
<?php

    /////////////////////////////////////////////////////////

    //

    //  AHNAR CSA3004 APT University of Malta

    //  Ian Bugeja 2005

    //

    //////////////////////////////////////////////////////////

    define("AHNARLOC", "http://ahnar.iannet.org/ahnar.php?ahnarurl=");



    class HTML_Parser

    {



        public $stshlink = ''; //<link stylesheet tag

        private $domain = '';  //page domain

        public $conframes = false;  //if page is a frames page

        public $atagcount = 0;      //number of <a> tags

        public $words = '';         //page contents excluding tags

        public $ataglist;           //position of <a> tags

        public $pagetitle;          //page title

        public $middleatext;        //contains text in middle of a tag

        public $docutype;           //!DOCTYPE tag of html page

        public $bodytag;            //<body> tag

        private $bold = false;

        public $adest;              //<a> tags destination





        public function __construct($stlink, $dom)

        {

                $this->stshlink = $stlink;

                $siz = strlen($dom);

                if ($dom[$siz-1] == '/') $dom = substr($dom, 0, $siz-1);

                $this->domain = $dom;

                $this->atagcount = 0;

                $this->words = '';

                $this->pagetitle = '';

                $this->docutype = '';

                $this->bodytag = '';

                $this->bold = false;

                $this->ataglist = array();

                $this->middleatext = array();

        }





        private function id_tag($tag)    //identify tag (removing < and >)

        {

                $len = strlen($tag);

                $tag[0] = chr(32);

                $tag[$len-1] = chr(32);

                $tag = str_replace("\r", " ", $tag);

                $tag = str_replace("\n", " ", $tag);

                $tag = trim($tag);



                $tag = strtolower($tag);

                $pos = strpos($tag, ' ');



                if ($pos === false)

                {

                        return $tag;

                }

                else

                {

                        return substr($tag, 0, $pos);

                }

        }



        public function html_parse($code)

        {

            $newcontents = '';

            $subcontents = '';

            $boldcontents = '';

            $pos = -1;

            $len = strlen($code);

            $scriptparse = false;

            $titleparse = false;

            $midaflag = 0;

            $midatext = '';

            $aname = '';

            while ($pos < $len)

            {

                $pos++;

                //if (($code{$pos} == '<') && (($scriptparse == false) || (($code[$pos+2] == 's') || ($code[$pos+2] == 'S'))))

                if ($code[$pos] == '<')

                {

                    $tpos = $pos;

                    while (($code[$tpos] != '>') && ($tpos < $len)) //find end tag marker

                    {

                        $tpos++;

                        if ($code[$tpos] == '<') { $tpos--; break; } //in case of incorrect closed tag...

                    }

                    $tag = substr($code, $pos, $tpos-$pos+1); //get tag

                    //identify tag type

                    $idtag = $this->id_tag($tag);

                    if ($idtag[0] == '/')  //closing tag

                    {

                        $subcontents .= ' ';

                        if ($this->bold == true)

                        {

                                $subcontents .= ' '.$boldcontents . ' ' . $boldcontents;

                                $this->bold = false;

                        }

                    }

                    else

                    {

                        $bt = $this->checkBoldCSS($tag);

                        if ($bt !== false) $this->bold = $bt;

                        if ($this->bold == true) $boldcontents = '';

                    }



                    switch ($idtag)

                    {

                        case '!doctype' : $this->docutype = $tag; break;

                        case 'link': $this->stshlink .= $this->modify_src($tag, 'href', false); break;

                        case 'html': break;

                        case '/html': break;

                        case 'title':  if ($this->pagetitle == "") $titleparse = true; $newcontents .= "<!--"; break;

                        case '/title': $titleparse = false; $newcontents .= "-->"; break;

                        case 'head' : break;

                        case '/head' : break;

                        case 'meta' : $newcontents .= "<!--ahnar rem-meta-->";

                                        if (stripos($tag, "refresh") !== false) //not to remove meta refresh tag

                                        {

                                                $metaext1 = stripos($tag, "url"); echo $metaext1.'~';

                                                $metaext2 = stripos($tag, "content");  echo $metaext2.'~';

                                                $metaurl = substr($tag, $metaext2+8, $metaext1+4-($metaext2+8)); echo $metaurl;

                                                $tag = str_replace($metaurl, "", $tag);

                                                $tag = $this->modify_src($tag, "content", true);

                                                $tag = str_ireplace("content=", "content=".$metaurl, $tag);

                                                $newcontents .= $tag;

