--- spam.module.orig	2005-01-25 21:46:14.000000000 +0000
+++ spam.module	2005-01-27 00:16:54.000000000 +0000
@@ -122,7 +122,7 @@
         $tokens = spam_tokenize($node->title, 'title*');
         $tokens = array_merge($tokens, spam_tokenize($node->body));
         $weight += spam_limit_urls('content', spam_count_urls());
-        if (($rating = _spam_rating($tokens, $weight)) >= 
+        if (($rating = _spam_rating($tokens, $weight)) >=
            variable_get('spam_threshold', 80)) {
           spam_node_actions($node, $tokens, 1, $op);
         }
@@ -236,7 +236,7 @@
     $group .= $inner_output;
   }
   $output .= form_group('Advanced configuration', $group);
-  
+
   return $output;
 }
 
@@ -260,7 +260,7 @@
   $items = array();
   if ($may_cache) {
     $items[] = array('path' => 'admin/comment/list/spam', 'title' => t('spam'),
-      'access' => user_access('administer spam rating'), 
+      'access' => user_access('administer spam rating'),
       'callback' => 'spam_admin', 'type' => MENU_LOCAL_TASK);
     $items[] = array('path' => 'admin/node/spam', 'title' => t('spam'),
       'access' => user_access('administer spam rating'),
@@ -333,7 +333,7 @@
   $sql .= tablesort_sql($header);
   $result = pager_query($sql,  50);
 
-  // Make sure the update controls are disabled if we don't have any rows 
+  // Make sure the update controls are disabled if we don't have any rows
   // to select from.
   $disabled = !db_num_rows($result);
 
@@ -396,7 +396,7 @@
   $sql .= tablesort_sql($header);
   $result = pager_query($sql,  50);
 
-  // Make sure the update controls are disabled if we don't have any rows 
+  // Make sure the update controls are disabled if we don't have any rows
   // to select from.
   $disabled = !db_num_rows($result);
 
@@ -847,7 +847,7 @@
 function spam_page() {
   $content = arg(1);
   $op = arg(3);
-  
+
   switch ($content) {
     case 'comment':
       if ($op) {
@@ -883,7 +883,7 @@
           }
           drupal_set_message(t('Comment marked as spam.'));
           drupal_goto("node/$comment->nid#comment-$comment->cid");
-        } 
+        }
         else {
           if (variable_get('spam_unpublish', 0)) {
             spam_admin_publish_comment($comment->cid);
@@ -927,7 +927,7 @@
           }
           drupal_set_message(t("$node->type marked as spam."));
           drupal_goto("node/$node->nid");
-        } 
+        }
         else {
           if (variable_get('spam_unpublish', 0)) {
             spam_admin_publish_node($node->nid);
@@ -944,7 +944,7 @@
 
 /**
  * Break text into words. Special handling currently exists for urls.
- *  
+ *
  * @param $string A string of text to tokenize
  * @param $tag An optional tag to prepend to each token
  * @return An array of all words obtained from string
@@ -1097,40 +1097,37 @@
 
