following should be sufficient:

#include <Rcpp.h>

// [[Rcpp::export]]
int LevenshteinDistance(std::string s,
                        std::string t) {

  // Number of elements
  int n = s.size();
  int m = t.size();
  Rcpp::IntegerMatrix d(n+1, m+1);
  Rcpp::Rcout << "n:" << n << ", m:" << m << std::endl;

  if (n == 0){
    return m;
  }

  if (m == 0){
    return n;
  }

  for (int i = 0; i <= n; i++){
    d(i, 0) = i;
  }

  // No sense to revisit the (0,0) coordinate
  for (int j = 1; j <= m; j++){
    d(0, j) = j;
  }

  for (int j = 1; j <= m; j++){

    for (int i = 1; i <= n; i++){

      if (s[i - 1] == t[j - 1]){

        d(i, j) = d(i - 1, j - 1);  // no operation

      } else {

        d(i, j) = std::min(d(i - 1, j) + 1,    //a deletion
                           std::min(d(i, j - 1) + 1,   //an insertion
                           d(i - 1, j - 1) + 1)); //a substitution

      } // end if

    } // end inner for

  } // end outer for

  return d(n, m);
}

With this function being written, we can opt to write a quick wrapper to process different dimensions of comparisons such multiple sources to a target and so on...

// [[Rcpp::export]]
Rcpp::IntegerVector LevenshteinDistance_Vector(std::vector<std::string> s,
                                               std::string t){

  // Number of Sources
  unsigned int sn = s.size();

  Rcpp::IntegerVector o(sn); // populates automatically with 0.

  for(unsigned int i = 0; i < sn; i++){
    o(i) = LevenshteinDistance(s[i],t);
  }

  return o;
}