private static function _markTicketAs($ticket_id, $spam = true)
{
// [TODO] Make sure we can't retrain tickets which are already spam trained
// [TODO] This is a performance killer
$ticket = DAO_Ticket::getTicket($ticket_id);
if ($ticket->spam_training != CerberusTicketSpamTraining::BLANK) {
return TRUE;
}
// pull up text of first ticket message
// [TODO] This is a performance killer
$first_message = DAO_Ticket::getMessage($ticket->first_message_id);
if (empty($first_message)) {
return FALSE;
}
// [TODO] This is a performance killer
$headers = DAO_MessageHeader::getAll($first_message->id);
// Pass text to analyze() to get back interesting words
$content = '';
if (!empty($ticket->subject)) {
// SplitCamelCapsSubjects
$hits = preg_split("{(?<=[a-z])(?=[A-Z])}x", $ticket->subject);
if (is_array($hits) && !empty($hits)) {
$content .= implode(' ', $hits);
}
}
$content .= ' ' . $first_message->getContent();
if (strlen($content) > self::MAX_BODY_LENGTH) {
$content = substr($content, 0, strrpos(substr($content, 0, self::MAX_BODY_LENGTH), ' '));
}
$words = self::processText($content);
// Train interesting words as spam/notspam
// $out = self::_calculateSpamProbability($words);
// self::_trainWords($out['words'],$spam);
self::_trainWords($words, $spam);
// [TODO] Testing, train all words
// Increase the bayes_stats spam or notspam total count by 1
// [TODO] This is a performance killer (could be done in batches)
if ($spam) {
DAO_Bayes::addOneToSpamTotal();
DAO_Address::addOneToSpamTotal($ticket->first_wrote_address_id);
} else {
DAO_Bayes::addOneToNonSpamTotal();
DAO_Address::addOneToNonSpamTotal($ticket->first_wrote_address_id);
}
// Forced training should leave a cache of 0.0001 or 0.9999 on the ticket table
$fields = array('spam_score' => $spam ? 0.9999 : 0.0001, 'spam_training' => $spam ? CerberusTicketSpamTraining::SPAM : CerberusTicketSpamTraining::NOT_SPAM);
DAO_Ticket::updateTicket($ticket_id, $fields);
return TRUE;
}