inc/3rdparty/makefulltextfeed.php

   1 <?php
   2 // Full-Text RSS: Create Full-Text Feeds
   3 // Author: Keyvan Minoukadeh
   4 // Copyright (c) 2013 Keyvan Minoukadeh
   5 // License: AGPLv3
   6 // Version: 3.2
   7 // Date: 2013-05-13
   8 // More info: http://fivefilters.org/content-only/
   9 // Help: http://help.fivefilters.org
  10
  11 /*
  12 This program is free software: you can redistribute it and/or modify
  13 it under the terms of the GNU Affero General Public License as published by
  14 the Free Software Foundation, either version 3 of the License, or
  15 (at your option) any later version.
  16
  17 This program is distributed in the hope that it will be useful,
  18 but WITHOUT ANY WARRANTY; without even the implied warranty of
  19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20 GNU Affero General Public License for more details.
  21
  22 You should have received a copy of the GNU Affero General Public License
  23 along with this program.  If not, see <http://www.gnu.org/licenses/>.
  24 */
  25
  26 // Usage
  27 // -----
  28 // Request this file passing it a web page or feed URL in the querystring: makefulltextfeed.php?url=example.org/article
  29 // For more request parameters, see http://help.fivefilters.org/customer/portal/articles/226660-usage
  30
  31 error_reporting(E_ALL ^ E_NOTICE);
  32 ini_set("display_errors", 1);
  33 @set_time_limit(120);
  34
  35 // Deal with magic quotes
  36 if (get_magic_quotes_gpc()) {
  37         $process = array(&$_GET, &$_POST, &$_REQUEST);
  38         while (list($key, $val) = each($process)) {
  39                 foreach ($val as $k => $v) {
  40                         unset($process[$key][$k]);
  41                         if (is_array($v)) {
  42                                 $process[$key][stripslashes($k)] = $v;
  43                                 $process[] = &$process[$key][stripslashes($k)];
  44                         } else {
  45                                 $process[$key][stripslashes($k)] = stripslashes($v);
  46                         }
  47                 }
  48         }
  49         unset($process);
  50 }
  51
  52 // set include path
  53 set_include_path(realpath(dirname(__FILE__).'/libraries').PATH_SEPARATOR.get_include_path());
  54
  55 require_once dirname(__FILE__).'/makefulltextfeedHelpers.php';
  56
  57 ////////////////////////////////
  58 // Load config file
  59 ////////////////////////////////
  60 require dirname(__FILE__).'/config.php';
  61
  62 ////////////////////////////////
  63 // Prevent indexing/following by search engines because:
  64 // 1. The content is already public and presumably indexed (why create duplicates?)
  65 // 2. Not doing so might increase number of requests from search engines, thus increasing server load
  66 // Note: feed readers and services such as Yahoo Pipes will not be affected by this header.
  67 // Note: Using Disallow in a robots.txt file will be more effective (search engines will check
  68 // that before even requesting makefulltextfeed.php).
  69 ////////////////////////////////
  70 header('X-Robots-Tag: noindex, nofollow');
  71
  72 ////////////////////////////////
  73 // Check if service is enabled
  74 ////////////////////////////////
  75 if (!$options->enabled) {
  76         die('The full-text RSS service is currently disabled');
  77 }
  78
  79 ////////////////////////////////
  80 // Debug mode?
  81 // See the config file for debug options.
  82 ////////////////////////////////
  83 $debug_mode = false;
  84 if (isset($_GET['debug'])) {
  85         if ($options->debug === true || $options->debug == 'user') {
  86                 $debug_mode = true;
  87         } elseif ($options->debug == 'admin') {
  88                 session_start();
  89                 $debug_mode = (@$_SESSION['auth'] == 1);
  90         }
  91         if ($debug_mode) {
  92                 header('Content-Type: text/plain; charset=utf-8');
  93         } else {
  94                 if ($options->debug == 'admin') {
  95                         die('You must be logged in to the <a href="admin/">admin area</a> to see debug output.');
  96                 } else {
  97                         die('Debugging is disabled.');
  98                 }
  99         }
 100 }
 101
 102 ////////////////////////////////
 103 // Check for APC
 104 ////////////////////////////////
 105 $options->apc = $options->apc && function_exists('apc_add');
 106 if ($options->apc) {
 107         debug('APC is enabled and available on server');
 108 } else {
 109         debug('APC is disabled or not available on server');
 110 }
 111
 112 ////////////////////////////////
 113 // Check for smart cache
 114 ////////////////////////////////
 115 $options->smart_cache = $options->smart_cache && function_exists('apc_inc');
 116
 117 ////////////////////////////////
 118 // Check for feed URL
 119 ////////////////////////////////
 120 if (!isset($_GET['url'])) {
 121         die('No URL supplied');
 122 }
 123 $url = trim($_GET['url']);
 124 if (strtolower(substr($url, 0, 7)) == 'feed://') {
 125         $url = 'http://'.substr($url, 7);
 126 }
 127 if (!preg_match('!^https?://.+!i', $url)) {
 128         $url = 'http://'.$url;
 129 }
 130