 /**
  * Test array of words against known words, determine probability is spam
- *  
+ *
  * @param $tokens An array of tokens
  * @return An int from 1 to 99 which is the probability that $tokens are spam
  */
 // test whether or not current content in form of token array is spam
 function _spam_rating($tokens = array(), $weight = 0) {
-  // build token array, use drift as index
-  foreach ($tokens as $token) {
-    $p = db_fetch_object(db_query("SELECT probability FROM {spam_tokens} WHERE token = '%s'", $token));
-    if (!$p->probability) {
-      $p->probability = variable_get('spam_unknown_probability', 40);
-    }
-    // get drift from median of 50
-    $t["$token,$p->probability"] = abs($p->probability - 50);
-  }
-  // sort so largest drift is first
-  asort($t);
-  $keys = array_keys($t);
-  $max = variable_get('spam_interesting_tokens', 15);
-  $total = 0;
-  // grab n tokens with largest drift
-  for ($i = 0; $i < $max; $i++) {
-    if ($pair = array_shift($keys)) {
-      $p = explode(',',$pair);
-      // add up combined probabilities
-      $total = $total + $p[1];
+
+  // $drift - minimal drift from the median - a soft-ish shoulder of the filter
+  // $max - maximum number of tokens to evaluate, a hard shoulder of the filter
+  $probs = array();
+  $drift = variable_get('spam_min_drift', 40);
+  //since the tokens are already filtered by drift, it may prove useful to
+  //get a few more anyway - this is more a memory saving and speed improvement
+  //measure than precision. It is somewhat akin to brain surgery, but fast.
+  $max = variable_get('spam_interesting_tokens', 40);
+  $num=0;
+
+  //Speed increase depends on asort() and the number of interesting tokens considered
+  //vs the added comparison and evaluation in SQL
+  foreach($tokens as $token){
+    $p = db_fetch_object(db_query("SELECT probability FROM {spam_tokens} WHERE token='%s' AND (ABS(50 - probablility) >= %d) AND last >= %d SORT BY ABS(50 - probability)", $token,$drift,$drift));
+    if($p->probability){
+      $probs += $p->probability;
+      $num++;
     }
-    else {
-      // no more tokens
+    if($num>$max) {
       break;
     }
   }
 
-  $rating = ($total + $weight) / $i;
+  $rating = ($probs + $weight) / $num;
   if ($rating > 99)
     $rating = 99;
   else if ($rating < 1)
@@ -1144,7 +1141,7 @@
  * bayesian filter.  When passed in, each counter name should start with a
  * '+' or a '-', indicating whether the counter should be incremented or
  * decremented.
- *  
+ *
  * @param $counters An array of counter names
  */
 function _count($counters) {
@@ -1170,9 +1167,9 @@
  * Saves an array of tokens to the database, marking as 'spam' or 'notspam'.
  * If the token already exists in the database, then the appropriate counter
  * is incremented, tracking how many times it has been seen.
- *  
+ *
  * @param $tokens An array of tokens to be saved in the database
- * @param $type String, either 'spam' or 'notspam' 
+ * @param $type String, either 'spam' or 'notspam'
  */
 function spam_save_tokens($tokens, $type) {
   foreach ($tokens as $token) {
@@ -1198,9 +1195,9 @@
  * on what type of action is requested, either the token's 'spam' counter or
  * 'notspam' counter will be decremented.  This function is used to unlearn
  * specific content that was mis-marked as spam or notspam.
- *  
+ *
  * @param $tokens An array of tokens to be saved in the database
- * @param $type String, either 'spam' or 'notspam' 
+ * @param $type String, either 'spam' or 'notspam'
  */
 function spam_unsave_tokens($tokens, $type) {
   foreach ($tokens as $token) {
@@ -1213,7 +1210,7 @@
   }
   if ($type == 'spam')
       _count(array("+unlearned_notspam"));
-  else 
+  else
       _count(array("+unlearned_spam"));
   variable_set('spam_calculate_probabilities', 1);
 }
@@ -1230,7 +1227,7 @@
  * the module's configuration page.
  *
  * TODO: Measure the effectiveness of both of these algorithms.  Pick the best,
- * or come up with something better.  Ultimately this should not be a 
+ * or come up with something better.  Ultimately this should not be a
  * configurable option.
  */
 function spam_calculate_probabilities($since = 0) {
@@ -1452,7 +1449,7 @@
  * based on the list of spamm comments and nodes in found in their respective
  * tables.  This function is useful for three reasons: 1) if upgrading to a new
  * version of the tokenizer, you need to resynchronize your known tokens, 2)
- * if your spam filter seems to be well trained, and you want to cleanly 
+ * if your spam filter seems to be well trained, and you want to cleanly
  * measure its performance, and 3) if testing/developing an improved tokenizer
  * or spam probability logic.
  */