include/Scrape.php

   1 <?php
   2
   3 require_once('library/HTML5/Parser.php');
   4
   5 if(! function_exists('scrape_dfrn')) {
   6 function scrape_dfrn($url) {
   7
   8         $a = get_app();
   9
  10         $ret = array();
  11
  12         logger('scrape_dfrn: url=' . $url);
  13
  14         $s = fetch_url($url);
  15
  16         if(! $s)
  17                 return $ret;
  18
  19         $headers = $a->get_curl_headers();
  20         logger('scrape_dfrn: headers=' . $headers, LOGGER_DEBUG);
  21
  22
  23         $lines = explode("\n",$headers);
  24         if(count($lines)) {
  25                 foreach($lines as $line) {
  26                         // don't try and run feeds through the html5 parser
  27                         if(stristr($line,'content-type:') && ((stristr($line,'application/atom+xml')) || (stristr($line,'application/rss+xml'))))
  28                                 return ret;
  29                 }
  30         }
  31
  32
  33         $dom = HTML5_Parser::parse($s);
  34
  35         if(! $dom)
  36                 return $ret;
  37
  38         $items = $dom->getElementsByTagName('link');
  39
  40         // get DFRN link elements
  41
  42         foreach($items as $item) {
  43                 $x = $item->getAttribute('rel');
  44                 if(($x === 'alternate') && ($item->getAttribute('type') === 'application/atom+xml'))
  45                         $ret['feed_atom'] = $item->getAttribute('href');
  46                 if(substr($x,0,5) == "dfrn-")
  47                         $ret[$x] = $item->getAttribute('href');
  48                 if($x === 'lrdd') {
  49                         $decoded = urldecode($item->getAttribute('href'));
  50                         if(preg_match('/acct:([^@]*)@/',$decoded,$matches))
  51                                 $ret['nick'] = $matches[1];
  52                 }
  53         }
  54
  55         // Pull out hCard profile elements
  56
  57         $items = $dom->getElementsByTagName('*');
  58         foreach($items as $item) {
  59                 if(attribute_contains($item->getAttribute('class'), 'vcard')) {
  60                         $level2 = $item->getElementsByTagName('*');
  61                         foreach($level2 as $x) {
  62                                 if(attribute_contains($x->getAttribute('class'),'fn'))
  63                                         $ret['fn'] = $x->textContent;
  64                                 if(attribute_contains($x->getAttribute('class'),'photo'))
  65                                         $ret['photo'] = $x->getAttribute('src');
  66                                 if(attribute_contains($x->getAttribute('class'),'key'))
  67                                         $ret['key'] = $x->textContent;
  68                         }
  69                 }
  70         }
  71
  72         return $ret;
  73 }}
  74
  75
  76
  77
  78
  79
  80 if(! function_exists('validate_dfrn')) {
  81 function validate_dfrn($a) {
  82         $errors = 0;
  83         if(! x($a,'key'))
  84                 $errors ++;
  85         if(! x($a,'dfrn-request'))
  86                 $errors ++;
  87         if(! x($a,'dfrn-confirm'))
  88                 $errors ++;
  89         if(! x($a,'dfrn-notify'))
  90                 $errors ++;
  91         if(! x($a,'dfrn-poll'))
  92                 $errors ++;
  93         return $errors;
  94 }}
  95
  96 if(! function_exists('scrape_meta')) {
  97 function scrape_meta($url) {
  98
  99         $a = get_app();
 100
 101         $ret = array();
 102
 103         logger('scrape_meta: url=' . $url);
 104
 105         $s = fetch_url($url);
 106
 107         if(! $s)
 108                 return $ret;
 109
 110         $headers = $a->get_curl_headers();
 111         logger('scrape_meta: headers=' . $headers, LOGGER_DEBUG);
 112
 113         $lines = explode("\n",$headers);
 114         if(count($lines)) {
 115                 foreach($lines as $line) {
 116                         // don't try and run feeds through the html5 parser
 117                         if(stristr($line,'content-type:') && ((stristr($line,'application/atom+xml')) || (stristr($line,'application/rss+xml'))))
 118                                 return ret;
 119                 }
 120         }
 121
 122
 123
 124         $dom = HTML5_Parser::parse($s);
 125
 126         if(! $dom)
 127                 return $ret;
 128
 129         $items = $dom->getElementsByTagName('meta');
 130
 131         // get DFRN link elements
 132
 133         foreach($items as $item) {
 134                 $x = $item->getAttribute('name');