 131 $url = filter_var($url, FILTER_SANITIZE_URL);
 132 $test = filter_var($url, FILTER_VALIDATE_URL, FILTER_FLAG_SCHEME_REQUIRED);
 133 // deal with bug http://bugs.php.net/51192 (present in PHP 5.2.13 and PHP 5.3.2)
 134 if ($test === false) {
 135         $test = filter_var(strtr($url, '-', '_'), FILTER_VALIDATE_URL, FILTER_FLAG_SCHEME_REQUIRED);
 136 }
 137 if ($test !== false && $test !== null && preg_match('!^https?://!', $url)) {
 138         // all okay
 139         unset($test);
 140 } else {
 141         die('Invalid URL supplied');
 142 }
 143 debug("Supplied URL: $url");
 144
 145 /////////////////////////////////
 146 // Redirect to hide API key
 147 /////////////////////////////////
 148 if (isset($_GET['key']) && ($key_index = array_search($_GET['key'], $options->api_keys)) !== false) {
 149         $host = $_SERVER['HTTP_HOST'];
 150         $path = rtrim(dirname($_SERVER['SCRIPT_NAME']), '/\\');
 151         $_qs_url = (strtolower(substr($url, 0, 7)) == 'http://') ? substr($url, 7) : $url;
 152         $redirect = 'http://'.htmlspecialchars($host.$path).'/makefulltextfeed.php?url='.urlencode($_qs_url);
 153         $redirect .= '&key='.$key_index;
 154         $redirect .= '&hash='.urlencode(sha1($_GET['key'].$url));
 155         if (isset($_GET['html'])) $redirect .= '&html='.urlencode($_GET['html']);
 156         if (isset($_GET['max'])) $redirect .= '&max='.(int)$_GET['max'];
 157         if (isset($_GET['links'])) $redirect .= '&links='.urlencode($_GET['links']);
 158         if (isset($_GET['exc'])) $redirect .= '&exc='.urlencode($_GET['exc']);
 159         if (isset($_GET['format'])) $redirect .= '&format='.urlencode($_GET['format']);
 160         if (isset($_GET['callback'])) $redirect .= '&callback='.urlencode($_GET['callback']);
 161         if (isset($_GET['l'])) $redirect .= '&l='.urlencode($_GET['l']);
 162         if (isset($_GET['xss'])) $redirect .= '&xss';
 163         if (isset($_GET['use_extracted_title'])) $redirect .= '&use_extracted_title';
 164         if (isset($_GET['content'])) $redirect .= '&content='.urlencode($_GET['content']);
 165         if (isset($_GET['summary'])) $redirect .= '&summary='.urlencode($_GET['summary']);
 166         if (isset($_GET['debug'])) $redirect .= '&debug';
 167         if ($debug_mode) {
 168                 debug('Redirecting to hide access key, follow URL below to continue');
 169                 debug("Location: $redirect");
 170         } else {
 171                 header("Location: $redirect");
 172         }
 173         exit;
 174 }
 175
 176 ///////////////////////////////////////////////
 177 // Set timezone.
 178 // Prevents warnings, but needs more testing -
 179 // perhaps if timezone is set in php.ini we
 180 // don't need to set it at all...
 181 ///////////////////////////////////////////////
 182 if (!ini_get('date.timezone') || !@date_default_timezone_set(ini_get('date.timezone'))) {
 183         date_default_timezone_set('UTC');
 184 }
 185
 186 ///////////////////////////////////////////////
 187 // Check if the request is explicitly for an HTML page
 188 ///////////////////////////////////////////////
 189 $html_only = (isset($_GET['html']) && ($_GET['html'] == '1' || $_GET['html'] == 'true'));
 190
 191 ///////////////////////////////////////////////
 192 // Check if valid key supplied
 193 ///////////////////////////////////////////////
 194 $valid_key = false;
 195 if (isset($_GET['key']) && isset($_GET['hash']) && isset($options->api_keys[(int)$_GET['key']])) {
 196         $valid_key = ($_GET['hash'] == sha1($options->api_keys[(int)$_GET['key']].$url));
 197 }
 198 $key_index = ($valid_key) ? (int)$_GET['key'] : 0;
 199 if (!$valid_key && $options->key_required) {
 200         die('A valid key must be supplied');
 201 }
 202 if (!$valid_key && isset($_GET['key']) && $_GET['key'] != '') {
 203         die('The entered key is invalid');
 204 }
 205
 206 if (file_exists('custom_init.php')) require 'custom_init.php';
 207
 208 ///////////////////////////////////////////////
 209 // Check URL against list of blacklisted URLs
 210 ///////////////////////////////////////////////
 211 if (!url_allowed($url)) die('URL blocked');
 212
 213 ///////////////////////////////////////////////
 214 // Max entries
 215 // see config.php to find these values
 216 ///////////////////////////////////////////////
 217 if (isset($_GET['max'])) {
 218         $max = (int)$_GET['max'];
 219         if ($valid_key) {
 220                 $max = min($max, $options->max_entries_with_key);
 221         } else {
 222                 $max = min($max, $options->max_entries);
 223         }
 224 } else {
 225         if ($valid_key) {
 226                 $max = $options->default_entries_with_key;
 227         } else {
 228                 $max = $options->default_entries;
 229         }
 230 }
 231
 232 ///////////////////////////////////////////////
 233 // Link handling
 234 ///////////////////////////////////////////////
 235 if (isset($_GET['links']) && in_array($_GET['links'], array('preserve', 'footnotes', 'remove'))) {
 236         $links = $_GET['links'];
 237 } else {
 238         $links = 'preserve';
 239 }
 240
 241 ///////////////////////////////////////////////
 242 // Favour item titles in feed?
