Why does this PHP spintax code repeat identical iterations?

Question

http://ronaldarichardson.com/2011/09/23/recursive-php-spintax-class-3-0/

I like this script, but it isn't perfect. If you use this test input case:

{This is my {spintax|spuntext} formatted string, my {spintax|spuntext} formatted string, my {spintax|spuntext} formatted string example.}

You can see that the result ALWAYS contains 3 repetitions of either "spintax" or "spuntext". It never contains 1 "spintax" and 2 "spuntext", for example.

Example:

This is my spuntext formatted string, my spuntext formatted string, my spuntext formatted string example.

To be truly random it needs to generate a random iteration for each spintax {|} block and not repeat the same selection for identical blocks, like {spintax|spuntext}.

If you look at comment #7 on that page, fransberns is onto something, however when using his modified code in a live environment, the script would repeatedly run in an infinite loop and eat up all the server memory. So there must be a bug there, but I'm not sure what it is.

Any ideas? Or does anyone know of a robust PHP spintax script that allows for nested spintax and is truly random?

Answer 1

Please check this gist, it is working (and it is far simpler than original code ..).

Answer 2

The reason the Spintax class replaces all instances of {spintax|spuntext} with the same randomly chosen option is because of this line in the class:

  $str = str_replace($match[0], $new_str, $str);

The str_replace function replaces all instances of the substring with the replacement in the search string. To replace only the first instance, progressing in a serial fashion as you desired, we need to use the function preg_replace with a passed "count" argument of 1. However, when I looked over your link to the Spintax class and reference to post #7 I noticed an error in his suggested augmentation to the Spintax class.

fransberns suggested replacing:

$str = str_replace($match[0], $new_str, $str);

with this:

//one match at a time
$match_0 = str_replace("|", "\|", $match[0]);
$match_0 = str_replace("{", "\{", $match_0);
$match_0 = str_replace("}", "\}", $match_0);
$reg_exp = "/".$match_0."/";
$str = preg_replace($reg_exp, $new_str, $str, 1);

The problem with fransbergs' suggestion is that in his code he did not properly construct the regular expression for the preg_replace function. His error came from not properly escaping the \ character. His replacement code should have looked like this:

//one match at a time
$match_0 = str_replace("|", "\\|", $match[0]);
$match_0 = str_replace("{", "\\{", $match_0);
$match_0 = str_replace("}", "\\}", $match_0);
$reg_exp = "/".$match_0."/";
$str = preg_replace($reg_exp, $new_str, $str, 1);

Consider replacing the original class with this augmented version utilizing my correction on fransberns' suggested replacemnet:

class Spintax {

   function spin($str, $test=false)
   {
      if(!$test){
         do {
            $str = $this->regex($str);
         } while ($this->complete($str));
         return $str;
      } else {
         do {
            echo "<b>PROCESS: </b>";var_dump($str = $this->regex($str));echo "<br><br>";
         } while ($this->complete($str));
         return false;
      }
   }

   function regex($str)
   {
      preg_match("/{[^{}]+?}/", $str, $match);
      // Now spin the first captured string
      $attack = explode("|", $match[0]);
      $new_str = preg_replace("/[{}]/", "", $attack[rand(0,(count($attack)-1))]);
//      $str = str_replace($match[0], $new_str, $str); //this line was replaced
      $match_0 = str_replace("|", "\\|", $match[0]);
      $match_0 = str_replace("{", "\\{", $match_0);
      $match_0 = str_replace("}", "\\}", $match_0);
      $reg_exp = "/".$match_0."/";
      $str = preg_replace($reg_exp, $new_str, $str, 1);    
      return $str;
   }

   function complete($str)
   {
      $complete = preg_match("/{[^{}]+?}/", $str, $match);
      return $complete;
   }
}

When I tried using fransberns' suggested replacement "as is", because of the improper escaping of the \ character, I got an infinite loop. I assume that this is where your memory problem came from. After correcting fransberns' suggested replacement with the correct escaping of the \ character I did not enter an infinite loop.

Try the class above with the corrected augmentation and see if it works on your server (I can't see a reason why it shouldn't).