 135                 if(substr($x,0,5) == "dfrn-")
 136                         $ret[$x] = $item->getAttribute('content');
 137         }
 138
 139         return $ret;
 140 }}
 141
 142
 143 if(! function_exists('scrape_vcard')) {
 144 function scrape_vcard($url) {
 145
 146         $a = get_app();
 147
 148         $ret = array();
 149
 150         logger('scrape_vcard: url=' . $url);
 151
 152         $s = fetch_url($url);
 153
 154         if(! $s)
 155                 return $ret;
 156
 157         $headers = $a->get_curl_headers();
 158         $lines = explode("\n",$headers);
 159         if(count($lines)) {
 160                 foreach($lines as $line) {
 161                         // don't try and run feeds through the html5 parser
 162                         if(stristr($line,'content-type:') && ((stristr($line,'application/atom+xml')) || (stristr($line,'application/rss+xml'))))
 163                                 return ret;
 164                 }
 165         }
 166
 167         $dom = HTML5_Parser::parse($s);
 168
 169         if(! $dom)
 170                 return $ret;
 171
 172         // Pull out hCard profile elements
 173
 174         $items = $dom->getElementsByTagName('*');
 175         foreach($items as $item) {
 176                 if(attribute_contains($item->getAttribute('class'), 'vcard')) {
 177                         $level2 = $item->getElementsByTagName('*');
 178                         foreach($level2 as $x) {
 179                                 if(attribute_contains($x->getAttribute('class'),'fn'))
 180                                         $ret['fn'] = $x->textContent;
 181                                 if((attribute_contains($x->getAttribute('class'),'photo'))
 182                                         || (attribute_contains($x->getAttribute('class'),'avatar')))
 183                                         $ret['photo'] = $x->getAttribute('src');
 184                                 if((attribute_contains($x->getAttribute('class'),'nickname'))
 185                                         || (attribute_contains($x->getAttribute('class'),'uid')))
 186                                         $ret['nick'] = $x->textContent;
 187                         }
 188                 }
 189         }
 190
 191         return $ret;
 192 }}
 193
 194
 195 if(! function_exists('scrape_feed')) {
 196 function scrape_feed($url) {
 197
 198         $a = get_app();
 199
 200         $ret = array();
 201         $s = fetch_url($url);
 202
 203         if(! $s)
 204                 return $ret;
 205
 206         $headers = $a->get_curl_headers();
 207         logger('scrape_feed: headers=' . $headers, LOGGER_DEBUG);
 208
 209         $lines = explode("\n",$headers);
 210         if(count($lines)) {
 211                 foreach($lines as $line) {
 212                         if(stristr($line,'content-type:')) {
 213                                 if(stristr($line,'application/atom+xml') || stristr($s,'<feed')) {
 214                                         $ret['feed_atom'] = $url;
 215                                         return $ret;
 216                                 }
 217                                 if(stristr($line,'application/rss+xml') || stristr($s,'<rss')) {
 218                                         $ret['feed_rss'] = $url;
 219                                         return $ret;
 220                                 }
 221                         }
 222                 }
 223         }
 224
 225         $dom = HTML5_Parser::parse($s);
 226
 227         if(! $dom)
 228                 return $ret;
 229
 230
 231         $items = $dom->getElementsByTagName('img');
 232
 233         // get img elements (twitter)
 234
 235         if($items) {
 236                 foreach($items as $item) {
 237                         $x = $item->getAttribute('id');
 238                         if($x === 'profile-image') {
 239                                 $ret['photo'] = $item->getAttribute('src');
 240                         }
 241                 }
 242         }
 243
 244
 245         $head = $dom->getElementsByTagName('base');
 246         if($head) {
 247                 foreach($head as $head0) {
 248                         $basename = $head0->getAttribute('href');
 249                         break;
 250                 }
 251         }
 252         if(! $basename)
 253                 $basename = substr($url,0,strrpos($url,'/')) . '/';
 254
 255         $items = $dom->getElementsByTagName('link');
 256
 257         // get Atom/RSS link elements, take the first one of either.