 243 ///////////////////////////////////////////////
 244 $favour_feed_titles = true;
 245 if ($options->favour_feed_titles == 'user') {
 246         $favour_feed_titles = !isset($_GET['use_extracted_title']);
 247 } else {
 248         $favour_feed_titles = $options->favour_feed_titles;
 249 }
 250
 251 ///////////////////////////////////////////////
 252 // Include full content in output?
 253 ///////////////////////////////////////////////
 254 if ($options->content === 'user') {
 255         if (isset($_GET['content']) && $_GET['content'] === '0') {
 256                 $options->content = false;
 257         } else {
 258                 $options->content = true;
 259         }
 260 }
 261
 262 ///////////////////////////////////////////////
 263 // Include summaries in output?
 264 ///////////////////////////////////////////////
 265 if ($options->summary === 'user') {
 266         if (isset($_GET['summary']) && $_GET['summary'] === '1') {
 267                 $options->summary = true;
 268         } else {
 269                 $options->summary = false;
 270         }
 271 }
 272
 273 ///////////////////////////////////////////////
 274 // Exclude items if extraction fails
 275 ///////////////////////////////////////////////
 276 if ($options->exclude_items_on_fail === 'user') {
 277         $exclude_on_fail = (isset($_GET['exc']) && ($_GET['exc'] == '1'));
 278 } else {
 279         $exclude_on_fail = $options->exclude_items_on_fail;
 280 }
 281
 282 ///////////////////////////////////////////////
 283 // Detect language
 284 ///////////////////////////////////////////////
 285 if ($options->detect_language === 'user') {
 286         if (isset($_GET['l'])) {
 287                 $detect_language = (int)$_GET['l'];
 288         } else {
 289                 $detect_language = 1;
 290         }
 291 } else {
 292         $detect_language = $options->detect_language;
 293 }
 294
 295 $use_cld = extension_loaded('cld') && (version_compare(PHP_VERSION, '5.3.0') >= 0);
 296
 297 /////////////////////////////////////
 298 // Check for valid format
 299 // (stick to RSS (or RSS as JSON) for the time being)
 300 /////////////////////////////////////
 301 if (isset($_GET['format']) && $_GET['format'] == 'json') {
 302         $format = 'json';
 303 } else {
 304         $format = 'rss';
 305 }
 306
 307 /////////////////////////////////////
 308 // Should we do XSS filtering?
 309 /////////////////////////////////////
 310 if ($options->xss_filter === 'user') {
 311         $xss_filter = isset($_GET['xss']);
 312 } else {
 313         $xss_filter = $options->xss_filter;
 314 }
 315 if (!$xss_filter && isset($_GET['xss'])) {
 316         die('XSS filtering is disabled in config');
 317 }
 318
 319 /////////////////////////////////////
 320 // Check for JSONP
 321 // Regex from https://gist.github.com/1217080
 322 /////////////////////////////////////
 323 $callback = null;
 324 if ($format =='json' && isset($_GET['callback'])) {
 325         $callback = trim($_GET['callback']);
 326         foreach (explode('.', $callback) as $_identifier) {
 327                 if (!preg_match('/^[a-zA-Z_$][0-9a-zA-Z_$]*(?:\[(?:".+"|\'.+\'|\d+)\])*?$/', $_identifier)) {
 328                         die('Invalid JSONP callback');
 329                 }
 330         }
 331         debug("JSONP callback: $callback");
 332 }
 333
 334 //////////////////////////////////
 335 // Enable Cross-Origin Resource Sharing (CORS)
 336 //////////////////////////////////
 337 if ($options->cors) header('Access-Control-Allow-Origin: *');
 338
 339 //////////////////////////////////
 340 // Check for cached copy
 341 //////////////////////////////////
 342 if ($options->caching) {
 343         debug('Caching is enabled...');
 344         $cache_id = md5($max.$url.(int)$valid_key.$links.(int)$favour_feed_titles.(int)$options->content.(int)$options->summary.(int)$xss_filter.(int)$exclude_on_fail.$format.$detect_language.(int)isset($_GET['pubsub']));
 345         $check_cache = true;
 346         if ($options->apc && $options->smart_cache) {
 347                 apc_add("cache.$cache_id", 0, 10*60);
 348                 $apc_cache_hits = (int)apc_fetch("cache.$cache_id");
 349                 $check_cache = ($apc_cache_hits >= 2);
 350                 apc_inc("cache.$cache_id");
 351                 if ($check_cache) {
 352                         debug('Cache key found in APC, we\'ll try to load cache file from disk');
 353                 } else {
 354                         debug('Cache key not found in APC');
 355                 }
 356         }
 357         if ($check_cache) {
 358                 $cache = get_cache();
 359                 if ($data = $cache->load($cache_id)) {
 360                         if ($debug_mode) {
 361                                 debug('Loaded cached copy');
 362                                 exit;
 363                         }
 364                         if ($format == 'json') {
 365                                 if ($callback === null) {
 366                                         header('Content-type: application/json; charset=UTF-8');
 367                                 } else {
 368                                         header('Content-type: application/javascript; charset=UTF-8');
 369                                 }
 370                         } else {
 371                                 header('Content-type: text/xml; charset=UTF-8');
 372                                 header('X-content-type-options: nosniff');
 373                         }
