scottreynolds/modules/cre.module

<?php
// $Id:$
/* 
   A generalized recommendation engine. Depends on votingapi.module. see README.txt
*/

function cre_help($section) {
  switch ($section) {
    case 'admin/modules#description':
      return t("Provides api access any votable content used in votingapi. This is the backend engine");
    case 'admin/settings/cre':
      return t("Provides an recommendataion api to access recommended content");
  }
}
/*
 *  Creates or updates a record in cre_similarity_matrix. Maintains the sum and count
 *  Therefore, the avgerage difference between the two pieces of content can be calculated
 *
 * @param $userID
 * The uid of the vote
 *
 * @param $itemID
 * content_id of the new vote
 *
 * @return
 * NONE
 *  
*/
function cre_modify_avg_difference($userID,$itemID)
{
	// switch globals on database type!!!
	// Run my magik! put the values in cre_similarity_matrix table
   
   $db_result = db_query("SELECT DISTINCT r.content_id, r2.value-r.value as rating_difference, r.content_type as content_type1, r2.content_type as content_type2
   FROM {votingapi_vote} r, {votingapi_vote} r2
   WHERE r.uid=%d AND r2.content_id=%d AND r2.uid=r.uid", $userID, $itemID);
   if(!$db_result)
   {
   	drupal_set_message(t('ERROR'));
   }
   else
   {
   	while($row = db_fetch_array($db_result))
   	{
   	   $other_node=$row["content_id"];
   	   $rating_difference=$row["rating_difference"];
   	   
   	   // check to see if the pair of content ($itemID and $other_node) are already in the cre_similarity_matrix table
   	   if(db_num_rows(db_query("SELECT content_id1 from {cre_similarity_matrix} WHERE content_id1=%d AND content_id2=%d",$itemID,$other_node)) > 0)
   	   {
   	      //update the two rows
   	      //drupal_set_message(t('updated!'));
   	      db_query("UPDATE {cre_similarity_matrix} SET count=count+1, sum=sum+%d 
   	      WHERE content_id1=%d AND content_id2=%d", $rating_difference, $itemID, $other_node);
   	      // update the second row ONLY if the two nodes are different
   	      if($itemID != $other_node)
   	      {
   	         db_query("UPDATE {cre_similarity_matrix} SET count=count+1, sum=sum-%d 
   	         WHERE content_id1=%d AND content_id2=%d", $rating_difference, $other_node, $itemID);
   	      }
   	   }
   	   // if this is  a new 'pairing' create two rows
   	   else
   	   {
   	      $content_type1 = $row['content_type1'];
   	      $content_type2 = $row['content_type2'];
   	      // insert 2 rows only if the nodes are different
   	      db_query("INSERT INTO {cre_similarity_matrix} VALUES (%d,%d,'%s','%s',1, %d)",$itemID,$other_node,$content_type1,$content_type2,$rating_difference);
   	      
   	      if($itemID != $other_node)
   	      {
   	         db_query("INSERT INTO {cre_similarity_matrix} VALUES(%d, %d,'%s','%s',1,%d)",$other_node,$itemID,$content_type2,$content_type1,-$rating_difference); 
   	      }
   	   }
   	}
   }
}
  
/*
 * Returns an array of the top n content
 *
 * @param $uid 
 * user id for the personalized recommendation
 *
 * @param $n default 10
 * specifies the number of contents to return
 *
 * @param $target_type
 * the type of content that the calling function would like
 *
 * @param $tag default 'vote'
 * same as in votingapi. See votingapi
 *
 * @param $reference_type default NULL
 * When set only those votes whos content_type is equal to this
 * will be considered for the recommendations.
 *
 * @return
 * returns an array of objs with content_id and score
*/
function cre_top($uid, $n=10, $target_type = 'node', $tag = 'vote',$reference_type = NULL)
{
  
   if(isset($reference_type)) {
      $db_result = db_query("SELECT d.content_id1 as 'content_id', sum(d.sum+d.count*r.value)/sum(d.count) as 'score'
      FROM {votingapi_vote} r, {cre_similarity_matrix} d
      WHERE r.uid=%d AND d.content_id1<>r.content_id AND d.content_id2=r.content_id AND d.content_type1='%s' AND d.content_type2='%s' AND r.tag='%s'
      GROUP BY d.content_id1 ORDER BY score DESC",$uid,$target_type,$reference_type, $tag);
      