 258
 259         if($items) {
 260                 foreach($items as $item) {
 261                         $x = $item->getAttribute('rel');
 262                         if(($x === 'alternate') && ($item->getAttribute('type') === 'application/atom+xml')) {
 263                                 if(! x($ret,'feed_atom'))
 264                                         $ret['feed_atom'] = $item->getAttribute('href');
 265                         }
 266                         if(($x === 'alternate') && ($item->getAttribute('type') === 'application/rss+xml')) {
 267                                 if(! x($ret,'feed_rss'))
 268                                         $ret['feed_rss'] = $item->getAttribute('href');
 269                         }
 270                 }
 271         }
 272
 273         // Drupal and perhaps others only provide relative URL's. Turn them into absolute.
 274
 275         if(x($ret,'feed_atom') && (! strstr($ret['feed_atom'],'://')))
 276                 $ret['feed_atom'] = $basename . $ret['feed_atom'];
 277         if(x($ret,'feed_rss') && (! strstr($ret['feed_rss'],'://')))
 278                 $ret['feed_rss'] = $basename . $ret['feed_rss'];
 279
 280         return $ret;
 281 }}
 282
 283
 284 function probe_url($url) {
 285         require_once('include/email.php');
 286
 287         $result = array();
 288
 289         if(! $url)
 290                 return $result;
 291
 292         $diaspora = false;
 293         $email_conversant = false;
 294
 295         $twitter = ((strpos($url,'twitter.com') !== false) ? true : false);
 296
 297         if(! $twitter) {
 298                 $links = lrdd($url);
 299
 300                 if(count($links)) {
 301                         logger('probe_url: found lrdd links: ' . print_r($links,true), LOGGER_DATA);
 302                         foreach($links as $link) {
 303                                 if($link['@attributes']['rel'] === NAMESPACE_ZOT)
 304                                         $zot = unamp($link['@attributes']['href']);
 305                                 if($link['@attributes']['rel'] === NAMESPACE_DFRN)
 306                                         $dfrn = unamp($link['@attributes']['href']);
 307                                 if($link['@attributes']['rel'] === 'salmon')
 308                                         $notify = unamp($link['@attributes']['href']);
 309                                 if($link['@attributes']['rel'] === NAMESPACE_FEED)
 310                                         $poll = unamp($link['@attributes']['href']);
 311                                 if($link['@attributes']['rel'] === 'http://microformats.org/profile/hcard')
 312                                         $hcard = unamp($link['@attributes']['href']);
 313                                 if($link['@attributes']['rel'] === 'http://webfinger.net/rel/profile-page')
 314                                         $profile = unamp($link['@attributes']['href']);
 315                                 if($link['@attributes']['rel'] === 'http://joindiaspora.com/seed_location')
 316                                         $diaspora = true;
 317                         }
 318
 319                         // Status.Net can have more than one profile URL. We need to match the profile URL
 320                         // to a contact on incoming messages to prevent spam, and we won't know which one
 321                         // to match. So in case of two, one of them is stored as an alias. Only store URL's
 322                         // and not webfinger user@host aliases. If they've got more than two non-email style
 323                         // aliases, let's hope we're lucky and get one that matches the feed author-uri because
 324                         // otherwise we're screwed.