 374                         if (headers_sent()) die('Some data has already been output, can\'t send RSS file');
 375                         if ($callback) {
 376                                 echo "$callback($data);";
 377                         } else {
 378                                 echo $data;
 379                         }
 380                         exit;
 381                 }
 382         }
 383 }
 384
 385 //////////////////////////////////
 386 // Set Expires header
 387 //////////////////////////////////
 388 if (!$debug_mode) {
 389         header('Expires: ' . gmdate('D, d M Y H:i:s', time()+(60*10)) . ' GMT');
 390 }
 391
 392 //////////////////////////////////
 393 // Set up HTTP agent
 394 //////////////////////////////////
 395 global $http;
 396 $http = new HumbleHttpAgent();
 397 $http->debug = $debug_mode;
 398 $http->userAgentMap = $options->user_agents;
 399 $http->headerOnlyTypes = array_keys($options->content_type_exc);
 400 $http->rewriteUrls = $options->rewrite_url;
 401
 402 //////////////////////////////////
 403 // Set up Content Extractor
 404 //////////////////////////////////
 405 global $extractor;
 406 $extractor = new ContentExtractor(dirname(__FILE__).'/site_config/custom', dirname(__FILE__).'/site_config/standard');
 407 $extractor->debug = $debug_mode;
 408 SiteConfig::$debug = $debug_mode;
 409 SiteConfig::use_apc($options->apc);
 410 $extractor->fingerprints = $options->fingerprints;
 411 $extractor->allowedParsers = $options->allowed_parsers;
 412
 413 ////////////////////////////////
 414 // Get RSS/Atom feed
 415 ////////////////////////////////
 416 if (!$html_only) {
 417         debug('--------');
 418         debug("Attempting to process URL as feed");
 419         // Send user agent header showing PHP (prevents a HTML response from feedburner)
 420         $http->userAgentDefault = HumbleHttpAgent::UA_PHP;
 421         // configure SimplePie HTTP extension class to use our HumbleHttpAgent instance
 422         SimplePie_HumbleHttpAgent::set_agent($http);
 423         $feed = new SimplePie();
 424         // some feeds use the text/html content type - force_feed tells SimplePie to process anyway
 425         $feed->force_feed(true);
 426         $feed->set_file_class('SimplePie_HumbleHttpAgent');
 427         //$feed->set_feed_url($url); // colons appearing in the URL's path get encoded
 428         $feed->feed_url = $url;
 429         $feed->set_autodiscovery_level(SIMPLEPIE_LOCATOR_NONE);
 430         $feed->set_timeout(20);
 431         $feed->enable_cache(false);
 432         $feed->set_stupidly_fast(true);
 433         $feed->enable_order_by_date(false); // we don't want to do anything to the feed
 434         $feed->set_url_replacements(array());
 435         // initialise the feed
 436         // the @ suppresses notices which on some servers causes a 500 internal server error
 437         $result = @$feed->init();
 438         //$feed->handle_content_type();
 439         //$feed->get_title();
 440         if ($result && (!is_array($feed->data) || count($feed->data) == 0)) {
 441                 die('Sorry, no feed items found');
 442         }
 443         // from now on, we'll identify ourselves as a browser
 444         $http->userAgentDefault = HumbleHttpAgent::UA_BROWSER;
 445 }
 446
 447 ////////////////////////////////////////////////////////////////////////////////
 448 // Our given URL is not a feed, so let's create our own feed with a single item:
 449 // the given URL. This basically treats all non-feed URLs as if they were
 450 // single-item feeds.
 451 ////////////////////////////////////////////////////////////////////////////////
 452 $isDummyFeed = false;
 453 if ($html_only || !$result) {
 454         debug('--------');
 455         debug("Constructing a single-item feed from URL");
 456         $isDummyFeed = true;
 457         unset($feed, $result);
 458         // create single item dummy feed object
 459         $feed = new DummySingleItemFeed($url);
 460 }
 461
 462 ////////////////////////////////////////////
 463 // Create full-text feed
 464 ////////////////////////////////////////////
 465 $output = new FeedWriter();
 466 $output->setTitle(strip_tags($feed->get_title()));
 467 $output->setDescription(strip_tags($feed->get_description()));
 468 $output->setXsl('css/feed.xsl'); // Chrome uses this, most browsers ignore it
 469 if ($valid_key && isset($_GET['pubsub'])) { // used only on fivefilters.org at the moment
 470         $output->addHub('http://fivefilters.superfeedr.com/');
 471         $output->addHub('http://pubsubhubbub.appspot.com/');
 472         $output->setSelf('http://'.$_SERVER['HTTP_HOST'].$_SERVER['REQUEST_URI']);
 473 }
 474 $output->setLink($feed->get_link()); // Google Reader uses this for pulling in favicons
 475 if ($img_url = $feed->get_image_url()) {
 476         $output->setImage($feed->get_title(), $feed->get_link(), $img_url);
 477 }
 478
 479 ////////////////////////////////////////////
 480 // Loop through feed items
 481 ////////////////////////////////////////////
 482 $items = $feed->get_items(0, $max);
 483 // Request all feed items in parallel (if supported)
 484 $urls_sanitized = array();
 485 $urls = array();
 486 foreach ($items as $key => $item) {
 487         $permalink = htmlspecialchars_decode($item->get_permalink());
 488         // Colons in URL path segments get encoded by SimplePie, yet some sites expect them unencoded
 489         $permalink = str_replace('%3A', ':', $permalink);
 490         // validateUrl() strips non-ascii characters
 491         // simplepie already sanitizes URLs so let's not do it again here.