   }
   else
   {
      $db_result = db_query("SELECT d.content_id1 as 'content_id', sum(d.sum+d.count*r.value)/sum(d.count) as 'score'
      FROM {votingapi_vote} r, {cre_similarity_matrix} d
      WHERE r.uid=%d AND d.content_id1<>r.content_id AND d.content_id2=r.content_id AND d.content_type1='%s' AND d.content_type2=d.content_type1 AND r.tag='%s'
      GROUP BY d.content_id1 ORDER BY score DESC",$uid,$target_type,$tag);
   }
   
   $count = 0;

   while($matrixobj = db_fetch_object($db_result)) {
      if($count >= $n) {
         break;
      }
      // do a check to see if the content was rated by the uid
      $content_id = $matrixobj->content_id;
      $user_check_result = votingapi_get_vote($target_type, $content_id,'percent', 'vote', $uid);
      
      if(!$user_check_result) {
         $return_value[] = $matrixobj;
         $count++;
     }
   }
   if(isset($return_value)) {
      return $return_value;   
   }
   else
   {
      // do something special when user has rated all nodes
      // return top rated nodes?
      //return "No nodes";
   }
}

/*
 * Returns an array of content objs
 * 
 * @param $n
 * The number of objs to return
 * 
 * @param $nid
 * content_id that the return objects will be similar too
 *
 * @param $target_type
 * Type of content to return from this function
 *
 * @param $tag
 * votingapi tag. See Votingapi.module
 *
 * @param $reference_type
 * specifies the type of content that will be only be used to determine the revelenance of
 * a certain piece of content
 *
 * @return
 * $return_value[] is an array of objs that have two fields, content_id and average.
 *
*/

function cre_similar($n,$nid,$target_type,$tag='vote',$reference_type = NULL) // reference_type is sets which content_types are allowed to be reference when determining score
{
   // need to allow for reference_type to be an array!

   if(isset($reference_type)) {
      $sql_result = db_query("SELECT d.content_id2 as 'content_id', (d.sum / d.count) AS 'average'
      FROM {cre_similarity_matrix} d
      WHERE d.content_id1=%d AND d.content_id2 <> %d AND content_type1='%s' AND content_type2='%s'
      ORDER BY (sum/count) DESC LIMIT %d",$nid,$nid, $reference_type,$target_type,$n);
   }
   else {
      $sql_result = db_query("SELECT d.content_id2 as 'content_id', (d.sum / d.count) AS 'average' 
      FROM {cre_similarity_matrix} d
      WHERE d.content_id1=%d AND d.content_id2 <> %d AND content_type2='%s'
      ORDER BY (sum/count) DESC LIMIT %d",$nid,$nid,$target_type,$n);
   }
   while($recommend_obj = db_fetch_object($sql_result)) {
      $return_value[] = $recommend_obj;
   }
   return $return_value;

}

/*
 * implementation of votingapi's hook_insert()
*/

function cre_votingapi_insert(&$vote)
{
		// run my magik!
		// $vote obj has all the fields for a vote record! extract uid, nid
		cre_modify_avg_difference($vote->uid,$vote->content_id);
		//drupal_set_message(t('function Called'));
	return;
}

/*
 * implementation of votingapi's hook_update
*/

function cre_votingapi_update(&$vote,$new_value)
{
   // when a vote is updated need to just adjust sum NOT COUNT!!
   // for all records in cre_similarity_matrix where content_id1 or content_id2 equals vote->content_id
   // and the adjustment to sum MUST BE THE RATING_DIFFERENCE

   $rating_difference = $new_value-$vote->value;
   //drupal_set_message(t('difference = ' . $rating_difference));
      
   db_query("UPDATE {cre_similarity_matrix} SET sum=sum+%d 
   WHERE content_id1=%d", $rating_difference, $vote->content_id);
   
   db_query("UPDATE {cre_similarity_matrix} SET sum=sum-%d
   WHERE content_id2=%d",$rating_difference, $vote->content_id);
   
}
/*
 * implementation of votingapi's hook_delete
*/
function cre_votingapi_delete(&$vote)
{
   // TODO: modifiy the cre_similarity_matrix table by subratcting from count
   // and subtracting vote value from sum
   // OVER ALL REFERENCES TO THE CONTENT_ID
   // THINK !!! reverse of build!!
   $userID = $vote->uid;
   $itemID = $vote->content_id;
  $db_result = db_query("SELECT DISTINCT r.content_id, r2.value-r.value as rating_difference, r.content_type as content_type1, r2.content_type as content_type2
   FROM {votingapi_vote} r, {votingapi_vote} r2
   WHERE r.uid=%d AND r2.content_id=%d AND r2.uid=r.uid", $userID, $itemID);
   if(!$db_result)
   {
   	drupal_set_message(t('ERROR $userID, $itemID'));
   }
   else
   {
   	while($row = db_fetch_array($db_result))
   	{
   	   $other_node=$row["content_id"];
   	   $rating_difference=$row["rating_difference"];
   	   
