How do you implement Levenshtein distance in Delphi?

Question

I'm posting this in the spirit of answering your own questions.

The question I had was: How can I implement the Levenshtein algorithm for calculating edit-distance between two strings, as described here, in Delphi?

Just a note on performance: This thing is very fast. On my desktop (2.33 Ghz dual-core, 2GB ram, WinXP), I can run through an array of 100K strings in less than one second.

Answer 1

function EditDistance(s, t: string): integer;
var
  d : array of array of integer;
  i,j,cost : integer;
begin
  {
  Compute the edit-distance between two strings.
  Algorithm and description may be found at either of these two links:
  http://en.wikipedia.org/wiki/Levenshtein_distance
  http://www.google.com/search?q=Levenshtein+distance
  }

  try
    //initialize our cost array
    SetLength(d,Length(s)+1);
    for i := Low(d) to High(d) do begin
      SetLength(d[i],Length(t)+1);
    end;

    for i := Low(d) to High(d) do begin
      d[i,0] := i;
      for j := Low(d[i]) to High(d[i]) do begin
        d[0,j] := j;
      end;
    end;

    //store our costs in a 2-d grid  
    for i := Low(d)+1 to High(d) do begin
      for j := Low(d[i])+1 to High(d[i]) do begin
        if s[i] = t[j] then begin
          cost := 0;
        end
        else begin
          cost := 1;
        end;

        //to use "Min", add "Math" to your uses clause!
        d[i,j] := Min(Min(
                   d[i-1,j]+1,      //deletion
                   d[i,j-1]+1),     //insertion
                   d[i-1,j-1]+cost  //substitution
                   );
      end;  //for j
    end;  //for i

    //now that we've stored the costs, return the final one
    Result := d[Length(s),Length(t)];
  finally
    //cleanup
    for i := Low(d) to High(d) do begin
      for j := Low(d[i]) to High(d[i]) do begin
        SetLength(d[i],0);
      end;  //for j
    end;  //for i
    SetLength(d,0);
  end;  //try-finally
end;