 492         //$permalink = $http->validateUrl($permalink);
 493         if ($permalink) {
 494                 $urls_sanitized[] = $permalink;
 495         }
 496         $urls[$key] = $permalink;
 497 }
 498 debug('--------');
 499 debug('Fetching feed items');
 500 $http->fetchAll($urls_sanitized);
 501 //$http->cacheAll();
 502
 503 // count number of items added to full feed
 504 $item_count = 0;
 505
 506 foreach ($items as $key => $item) {
 507         debug('--------');
 508         debug('Processing feed item '.($item_count+1));
 509         $do_content_extraction = true;
 510         $extract_result = false;
 511         $text_sample = null;
 512         $permalink = $urls[$key];
 513         debug("Item URL: $permalink");
 514         $extracted_title = '';
 515         $feed_item_title = $item->get_title();
 516         if ($feed_item_title !== null) {
 517                 $feed_item_title = strip_tags(htmlspecialchars_decode($feed_item_title));
 518         }
 519         $newitem = $output->createNewItem();
 520         $newitem->setTitle($feed_item_title);
 521         if ($valid_key && isset($_GET['pubsub'])) { // used only on fivefilters.org at the moment
 522                 if ($permalink !== false) {
 523                         $newitem->setLink('http://fivefilters.org/content-only/redirect.php?url='.urlencode($permalink));
 524                 } else {
 525                         $newitem->setLink('http://fivefilters.org/content-only/redirect.php?url='.urlencode($item->get_permalink()));
 526                 }
 527         } else {
 528                 if ($permalink !== false) {
 529                         $newitem->setLink($permalink);
 530                 } else {
 531                         $newitem->setLink($item->get_permalink());
 532                 }
 533         }
 534         //if ($permalink && ($response = $http->get($permalink, true)) && $response['status_code'] < 300) {
 535         // Allowing error codes - some sites return correct content with error status
 536         // e.g. prospectmagazine.co.uk returns 403
 537         if ($permalink && ($response = $http->get($permalink, true)) && ($response['status_code'] < 300 || $response['status_code'] > 400)) {
 538                 $effective_url = $response['effective_url'];
 539                 if (!url_allowed($effective_url)) continue;
 540                 // check if action defined for returned Content-Type
 541                 $mime_info = get_mime_action_info($response['headers']);
 542                 if (isset($mime_info['action'])) {
 543                         if ($mime_info['action'] == 'exclude') {
 544                                 continue; // skip this feed item entry
 545                         } elseif ($mime_info['action'] == 'link') {
 546                                 if ($mime_info['type'] == 'image') {
 547                                         $html = "<a href=\"$effective_url\"><img src=\"$effective_url\" alt=\"{$mime_info['name']}\" /></a>";
 548                                 } else {
 549                                         $html = "<a href=\"$effective_url\">Download {$mime_info['name']}</a>";
 550                                 }
 551                                 $extracted_title = $mime_info['name'];
 552                                 $do_content_extraction = false;
 553                         }
 554                 }
 555                 if ($do_content_extraction) {
 556                         $html = $response['body'];
 557                         // remove strange things
 558                         $html = str_replace('</[>', '', $html);
 559                         $html = convert_to_utf8($html, $response['headers']);
 560                         // check site config for single page URL - fetch it if found
 561                         $is_single_page = false;
 562                         if ($single_page_response = getSinglePage($item, $html, $effective_url)) {
 563                                 $is_single_page = true;
 564                                 $effective_url = $single_page_response['effective_url'];
 565                                 // check if action defined for returned Content-Type
 566                                 $mime_info = get_mime_action_info($single_page_response['headers']);
 567                                 if (isset($mime_info['action'])) {
 568                                         if ($mime_info['action'] == 'exclude') {
 569                                                 continue; // skip this feed item entry
 570                                         } elseif ($mime_info['action'] == 'link') {
 571                                                 if ($mime_info['type'] == 'image') {
 572                                                         $html = "<a href=\"$effective_url\"><img src=\"$effective_url\" alt=\"{$mime_info['name']}\" /></a>";
 573                                                 } else {
 574                                                         $html = "<a href=\"$effective_url\">Download {$mime_info['name']}</a>";
 575                                                 }
 576                                                 $extracted_title = $mime_info['name'];
 577                                                 $do_content_extraction = false;
 578                                         }
 579                                 }
 580                                 if ($do_content_extraction) {
 581                                         $html = $single_page_response['body'];
 582                                         // remove strange things
 583                                         $html = str_replace('</[>', '', $html);
 584                                         $html = convert_to_utf8($html, $single_page_response['headers']);
 585                                         debug("Retrieved single-page view from $effective_url");
 586                                 }
 587                                 unset($single_page_response);
 588                         }
 589                 }
 590                 if ($do_content_extraction) {
 591                         debug('--------');
 592                         debug('Attempting to extract content');
 593                         $extract_result = $extractor->process($html, $effective_url);
 594                         $readability = $extractor->readability;
 595                         $content_block = ($extract_result) ? $extractor->getContent() : null;
 596                         $extracted_title = ($extract_result) ? $extractor->getTitle() : '';
 597                         // Deal with multi-page articles
 598                         //die('Next: '.$extractor->getNextPageUrl());
 599                         $is_multi_page = (!$is_single_page && $extract_result && $extractor->getNextPageUrl());
 600                         if ($options->multipage && $is_multi_page && $options->content) {
 601                                 debug('--------');
 602                                 debug('Attempting to process multi-page article');
 603                                 $multi_page_urls = array();
 604                                 $multi_page_content = array();
 605                                 while ($next_page_url = $extractor->getNextPageUrl()) {
 606                                         debug('--------');
 607                                         debug('Processing next page: '.$next_page_url);
 608                                         // If we've got URL, resolve against $url
 609                                         if ($next_page_url = makeAbsoluteStr($effective_url, $next_page_url)) {
 610                                                 // check it's not what we have already!