   	  
   	   //update the two rows
   	   db_query("UPDATE {cre_similarity_matrix} SET count=count-1, sum=sum-%d 
   	   WHERE content_id1=%d AND content_id2=%d", $rating_difference, $itemID, $other_node);
   	   
   	   // update the second row ONLY if the two nodes are different
   	   if($itemID != $other_node)
   	   {
   	      db_query("UPDATE {cre_similarity_matrix} SET count=count-1, sum=sum+%d 
   	      WHERE content_id1=%d AND content_id2=%d", $rating_difference, $other_node, $itemID);
   	   }
      }
   }
}

/*
 * private function to install the module over existing voting data
*/

function _cre_load_all_diff_avg()
{
   // for every uid 
	$usersSQL = "SELECT DISTINCT uid from {votingapi_vote}";

	$db_result = db_query($usersSQL);
	if(!$db_result)
	{
	    drupal_set_message(t('NO USERS!!'));
	    return;
	}

	   while($row = db_fetch_array($db_result))
	   {
	      $userID = $row['uid'];
	      
	      // for every thing they rated
	      $vote_result = db_query("SELECT content_id
	      FROM {votingapi_vote}
	      WHERE uid=%d",$userID);
	      if(!$vote_result)
	      {
	         drupal_set_message(t('No votes have been placed yet'));
	         return;
	      }
	      while($voting_record = db_fetch_array($vote_result))
	      {      
            $content_id = $voting_record['content_id'];	         