 325
 326                         foreach($links as $link) {
 327                                 if($link['@attributes']['rel'] === 'alias') {
 328                                         if(strpos($link['@attributes']['href'],'@') === false) {
 329                                                 if(isset($profile)) {
 330                                                         if($link['@attributes']['href'] !== $profile)
 331                                                                 $alias = unamp($link['@attributes']['href']);
 332                                                 }
 333                                                 else
 334                                                         $profile = unamp($link['@attributes']['href']);
 335                                         }
 336                                 }
 337                         }
 338                 }
 339                 else {
 340
 341                         // Check email
 342
 343                         $orig_url = $url;
 344                         if((strpos($orig_url,'@')) && validate_email($orig_url)) {
 345                                 $x = q("SELECT `prvkey` FROM `user` WHERE `uid` = %d LIMIT 1",
 346                                         intval(local_user())
 347                                 );
 348                                 $r = q("SELECT * FROM `mailacct` WHERE `uid` = %d AND `server` != '' LIMIT 1",
 349                                         intval(local_user())
 350                                 );
 351                                 if(count($x) && count($r)) {
 352                                     $mailbox = construct_mailbox_name($r[0]);
 353                                         $password = '';
 354                                         openssl_private_decrypt(hex2bin($r[0]['pass']),$password,$x[0]['prvkey']);
 355                                         $mbox = email_connect($mailbox,$r[0]['user'],$password);
 356                                         unset($password);
 357                                 }
 358                                 if($mbox) {
 359                                         $msgs = email_poll($mbox,$orig_url);
 360                                         if(count($msgs)) {
 361                                                 $addr = $orig_url;
 362                                                 $network = NETWORK_MAIL;
 363                                                 $name = substr($url,0,strpos($url,'@'));
 364                                                 $profile = 'http://' . substr($url,strpos($url,'@')+1);
 365                                                 // fix nick character range
 366                                                 $vcard = array('fn' => $name, 'nick' => $name, 'photo' => gravatar_img($url));
 367                                                 $notify = 'smtp ' . random_string();
 368                                                 $poll = 'email ' . random_string();
 369                                                 $priority = 0;
 370                                                 $x = email_msg_meta($mbox,$msgs[0]);
 371                                                 if(stristr($x->from,$orig_url))
 372                                                         $adr = imap_rfc822_parse_adrlist($x->from,'');
 373                                                 elseif(stristr($x->to,$orig_url))
 374                                                         $adr = imap_rfc822_parse_adrlist($x->to,'');
 375                                                 if(isset($adr) && strlen($adr[0]->personal))
 376                                                         $vcard['fn'] = notags($adr[0]->personal);
 377                                         }
 378                                         imap_close($mbox);
 379                                 }
 380                         }
 381                 }
 382         }
 383
 384         if(strlen($zot)) {
 385                 $s = fetch_url($zot);
 386                 if($s) {
 387                         $j = json_decode($s);
 388                         if($j) {
 389                                 $network = NETWORK_ZOT;
 390                                 $vcard   = array(
 391                                         'fn'    => $j->fullname,
 392                                         'nick'  => $j->nickname,
 393                                         'photo' => $j->photo
 394                                 );
 395                                 $profile  = $j->url;
 396                                 $notify   = $j->post;
 397                                 $pubkey   = $j->pubkey;
 398                                 $poll     = 'N/A';
 399                         }
 400                 }
 401         }
 402
 403         if(strlen($dfrn)) {