 611                                                 if (!in_array($next_page_url, $multi_page_urls)) {
 612                                                         // it's not, so let's attempt to fetch it
 613                                                         $multi_page_urls[] = $next_page_url;
 614                                                         $_prev_ref = $http->referer;
 615                                                         if (($response = $http->get($next_page_url, true)) && $response['status_code'] < 300) {
 616                                                                 // make sure mime type is not something with a different action associated
 617                                                                 $page_mime_info = get_mime_action_info($response['headers']);
 618                                                                 if (!isset($page_mime_info['action'])) {
 619                                                                         $html = $response['body'];
 620                                                                         // remove strange things
 621                                                                         $html = str_replace('</[>', '', $html);
 622                                                                         $html = convert_to_utf8($html, $response['headers']);
 623                                                                         if ($extractor->process($html, $next_page_url)) {
 624                                                                                 $multi_page_content[] = $extractor->getContent();
 625                                                                                 continue;
 626                                                                         } else { debug('Failed to extract content'); }
 627                                                                 } else { debug('MIME type requires different action'); }
 628                                                         } else { debug('Failed to fetch URL'); }
 629                                                 } else { debug('URL already processed'); }
 630                                         } else { debug('Failed to resolve against '.$effective_url); }
 631                                         // failed to process next_page_url, so cancel further requests
 632                                         $multi_page_content = array();
 633                                         break;
 634                                 }
 635                                 // did we successfully deal with this multi-page article?
 636                                 if (empty($multi_page_content)) {
 637                                         debug('Failed to extract all parts of multi-page article, so not going to include them');
 638                                         $_page = $readability->dom->createElement('p');
 639                                         $_page->innerHTML = '<em>This article appears to continue on subsequent pages which we could not extract</em>';
 640                                         $multi_page_content[] = $_page;
 641                                 }
 642                                 foreach ($multi_page_content as $_page) {
 643                                         $_page = $content_block->ownerDocument->importNode($_page, true);
 644                                         $content_block->appendChild($_page);
 645                                 }
 646                                 unset($multi_page_urls, $multi_page_content, $page_mime_info, $next_page_url, $_page);
 647                         }
 648                 }
 649                 // use extracted title for both feed and item title if we're using single-item dummy feed
 650                 if ($isDummyFeed) {
 651                         $output->setTitle($extracted_title);
 652                         $newitem->setTitle($extracted_title);
 653                 } else {
 654                         // use extracted title instead of feed item title?
 655                         if (!$favour_feed_titles && $extracted_title != '') {
 656                                 debug('Using extracted title in generated feed');
 657                                 $newitem->setTitle($extracted_title);
 658                         }
 659                 }
 660         }
 661         if ($do_content_extraction) {
 662                 // if we failed to extract content...