                                        }

                                        break;

                        case '/meta' : break;

                        case 'body' :   //$tag = $this->remAttrib($tag, 'onLoad');

                                        if ($this->getValue($tag, 'background') != "") $tag = $this->modify_src($tag, 'background', false);

                                      $this->bodytag = $tag;

                                      break;

                        case '/body' : break;

                        case 'frame' :  $this->conframes = true;  //mark page as a Framed page (parsed differently)

                                        $tag = $this->modify_src($tag, 'src', true);

                                        $newcontents .= $tag; break;

                        case 'iframe' : $tag = $this->modify_src($tag, 'src', true);

                                        $newcontents .= $tag; break;

                        case 'a' : if (count($this->ataglist) != count($this->middleatext))

                                   {

                                        echo count($this->ataglist).'|'.count($this->middleatext)."ERRORRRRRRRRRRRRRRR\r\n";

                                        while (count($this->ataglist) > count($this->middleatext))

                                              $this->middleatext[] = " ";

                                        while (count($this->ataglist) < count($this->middleatext))

                                              $this->ataglist[] = 0;

                                   }

                                   //if (count($this->ataglist) != count($this->middleatext)) echo "ERRORO BIGG\r\n";

                                   $this->atagcount++;

                                   $dest = $this->getValue($tag, "href");

                                   $this->adest[] = $dest;

                                   $tag = $this->modify_src($tag, 'href', true);

                                   $tag = $this->modify_src($tag, 'src', false);

                                   $midaflag = 0;

                                   $aname = $this->getValue($tag, "name");

                                   if($aname != "")

                                   {        //echo $aname." ";

                                        $midaflag = 3; //name atag (anchor position)

                                   }

                                   elseif ($dest == "")

                                   {

                                        $midaflag = 7;

                                   }

                                   elseif (stripos($tag, 'mailto', 1) === false)  //NOT EMAIL

                                   {      //echo '|'.$dest.'|';

                                        if ($this->idDestType($dest) < 50)

                                        {

                                                $midaflag = 1; $midatext = '';  //normal atag

                                        }

                                        else

                                                $midaflag = 4;

                                   }

                                   else

                                        $midaflag = 2; //email

                                   $this->ataglist[] = strlen($newcontents); //position of 'a' tag to know where abouts...

                                   $newcontents .= $tag;  //echo $tag.$dest.$midaflag.'|';

                                   break;



                        case '/a' :     $newcontents .= $tag;

                                        switch ($midaflag)

                                        {

                                                case 1 : $this->middleatext[] = trim(str_replace("&nbsp;", " ", strtolower($midatext))); break;

                                                case 2 : $this->middleatext[] = 'email';

                                                case 3 : $this->middleatext[] = 'achname-'.$aname; $aname = ''; break;

                                                case 4 : $this->middleatext[] = 'applic'; break;

                                                case 5 : $this->middleatext[] = 'img'; break;

                                                default : $this->middleatext[] = 'ahnoo'; break;

                                        }

                                        $midaflag = 0;

                                        break;



                        case 'img' :    $tag = $this->modify_src($tag, 'src', false);

                                        $newcontents .= $tag;  //echo $tag;

                                        $subcontents .= " ".$this->getValue($tag, "alt"); //add alt to text (used for keyword count)

                                        if ($midaflag > 0) $midaflag = 5; //notify image

                                        break;

                        case 'area' :   $tag = $this->modify_src($tag, 'href', true);

                                        $newcontents .= $tag;

                                        break;

                        case 'script' : $scriptparse = true;

                                        $tag = $this->modify_src($tag, 'src', false);

                                        $newcontents .= $tag;



                                        $pos = $tpos+1;

                                        while (true)

                                        {

                                                $tpos++;

                                                if (($code[$tpos] == '<') && ($code[$tpos+1] == '/') && (($code[$tpos+2] == 's') || ($code[$tpos+2] == 'S')) && (($code[$tpos+3] == 'c') || ($code[$tpos+3] == 'C')))

                                                        break;

                                        }

                                        $tpos--;

                                        $tag = substr($code, $pos, $tpos-$pos+1);

                                        $newcontents .= $tag;



                                        break;

                        case '/script' : $scriptparse = false;

                                        $newcontents .= $tag; break;

                        case 'style' :  $scriptparse = true;

                                        $newcontents .= $tag; break;

                        case '/style' : $scriptparse = false;

                                        $newcontents .= $tag; break;