 404                 $ret = scrape_dfrn($dfrn);
 405                 if(is_array($ret) && x($ret,'dfrn-request')) {
 406                         $network = NETWORK_DFRN;
 407                         $request = $ret['dfrn-request'];
 408                         $confirm = $ret['dfrn-confirm'];
 409                         $notify  = $ret['dfrn-notify'];
 410                         $poll    = $ret['dfrn-poll'];
 411                 }
 412         }
 413
 414         if($network !== NETWORK_ZOT && $network !== NETWORK_DFRN && $network !== NETWORK_MAIL) {
 415                 $network  = NETWORK_OSTATUS;
 416                 $priority = 0;
 417
 418                 if($hcard) {
 419                         $vcard = scrape_vcard($hcard);
 420
 421                         // Google doesn't use absolute url in profile photos
 422
 423                         if((x($vcard,'photo')) && substr($vcard['photo'],0,1) == '/') {
 424                                 $h = @parse_url($hcard);
 425                                 if($h)
 426                                         $vcard['photo'] = $h['scheme'] . '://' . $h['host'] . $vcard['photo'];
 427                         }
 428
 429                         logger('probe_url: scrape_vcard: ' . print_r($vcard,true), LOGGER_DATA);
 430                 }
 431
 432                 if(! $profile) {
 433                         if($diaspora)
 434                                 $profile = $hcard;
 435                         else
 436                                 $profile = $url;
 437                 }
 438
 439                 if($twitter) {
 440                         logger('twitter: setup');
 441                         $tid = basename($url);
 442                         $tapi = 'https://api.twitter.com/1/statuses/user_timeline.rss';
 443                         if(intval($tid))
 444                                 $poll = $tapi . '?user_id=' . $tid;
 445                         else
 446                                 $poll = $tapi . '?screen_name=' . $tid;
 447                         $profile = 'http://twitter.com/#!/' . $tid;
 448                 }
 449
 450                 if(! x($vcard,'fn'))
 451                         if(x($vcard,'nick'))
 452                                 $vcard['fn'] = $vcard['nick'];
 453
 454
 455                 if(((! isset($vcard)) && (! $poll)) || ($twitter)) {
 456
 457                         $feedret = scrape_feed($url);
 458                         logger('probe_url: scrape_feed returns: ' . print_r($feedret,true), LOGGER_DATA);
 459                         if(count($feedret) && ($feedret['feed_atom'] || $feedret['feed_rss'])) {
 460                                 $poll = ((x($feedret,'feed_atom')) ? unamp($feedret['feed_atom']) : unamp($feedret['feed_rss']));
 461                                 $vcard = array();
 462                         }
 463
 464                         if(x($feedret,'photo'))
 465                                 $vcard['photo'] = $feedret['photo'];
 466                         require_once('library/simplepie/simplepie.inc');
 467                     $feed = new SimplePie();
 468                         $xml = fetch_url($poll);
 469
 470                         logger('probe_url: fetch feed: ' . $poll . ' returns: ' . $xml, LOGGER_DATA);
 471                         $a = get_app();
 472
 473                         logger('probe_url: scrape_feed: headers: ' . $a->get_curl_headers(), $LOGGER_DATA);
 474
 475                         $feed->set_raw_data($xml);
 476
 477                     $feed->init();
 478                         if($feed->error())
 479                                 logger('probe_url: scrape_feed: Error parsing XML: ' . $feed->error());
 480
 481                         if(! x($vcard,'photo'))
 482                                 $vcard['photo'] = $feed->get_image_url();
 483                         $author = $feed->get_author();
 484                         if($author) {
 485                                 $vcard['fn'] = unxmlify(trim($author->get_name()));
 486                                 if(! $vcard['fn'])
 487                                         $vcard['fn'] = trim(unxmlify($author->get_email()));
 488                                 if(strpos($vcard['fn'],'@') !== false)
 489                                         $vcard['fn'] = substr($vcard['fn'],0,strpos($vcard['fn'],'@'));
 490                                 $email = unxmlify($author->get_email());
 491                                 if(! $vcard['photo']) {
 492                                         $rawtags = $feed->get_feed_tags( SIMPLEPIE_NAMESPACE_ATOM_10, 'author');
 493                                 if($rawtags) {
 494                                                 $elems = $rawtags[0]['child'][SIMPLEPIE_NAMESPACE_ATOM_10];
 495                                                 if((x($elems,'link')) && ($elems['link'][0]['attribs']['']['rel'] === 'photo'))