 663                 if (!$extract_result) {
 664                         if ($exclude_on_fail) {
 665                                 debug('Failed to extract, so skipping (due to exclude on fail parameter)');
 666                                 continue; // skip this and move to next item
 667                         }
 668                         //TODO: get text sample for language detection
 669                         $html = $options->error_message;
 670                         // keep the original item description
 671                         $html .= $item->get_description();
 672                 } else {
 673                         $readability->clean($content_block, 'select');
 674                         if ($options->rewrite_relative_urls) makeAbsolute($effective_url, $content_block);
 675                         // footnotes
 676                         if (($links == 'footnotes') && (strpos($effective_url, 'wikipedia.org') === false)) {
 677                                 $readability->addFootnotes($content_block);
 678                         }
 679                         // remove nesting: <div><div><div><p>test</p></div></div></div> = <p>test</p>
 680                         while ($content_block->childNodes->length == 1 && $content_block->firstChild->nodeType === XML_ELEMENT_NODE) {
 681                                 // only follow these tag names
 682                                 if (!in_array(strtolower($content_block->tagName), array('div', 'article', 'section', 'header', 'footer'))) break;
 683                                 //$html = $content_block->firstChild->innerHTML; // FTR 2.9.5
 684                                 $content_block = $content_block->firstChild;
 685                         }
 686                         // convert content block to HTML string
 687                         // Need to preserve things like body: //img[@id='feature']
 688                         if (in_array(strtolower($content_block->tagName), array('div', 'article', 'section', 'header', 'footer'))) {
 689                                 $html = $content_block->innerHTML;
 690                         } else {
 691                                 $html = $content_block->ownerDocument->saveXML($content_block); // essentially outerHTML
 692                         }
 693                         //unset($content_block);
 694                         // post-processing cleanup
 695                         $html = preg_replace('!<p>[\s\h\v]*</p>!u', '', $html);
 696                         if ($links == 'remove') {
 697                                 $html = preg_replace('!</?a[^>]*>!', '', $html);
 698                         }
 699                         // get text sample for language detection
 700                         $text_sample = strip_tags(substr($html, 0, 500));
 701                         $html = make_substitutions($options->message_to_prepend).$html;
 702                         $html .= make_substitutions($options->message_to_append);
 703                 }
 704         }
 705
 706         if ($valid_key && isset($_GET['pubsub'])) { // used only on fivefilters.org at the moment
 707                 $newitem->addElement('guid', 'http://fivefilters.org/content-only/redirect.php?url='.urlencode($item->get_permalink()), array('isPermaLink'=>'false'));
 708         } else {
 709                 $newitem->addElement('guid', $item->get_permalink(), array('isPermaLink'=>'true'));
 710         }
 711         // filter xss?
 712         if ($xss_filter) {
 713                 debug('Filtering HTML to remove XSS');
 714                 $html = htmLawed::hl($html, array('safe'=>1, 'deny_attribute'=>'style', 'comment'=>1, 'cdata'=>1));
 715         }
 716
 717         // add content
 718         if ($options->summary === true) {
 719                 // get summary
 720                 $summary = '';
 721                 if (!$do_content_extraction) {
 722                         $summary = $html;
 723                 } else {
 724                         // Try to get first few paragraphs
 725                         if (isset($content_block) && ($content_block instanceof DOMElement)) {
 726                                 $_paras = $content_block->getElementsByTagName('p');
 727                                 foreach ($_paras as $_para) {
 728                                         $summary .= preg_replace("/[\n\r\t ]+/", ' ', $_para->textContent).' ';
 729                                         if (strlen($summary) > 200) break;
 730                                 }
 731                         } else {
 732                                 $summary = $html;
 733                         }
 734                 }
 735                 unset($_paras, $_para);
 736                 $summary = get_excerpt($summary);
 737                 $newitem->setDescription($summary);
 738                 if ($options->content) $newitem->setElement('content:encoded', $html);
 739         } else {
 740                 if ($options->content) $newitem->setDescription($html);
 741         }
 742
 743         // set date
 744         if ((int)$item->get_date('U') > 0) {
 745                 $newitem->setDate((int)$item->get_date('U'));
 746         } elseif ($extractor->getDate()) {
 747                 $newitem->setDate($extractor->getDate());
 748         }
 749
 750         // add authors
 751         if ($authors = $item->get_authors()) {
 752                 foreach ($authors as $author) {
 753                         // for some feeds, SimplePie stores author's name as email, e.g. http://feeds.feedburner.com/nymag/intel
 754                         if ($author->get_name() !== null) {
 755                                 $newitem->addElement('dc:creator', $author->get_name());
 756                         } elseif ($author->get_email() !== null) {
 757                                 $newitem->addElement('dc:creator', $author->get_email());
 758                         }
 759                 }
 760         } elseif ($authors = $extractor->getAuthors()) {
 761                 //TODO: make sure the list size is reasonable
 762                 foreach ($authors as $author) {
 763                         // TODO: xpath often selects authors from other articles linked from the page.
 764                         // for now choose first item
 765                         $newitem->addElement('dc:creator', $author);
 766                         break;
 767                 }
 768         }
 769
 770         // add language
 771         if ($detect_language) {
 772                 $language = $extractor->getLanguage();
 773                 if (!$language) $language = $feed->get_language();
 774                 if (($detect_language == 3 || (!$language && $detect_language == 2)) && $text_sample) {
 775                         try {
 776                                 if ($use_cld) {
 777                                         // Use PHP-CLD extension
 778                                         $php_cld = 'CLD\detect'; // in quotes to prevent PHP 5.2 parse error
 779                                         $res = $php_cld($text_sample);
 780                                         if (is_array($res) && count($res) > 0) {
 781                                                 $language = $res[0]['code'];
 782                                         }
 783                                 } else {
 784                                         //die('what');
 785                                         // Use PEAR's Text_LanguageDetect
 786                                         if (!isset($l)) {
 787                                           $l = new Text_LanguageDetect();
 788                                           $l->setNameMode(2); // return ISO 639-1 codes (e.g. "en")
 789                                         }
 790                                         $l_result = $l->detect($text_sample, 1);
 791                                         if (count($l_result) > 0) {
 792                                                 $language = key($l_result);
 793                                         }
 794                                 }
 795                         } catch (Exception $e) {
 796                                 //die('error: '.$e);
 797                                 // do nothing
 798                         }
 799                 }
 800                 if ($language && (strlen($language) < 7)) {
 801                         $newitem->addElement('dc:language', $language);
 802                 }
 803         }
 804
 805         // add MIME type (if it appeared in our exclusions lists)
 806         if (isset($mime_info['mime'])) $newitem->addElement('dc:format', $mime_info['mime']);
 807         // add effective URL (URL after redirects)
 808         if (isset($effective_url)) {
 809                 //TODO: ensure $effective_url is valid witout - sometimes it causes problems, e.g.