                        case 'table' : if ($this->getValue($tag, 'background') != "") $tag = $this->modify_src($tag, 'background', false);

                                       $newcontents .= $tag; break;

                        case 'td' : if ($this->getValue($tag, 'background') != "") $tag = $this->modify_src($tag, 'background', false);

                                    $newcontents .= $tag; break;

                        case 'droplet' : $tag = $this->modify_src($tag, 'src', true);

                                         $newcontents .= $tag; break;

                        case 'form' : $tag = $this->modify_src($tag, 'action', true);

                                      $ax = strpos($tag, "ahnarurl=", 0);

                                      $ax2 = strpos($tag, "\"", $ax); //add a hidden field to send url to script

                                      if ($ax2 <= 0) $ax2 = strpos($tag, " ", $ax); if ($ax2 <= 0) $ax2 = strpos($tag, ">", $ax);

                                      $actionsrc = substr($tag,$ax+9, $ax2-$ax-9);

                                      $newcontents .= $tag; break;

                        case '/form' : $newcontents .= "<input type='hidden' name='ahnarurl' value='$actionsrc'></form>"; break;



                        case 'b' :

                        case 'em' :

                        case 'strong' :

                        case 'h1' : $this->bold = true;

                                    $newcontents .= $tag;

                                    $boldcontents = ''; break;



                        case 'param' :  $paramvalue = strtolower($this->getValue($tag, "name"));

                                        if ($paramvalue == "movie")

                                                $tag = $this->modify_src($tag, 'value', false);

                                       $newcontents .= $tag; break;

                        case 'embed' : $tag = $this->modify_src($tag, 'src', false);

                                        $newcontents .= $tag; break;

                        case 'applet' : if (stripos($tag, 'codebase') === false)

                                                $tag = str_ireplace("<applet ", "<applet codebase=\".\" ", $tag);

                                        $tag = $this->modify_src($tag, 'codebase', false);

                                        $newcontents .= $tag; break;

                        //Tags not above just add unmodified

                        default : $newcontents .= $tag; break;

                    }

                    ////

                    $pos = $tpos;

                }

                else

                {

                    $newcontents .= $code[$pos];    //text not consisting of a tag (body text)



                    if (ord($code[$pos]) > 30)

                    if ($scriptparse == false) $subcontents .= $code[$pos]; //subcontents (used for page keywords)



                    if ($titleparse) $this->pagetitle .= $code[$pos];  //page title



                    if ($midaflag == 1) $midatext .= $code[$pos]; //text between <a> tag ie actual link



                    if ($this->bold == true) $boldcontents .= $code[$pos]; //Bold text

                }

            }

              //echo $subcontents;

            $subcontents = str_replace("&", " &", $subcontents);

            $subcontents = str_replace(";", "; ", $subcontents);

            $subcontents = str_replace("&nbsp;", " ", $subcontents);

            $this->words = str_replace("  ", " ", $subcontents);

            return $newcontents;

        }





        private function modify_src($imgtag, $attrib, $ahnr)

        {

                $pos = 1;

                $imgtag = str_replace("\r", " ", $imgtag); $imgtag = str_replace("\n", " ", $imgtag); //remove new line

                $len = strlen($imgtag);

                $newtag = $imgtag;

                while (($pos !== false) && ($pos < $len))

                {

                        $pos = stripos($imgtag, ' '.$attrib, $pos+1);

                        if ($pos !== false)

                        {

                                $qpos = strpos($imgtag, '=', $pos+1);

                                $i = $qpos;

                                do { $i++; } while (($imgtag[$i] == chr(32)) || ($imgtag[$i] == chr(34)) || ($imgtag[$i] == chr(39)) );



                                if ($attrib == 'href')

                                {

                                        if (stripos($imgtag, 'mailto', $i) !== false) return $imgtag; //case of mailto href

                                        if (stripos($imgtag, 'javascript:', $i) !== false) return $imgtag; //case of javascript href

                                }



                                $httppos = stripos($imgtag, 'http://', $i);   if ($httppos-$i > 5) $httppos=0;

                                $httpspos = stripos($imgtag, 'https://', $i);  if ($httpspos-$i > 5) $httpspos=0;



                                if ($imgtag[$i] == '#')

                                        $newtag = $imgtag; //instead of return

                                elseif (($httppos > 0) || ($httpspos > 0))

                                {

                                        $newtag = ''; //TO CHECK--------------

                                        $newtag = substr($imgtag, 0, $i);

                                        if ($ahnr == true) $newtag .= AHNARLOC;

                                        $newtag .= substr($imgtag, $i, strlen($imgtag)-$i+1);

                                        $imgtag = $newtag; //instead of return

                                }

                                else

                                {

                                        $newtag = '';

                                        $newtag = substr($imgtag, 0, $i);

                                        if ($ahnr == true) $newtag .= AHNARLOC;

                                        if (($imgtag[$i] != '/') && ($imgtag[$i] != '.'))