 496                                                         $vcard['photo'] = $elems['link'][0]['attribs']['']['href'];
 497                                 }
 498                                 }
 499                         }
 500                         else {
 501                                 $item = $feed->get_item(0);
 502                                 if($item) {
 503                                         $author = $item->get_author();
 504                                         if($author) {
 505                                                 $vcard['fn'] = trim(unxmlify($author->get_name()));
 506                                                 if(! $vcard['fn'])
 507                                                         $vcard['fn'] = trim(unxmlify($author->get_email()));
 508                                                 if(strpos($vcard['fn'],'@') !== false)
 509                                                         $vcard['fn'] = substr($vcard['fn'],0,strpos($vcard['fn'],'@'));
 510                                                 $email = unxmlify($author->get_email());
 511                                         }
 512                                         if(! $vcard['photo']) {
 513                                                 $rawmedia = $item->get_item_tags('http://search.yahoo.com/mrss/','thumbnail');
 514                                                 if($rawmedia && $rawmedia[0]['attribs']['']['url'])
 515                                                         $vcard['photo'] = unxmlify($rawmedia[0]['attribs']['']['url']);
 516                                         }
 517                                         if(! $vcard['photo']) {
 518                                                 $rawtags = $item->get_item_tags( SIMPLEPIE_NAMESPACE_ATOM_10, 'author');
 519                                         if($rawtags) {
 520                                                         $elems = $rawtags[0]['child'][SIMPLEPIE_NAMESPACE_ATOM_10];
 521                                                         if((x($elems,'link')) && ($elems['link'][0]['attribs']['']['rel'] === 'photo'))
 522                                                                 $vcard['photo'] = $elems['link'][0]['attribs']['']['href'];
 523                                         }
 524                                         }
 525                                 }
 526                         }
 527                         if((! $vcard['photo']) && strlen($email))
 528                                 $vcard['photo'] = gravatar_img($email);
 529                         if($poll === $profile)
 530                                 $lnk = $feed->get_permalink();
 531                         if(isset($lnk) && strlen($lnk))
 532                                 $profile = $lnk;
 533
 534                         if(! (x($vcard,'fn')))
 535                                 $vcard['fn'] = notags($feed->get_title());
 536                         if(! (x($vcard,'fn')))
 537                                 $vcard['fn'] = notags($feed->get_description());
 538
 539                         if(strpos($vcard['fn'],'Twitter / ') !== false) {
 540                                 $vcard['fn'] = substr($vcard['fn'],strpos($vcard['fn'],'/')+1);
 541                                 $vcard['fn'] = trim($vcard['fn']);
 542                         }
 543                         if(! x($vcard,'nick')) {
 544                                 $vcard['nick'] = strtolower(notags(unxmlify($vcard['fn'])));
 545                                 if(strpos($vcard['nick'],' '))
 546                                         $vcard['nick'] = trim(substr($vcard['nick'],0,strpos($vcard['nick'],' ')));
 547                         }
 548                         $network = 'feed';
 549                         $priority = 2;
 550                 }
 551         }
 552
 553         if(! x($vcard,'photo')) {
 554                 $a = get_app();
 555                 $vcard['photo'] = $a->get_baseurl() . '/images/default-profile.jpg' ;
 556         }
 557         $vcard['fn'] = notags($vcard['fn']);
 558         $vcard['nick'] = notags($vcard['nick']);
 559
 560
 561         $result['name'] = $vcard['fn'];
 562         $result['nick'] = $vcard['nick'];
 563         $result['url'] = $profile;
 564         $result['addr'] = $addr;
 565         $result['notify'] = $notify;
 566         $result['poll'] = $poll;
 567         $result['request'] = $request;
 568         $result['confirm'] = $confirm;
 569         $result['photo'] = $vcard['photo'];
 570         $result['priority'] = $priority;
 571         $result['network'] = $network;
 572         $result['alias'] = $alias;
 573         $result['pubkey'] = $pubkey;
 574
 575         logger('probe_url: ' . print_r($result,true), LOGGER_DEBUG);
 576
 577         return $result;
 578 }