	         // run the build function
            cre_modify_avg_difference($userID,$content_id);
	      }
	   }
}
1	<?php
2	// $Id:$
3	/*
4	A generalized recommendation engine. Depends on votingapi.module. see README.txt
5	*/
6
7	function cre_help($section) {
8	switch ($section) {
9	case 'admin/modules#description':
10	return t("Provides api access any votable content used in votingapi. This is the backend engine");
11	case 'admin/settings/cre':
12	return t("Provides an recommendataion api to access recommended content");
13	}
14	}
15	/*
16	* Creates or updates a record in cre_similarity_matrix. Maintains the sum and count
17	* Therefore, the avgerage difference between the two pieces of content can be calculated
18	*
19	* @param $userID
20	* The uid of the vote
21	*
22	* @param $itemID
23	* content_id of the new vote
24	*
25	* @return
26	* NONE
27	*
28	*/
29	function cre_modify_avg_difference($userID,$itemID)
30	{
31	// switch globals on database type!!!
32	// Run my magik! put the values in cre_similarity_matrix table
33
34	$db_result = db_query("SELECT DISTINCT r.content_id, r2.value-r.value as rating_difference, r.content_type as content_type1, r2.content_type as content_type2
35	FROM {votingapi_vote} r, {votingapi_vote} r2
36	WHERE r.uid=%d AND r2.content_id=%d AND r2.uid=r.uid", $userID, $itemID);
37	if(!$db_result)
38	{
39	drupal_set_message(t('ERROR'));
40	}
41	else
42	{
43	while($row = db_fetch_array($db_result))
44	{
45	$other_node=$row["content_id"];
46	$rating_difference=$row["rating_difference"];
47
48	// check to see if the pair of content ($itemID and $other_node) are already in the cre_similarity_matrix table
49	if(db_num_rows(db_query("SELECT content_id1 from {cre_similarity_matrix} WHERE content_id1=%d AND content_id2=%d",$itemID,$other_node)) > 0)
50	{
51	//update the two rows
52	//drupal_set_message(t('updated!'));
53	db_query("UPDATE {cre_similarity_matrix} SET count=count+1, sum=sum+%d
54	WHERE content_id1=%d AND content_id2=%d", $rating_difference, $itemID, $other_node);
55	// update the second row ONLY if the two nodes are different
56	if($itemID != $other_node)
57	{
58	db_query("UPDATE {cre_similarity_matrix} SET count=count+1, sum=sum-%d
59	WHERE content_id1=%d AND content_id2=%d", $rating_difference, $other_node, $itemID);
60	}
61	}
62	// if this is a new 'pairing' create two rows
63	else
64	{
65	$content_type1 = $row['content_type1'];
66	$content_type2 = $row['content_type2'];
67	// insert 2 rows only if the nodes are different
68	db_query("INSERT INTO {cre_similarity_matrix} VALUES (%d,%d,'%s','%s',1, %d)",$itemID,$other_node,$content_type1,$content_type2,$rating_difference);
69
70	if($itemID != $other_node)
71	{
72	db_query("INSERT INTO {cre_similarity_matrix} VALUES(%d, %d,'%s','%s',1,%d)",$other_node,$itemID,$content_type2,$content_type1,-$rating_difference);
73	}
74	}
75	}
76	}
77	}
78
79	/*
80	* Returns an array of the top n content
81	*
82	* @param $uid
83	* user id for the personalized recommendation
84	*
85	* @param $n default 10
86	* specifies the number of contents to return
87	*
88	* @param $target_type
89	* the type of content that the calling function would like
90	*
91	* @param $tag default 'vote'
92	* same as in votingapi. See votingapi
93	*
94	* @param $reference_type default NULL
95	* When set only those votes whos content_type is equal to this
96	* will be considered for the recommendations.
97	*
98	* @return
99	* returns an array of objs with content_id and score
100	*/
101	function cre_top($uid, $n=10, $target_type = 'node', $tag = 'vote',$reference_type = NULL)
102	{
103
104	if(isset($reference_type)) {
105	$db_result = db_query("SELECT d.content_id1 as 'content_id', sum(d.sum+d.count*r.value)/sum(d.count) as 'score'
106	FROM {votingapi_vote} r, {cre_similarity_matrix} d
107	WHERE r.uid=%d AND d.content_id1<>r.content_id AND d.content_id2=r.content_id AND d.content_type1='%s' AND d.content_type2='%s' AND r.tag='%s'
108	GROUP BY d.content_id1 ORDER BY score DESC",$uid,$target_type,$reference_type, $tag);
109
110	}
111	else
112	{
113	$db_result = db_query("SELECT d.content_id1 as 'content_id', sum(d.sum+d.count*r.value)/sum(d.count) as 'score'
114	FROM {votingapi_vote} r, {cre_similarity_matrix} d
115	WHERE r.uid=%d AND d.content_id1<>r.content_id AND d.content_id2=r.content_id AND d.content_type1='%s' AND d.content_type2=d.content_type1 AND r.tag='%s'
116	GROUP BY d.content_id1 ORDER BY score DESC",$uid,$target_type,$tag);
117	}
118
119	$count = 0;
120
121	while($matrixobj = db_fetch_object($db_result)) {
122	if($count >= $n) {
123	break;
124	}
125	// do a check to see if the content was rated by the uid
126	$content_id = $matrixobj->content_id;
127	$user_check_result = votingapi_get_vote($target_type, $content_id,'percent', 'vote', $uid);
128
129	if(!$user_check_result) {
130	$return_value[] = $matrixobj;
131	$count++;
132	}
133	}
134	if(isset($return_value)) {
135	return $return_value;
136	}
137	else
138	{
139	// do something special when user has rated all nodes
140	// return top rated nodes?
141	//return "No nodes";