                                                $newtag .= $this->domain;

                                        elseif ($imgtag[$i] == '/')

                                        {

                                                $slpos = stripos($this->domain, '/', 9);

                                                if ($slpos === false) $newtag .= $this->domain; else $newtag .= substr($this->domain, 0, $slpos);

                                        }

                                        elseif ($imgtag[$i] == '.')    ///??? CHECK

                                        {

                                                if ($imgtag[$i+1] == '.')

                                                        $slpos = strripos($this->domain, '/');

                                                else

                                                        $slpos = 0; //just for next line

                                                if ($slpos <= 6) $newtag .= $this->domain; else $newtag .= substr($this->domain, 0, $slpos);

                                        }

                                        while ($imgtag[$i] == '.') $i++;

                                        if ($imgtag[$i] != '/') $newtag .= '/';

                                        $newtag .= substr($imgtag, $i, strlen($imgtag)-$i+1);

                                        $imgtag = $newtag; //instead of return

                                }

                        }

                        $len = strlen($imgtag);

                }



                return $newtag;

        }



        private function checkBoldCSS($tag)  //returns true if style attribute present and bold is set

        {

                $stpos = stripos($tag, "style");

                if ($stpos === false) return;



                if (stripos($tag, "bold", $stpos + 5)) return true;



                return false;

        }





        private function getValue($atag, $attrib)  //returns destination of (href) <a> tag

        {

                $hrefpos = stripos($atag, $attrib);

                if ($hrefpos === false) return "";



                $epos = stripos($atag, "=", $hrefpos);

                if (($epos === false) || ($epos > $hrefpos+strlen($attrib)+3)) return "";

                $qpos = stripos($atag, '"', $epos);

                if (($qpos !== false) && ($qpos < $epos+3))

                        $fqpos = stripos($atag, '"', $qpos+1); //closing "

                else

                {

                        $qpos = stripos($atag, "'", $epos);



                        if (($qpos !== false) && ($qpos < $epos+3))

                                $fqpos = stripos($atag, "'", $qpos+1); //closing '

                        else

                        {

                                $qpos = $epos;

                                $fqpos = stripos($atag, " ", $qpos);

                                if ($fqpos === false) $fqpos = strlen($atag)-1;

                        }

                }



                if ($qpos+2 >= $fqpos) return "";



                return trim(substr($atag, $qpos+1, $fqpos-$qpos-1));

        }



        private function remAttrib($tag, $attrib)

        {

                //identical to above

                $attribpos = stripos($tag, $attrib);

                if ($attribpos === false) return $tag;



                $epos = stripos($tag, "=", $attribpos);

                if (($epos === false) || ($epos > $attribpos+strlen($attrib)+3)) return "";

                $qpos = stripos($tag, '"', $epos);

                if (($qpos !== false) && ($qpos < $epos+3))

                        $fqpos = stripos($tag, '"', $qpos+1); //closing "

                else

                {

                        $qpos = stripos($tag, "'", $epos);



                        if (($qpos !== false) && ($qpos < $epos+3))

                                $fqpos = stripos($tag, "'", $qpos+1); //closing '

                        else

                        {

                                $qpos = $epos;

                                $fqpos = stripos($tag, " ", $qpos);

                                if ($fqpos === false) $fqpos = strlen($tag)-1;

                        }

                }



                //addon

                $newtag = substr($tag, 0, $attribpos);

                $newtag .= substr($tag, $fqpos+1, strlen($tag)-$fqpos-1);



                return $newtag;

        }



        private function idDestType($path)

        {

                $extpos = strrpos($path, ".");

                if ($extpos === false) return 0; //not file



                $ext = substr($path, $extpos, 4);

                if ($ext == ".htm") return 1;

                if ($ext == ".pdf") return 3;

                if (($ext == ".jpg") || ($ext == ".gif") || ($ext == ".bmp") || ($ext == ".png") || ($ext == ".jpe"))

                        return 4;



                if ($ext == ".zip") return 55;

                if ($ext == ".rar") return 56;

                if ($ext == ".exe") return 57;



                return 1; //default

        }



    } //class



?>