 810                 //http://www.siasat.pk/forum/showthread.php?108883-Pakistan-Chowk-by-Rana-Mubashir-�-25th-March-2012-Special-Program-from-Liari-(Karachi)
 811                 //temporary measure: use utf8_encode()
 812                 $newitem->addElement('dc:identifier', remove_url_cruft(utf8_encode($effective_url)));
 813         } else {
 814                 $newitem->addElement('dc:identifier', remove_url_cruft($item->get_permalink()));
 815         }
 816
 817         // add categories
 818         if ($categories = $item->get_categories()) {
 819                 foreach ($categories as $category) {
 820                         if ($category->get_label() !== null) {
 821                                 $newitem->addElement('category', $category->get_label());
 822                         }
 823                 }
 824         }
 825
 826         // check for enclosures
 827         if ($options->keep_enclosures) {
 828                 if ($enclosures = $item->get_enclosures()) {
 829                         foreach ($enclosures as $enclosure) {
 830                                 // thumbnails
 831                                 foreach ((array)$enclosure->get_thumbnails() as $thumbnail) {
 832                                         $newitem->addElement('media:thumbnail', '', array('url'=>$thumbnail));
 833                                 }
 834                                 if (!$enclosure->get_link()) continue;
 835                                 $enc = array();
 836                                 // Media RSS spec ($enc): http://search.yahoo.com/mrss
 837                                 // SimplePie methods ($enclosure): http://simplepie.org/wiki/reference/start#methods4
 838                                 $enc['url'] = $enclosure->get_link();
 839                                 if ($enclosure->get_length()) $enc['fileSize'] = $enclosure->get_length();
 840                                 if ($enclosure->get_type()) $enc['type'] = $enclosure->get_type();
 841                                 if ($enclosure->get_medium()) $enc['medium'] = $enclosure->get_medium();
 842                                 if ($enclosure->get_expression()) $enc['expression'] = $enclosure->get_expression();
 843                                 if ($enclosure->get_bitrate()) $enc['bitrate'] = $enclosure->get_bitrate();
 844                                 if ($enclosure->get_framerate()) $enc['framerate'] = $enclosure->get_framerate();
 845                                 if ($enclosure->get_sampling_rate()) $enc['samplingrate'] = $enclosure->get_sampling_rate();
 846                                 if ($enclosure->get_channels()) $enc['channels'] = $enclosure->get_channels();
 847                                 if ($enclosure->get_duration()) $enc['duration'] = $enclosure->get_duration();
 848                                 if ($enclosure->get_height()) $enc['height'] = $enclosure->get_height();
 849                                 if ($enclosure->get_width()) $enc['width'] = $enclosure->get_width();
 850                                 if ($enclosure->get_language()) $enc['lang'] = $enclosure->get_language();
 851                                 $newitem->addElement('media:content', '', $enc);
 852                         }
 853                 }
 854         }
 855         $output->addItem($newitem);
 856         unset($html);
 857         $item_count++;
 858 }
 859
 860 // output feed
 861 debug('Done!');
 862 /*
 863 if ($debug_mode) {
 864         $_apc_data = apc_cache_info('user');
 865         var_dump($_apc_data); exit;
 866 }
 867 */
 868 if (!$debug_mode) {
 869         if ($callback) echo "$callback("; // if $callback is set, $format also == 'json'
 870         if ($format == 'json') $output->setFormat(($callback === null) ? JSON : JSONP);
 871         $add_to_cache = $options->caching;
 872         // is smart cache mode enabled?
 873         if ($add_to_cache && $options->apc && $options->smart_cache) {
 874                 // yes, so only cache if this is the second request for this URL
 875                 $add_to_cache = ($apc_cache_hits >= 2);
 876                 // purge cache
 877                 if ($options->cache_cleanup > 0) {
 878                         if (rand(1, $options->cache_cleanup) == 1) {
 879                                 // apc purge code adapted from from http://www.thimbleopensource.com/tutorials-snippets/php-apc-expunge-script
 880                                 $_apc_data = apc_cache_info('user');
 881                                 foreach ($_apc_data['cache_list'] as $_apc_item) {
 882                                   if ($_apc_item['ttl'] > 0 && ($_apc_item['ttl'] + $_apc_item['creation_time'] < time())) {
 883                                         apc_delete($_apc_item['info']);
 884                                   }
 885                                 }
 886                         }
 887                 }
 888         }
 889         if ($add_to_cache) {
 890                 ob_start();
 891                 $output->genarateFeed(false);
 892                 $output = ob_get_contents();
 893                 ob_end_clean();
 894                 if ($html_only && $item_count == 0) {
 895                         // do not cache - in case of temporary server glitch at source URL
 896                 } else {
 897                         $cache = get_cache();
 898                         if ($add_to_cache) $cache->save($output, $cache_id);
 899                 }
 900                 echo $output;
 901         } else {
 902                 $output->genarateFeed(false);
 903         }
 904         if ($callback) echo ');';
 905 }
 906