142	}
143	}
144
145	/*
146	* Returns an array of content objs
147	*
148	* @param $n
149	* The number of objs to return
150	*
151	* @param $nid
152	* content_id that the return objects will be similar too
153	*
154	* @param $target_type
155	* Type of content to return from this function
156	*
157	* @param $tag
158	* votingapi tag. See Votingapi.module
159	*
160	* @param $reference_type
161	* specifies the type of content that will be only be used to determine the revelenance of
162	* a certain piece of content
163	*
164	* @return
165	* $return_value[] is an array of objs that have two fields, content_id and average.
166	*
167	*/
168
169	function cre_similar($n,$nid,$target_type,$tag='vote',$reference_type = NULL) // reference_type is sets which content_types are allowed to be reference when determining score
170	{
171	// need to allow for reference_type to be an array!
172
173	if(isset($reference_type)) {
174	$sql_result = db_query("SELECT d.content_id2 as 'content_id', (d.sum / d.count) AS 'average'
175	FROM {cre_similarity_matrix} d
176	WHERE d.content_id1=%d AND d.content_id2 <> %d AND content_type1='%s' AND content_type2='%s'
177	ORDER BY (sum/count) DESC LIMIT %d",$nid,$nid, $reference_type,$target_type,$n);
178	}
179	else {
180	$sql_result = db_query("SELECT d.content_id2 as 'content_id', (d.sum / d.count) AS 'average'
181	FROM {cre_similarity_matrix} d
182	WHERE d.content_id1=%d AND d.content_id2 <> %d AND content_type2='%s'
183	ORDER BY (sum/count) DESC LIMIT %d",$nid,$nid,$target_type,$n);
184	}
185	while($recommend_obj = db_fetch_object($sql_result)) {
186	$return_value[] = $recommend_obj;
187	}
188	return $return_value;
189
190	}
191
192	/*
193	* implementation of votingapi's hook_insert()
194	*/
195
196	function cre_votingapi_insert(&$vote)
197	{
198	// run my magik!
199	// $vote obj has all the fields for a vote record! extract uid, nid
200	cre_modify_avg_difference($vote->uid,$vote->content_id);
201	//drupal_set_message(t('function Called'));
202	return;
203	}
204
205	/*
206	* implementation of votingapi's hook_update
207	*/
208
209	function cre_votingapi_update(&$vote,$new_value)
210	{
211	// when a vote is updated need to just adjust sum NOT COUNT!!
212	// for all records in cre_similarity_matrix where content_id1 or content_id2 equals vote->content_id
213	// and the adjustment to sum MUST BE THE RATING_DIFFERENCE
214
215	$rating_difference = $new_value-$vote->value;
216	//drupal_set_message(t('difference = ' . $rating_difference));
217
218	db_query("UPDATE {cre_similarity_matrix} SET sum=sum+%d
219	WHERE content_id1=%d", $rating_difference, $vote->content_id);
220
221	db_query("UPDATE {cre_similarity_matrix} SET sum=sum-%d
222	WHERE content_id2=%d",$rating_difference, $vote->content_id);
223
224	}
225	/*
226	* implementation of votingapi's hook_delete
227	*/
228	function cre_votingapi_delete(&$vote)
229	{
230	// TODO: modifiy the cre_similarity_matrix table by subratcting from count
231	// and subtracting vote value from sum
232	// OVER ALL REFERENCES TO THE CONTENT_ID
233	// THINK !!! reverse of build!!
234	$userID = $vote->uid;
235	$itemID = $vote->content_id;
236	$db_result = db_query("SELECT DISTINCT r.content_id, r2.value-r.value as rating_difference, r.content_type as content_type1, r2.content_type as content_type2
237	FROM {votingapi_vote} r, {votingapi_vote} r2
238	WHERE r.uid=%d AND r2.content_id=%d AND r2.uid=r.uid", $userID, $itemID);
239	if(!$db_result)
240	{
241	drupal_set_message(t('ERROR $userID, $itemID'));
242	}
243	else
244	{
245	while($row = db_fetch_array($db_result))
246	{
247	$other_node=$row["content_id"];
248	$rating_difference=$row["rating_difference"];
249
250
251	//update the two rows
252	db_query("UPDATE {cre_similarity_matrix} SET count=count-1, sum=sum-%d
253	WHERE content_id1=%d AND content_id2=%d", $rating_difference, $itemID, $other_node);
254
255	// update the second row ONLY if the two nodes are different
256	if($itemID != $other_node)
257	{
258	db_query("UPDATE {cre_similarity_matrix} SET count=count-1, sum=sum+%d
259	WHERE content_id1=%d AND content_id2=%d", $rating_difference, $other_node, $itemID);
260	}
261	}
262	}
263	}
264
265	/*
266	* private function to install the module over existing voting data
267	*/
268
269	function _cre_load_all_diff_avg()
270	{
271	// for every uid
272	$usersSQL = "SELECT DISTINCT uid from {votingapi_vote}";
273
274	$db_result = db_query($usersSQL);
275	if(!$db_result)
276	{
277	drupal_set_message(t('NO USERS!!'));
278	return;
279	}
280
281	while($row = db_fetch_array($db_result))
282	{
283	$userID = $row['uid'];
284
285	// for every thing they rated
286	$vote_result = db_query("SELECT content_id
287	FROM {votingapi_vote}
288	WHERE uid=%d",$userID);
289	if(!$vote_result)
290	{
291	drupal_set_message(t('No votes have been placed yet'));
292	return;
293	}
294	while($voting_record = db_fetch_array($vote_result))
295	{
296	$content_id = $voting_record['content_id'];
297
298	// run the build function
299	cre_modify_avg_difference($userID,$content_id);
300	}
301	}
302	}