aboutsummaryrefslogtreecommitdiffhomepage
path: root/inc/3rdparty/makefulltextfeed.php
diff options
context:
space:
mode:
Diffstat (limited to 'inc/3rdparty/makefulltextfeed.php')
-rwxr-xr-xinc/3rdparty/makefulltextfeed.php714
1 files changed, 214 insertions, 500 deletions
diff --git a/inc/3rdparty/makefulltextfeed.php b/inc/3rdparty/makefulltextfeed.php
index 2852c4c2..a081f88b 100755
--- a/inc/3rdparty/makefulltextfeed.php
+++ b/inc/3rdparty/makefulltextfeed.php
@@ -3,8 +3,8 @@
3// Author: Keyvan Minoukadeh 3// Author: Keyvan Minoukadeh
4// Copyright (c) 2013 Keyvan Minoukadeh 4// Copyright (c) 2013 Keyvan Minoukadeh
5// License: AGPLv3 5// License: AGPLv3
6// Version: 3.1 6// Version: 3.2
7// Date: 2013-03-05 7// Date: 2013-05-13
8// More info: http://fivefilters.org/content-only/ 8// More info: http://fivefilters.org/content-only/
9// Help: http://help.fivefilters.org 9// Help: http://help.fivefilters.org
10 10
@@ -25,14 +25,10 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
25 25
26// Usage 26// Usage
27// ----- 27// -----
28// Request this file passing it your feed in the querystring: makefulltextfeed.php?url=mysite.org 28// Request this file passing it a web page or feed URL in the querystring: makefulltextfeed.php?url=example.org/article
29// The following options can be passed in the querystring: 29// For more request parameters, see http://help.fivefilters.org/customer/portal/articles/226660-usage
30// * URL: url=[feed or website url] (required, should be URL-encoded - in php: urlencode($url))
31// * URL points to HTML (not feed): html=true (optional, by default it's automatically detected)
32// * API key: key=[api key] (optional, refer to config.php)
33// * Max entries to process: max=[max number of items] (optional)
34 30
35error_reporting(E_ALL ^ E_NOTICE); 31//error_reporting(E_ALL ^ E_NOTICE);
36ini_set("display_errors", 1); 32ini_set("display_errors", 1);
37@set_time_limit(120); 33@set_time_limit(120);
38 34
@@ -55,42 +51,8 @@ if (get_magic_quotes_gpc()) {
55 51
56// set include path 52// set include path
57set_include_path(realpath(dirname(__FILE__).'/libraries').PATH_SEPARATOR.get_include_path()); 53set_include_path(realpath(dirname(__FILE__).'/libraries').PATH_SEPARATOR.get_include_path());
58// Autoloading of classes allows us to include files only when they're 54
59// needed. If we've got a cached copy, for example, only Zend_Cache is loaded. 55require_once dirname(__FILE__).'/makefulltextfeedHelpers.php';
60function autoload($class_name) {
61 static $dir = null;
62 if ($dir === null) $dir = dirname(__FILE__).'/libraries/';
63 static $mapping = array(
64 // Include FeedCreator for RSS/Atom creation
65 'FeedWriter' => 'feedwriter/FeedWriter.php',
66 'FeedItem' => 'feedwriter/FeedItem.php',
67 // Include ContentExtractor and Readability for identifying and extracting content from URLs
68 'ContentExtractor' => 'content-extractor/ContentExtractor.php',
69 'SiteConfig' => 'content-extractor/SiteConfig.php',
70 'Readability' => 'readability/Readability.php',
71 // Include Humble HTTP Agent to allow parallel requests and response caching
72 'HumbleHttpAgent' => 'humble-http-agent/HumbleHttpAgent.php',
73 'SimplePie_HumbleHttpAgent' => 'humble-http-agent/SimplePie_HumbleHttpAgent.php',
74 'CookieJar' => 'humble-http-agent/CookieJar.php',
75 // Include Zend Cache to improve performance (cache results)
76 'Zend_Cache' => 'Zend/Cache.php',
77 // Language detect
78 'Text_LanguageDetect' => 'language-detect/LanguageDetect.php',
79 // HTML5 Lib
80 'HTML5_Parser' => 'html5/Parser.php',
81 // htmLawed - used if XSS filter is enabled (xss_filter)
82 'htmLawed' => 'htmLawed/htmLawed.php'
83 );
84 if (isset($mapping[$class_name])) {
85 debug("** Loading class $class_name ({$mapping[$class_name]})");
86 require $dir.$mapping[$class_name];
87 return true;
88 } else {
89 return false;
90 }
91}
92spl_autoload_register('autoload');
93require dirname(__FILE__).'/libraries/simplepie/autoloader.php';
94 56
95//////////////////////////////// 57////////////////////////////////
96// Load config file 58// Load config file
@@ -110,8 +72,8 @@ header('X-Robots-Tag: noindex, nofollow');
110//////////////////////////////// 72////////////////////////////////
111// Check if service is enabled 73// Check if service is enabled
112//////////////////////////////// 74////////////////////////////////
113if (!$options->enabled) { 75if (!$options->enabled) {
114 die('The full-text RSS service is currently disabled'); 76 die('The full-text RSS service is currently disabled');
115} 77}
116 78
117//////////////////////////////// 79////////////////////////////////
@@ -155,8 +117,8 @@ $options->smart_cache = $options->smart_cache && function_exists('apc_inc');
155//////////////////////////////// 117////////////////////////////////
156// Check for feed URL 118// Check for feed URL
157//////////////////////////////// 119////////////////////////////////
158if (!isset($_GET['url'])) { 120if (!isset($_GET['url'])) {
159 die('No URL supplied'); 121 die('No URL supplied');
160} 122}
161$url = trim($_GET['url']); 123$url = trim($_GET['url']);
162if (strtolower(substr($url, 0, 7)) == 'feed://') { 124if (strtolower(substr($url, 0, 7)) == 'feed://') {
@@ -195,10 +157,12 @@ if (isset($_GET['key']) && ($key_index = array_search($_GET['key'], $options->ap
195 if (isset($_GET['links'])) $redirect .= '&links='.urlencode($_GET['links']); 157 if (isset($_GET['links'])) $redirect .= '&links='.urlencode($_GET['links']);
196 if (isset($_GET['exc'])) $redirect .= '&exc='.urlencode($_GET['exc']); 158 if (isset($_GET['exc'])) $redirect .= '&exc='.urlencode($_GET['exc']);
197 if (isset($_GET['format'])) $redirect .= '&format='.urlencode($_GET['format']); 159 if (isset($_GET['format'])) $redirect .= '&format='.urlencode($_GET['format']);
198 if (isset($_GET['callback'])) $redirect .= '&callback='.urlencode($_GET['callback']); 160 if (isset($_GET['callback'])) $redirect .= '&callback='.urlencode($_GET['callback']);
199 if (isset($_GET['l'])) $redirect .= '&l='.urlencode($_GET['l']); 161 if (isset($_GET['l'])) $redirect .= '&l='.urlencode($_GET['l']);
200 if (isset($_GET['xss'])) $redirect .= '&xss'; 162 if (isset($_GET['xss'])) $redirect .= '&xss';
201 if (isset($_GET['use_extracted_title'])) $redirect .= '&use_extracted_title'; 163 if (isset($_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']);
202 if (isset($_GET['debug'])) $redirect .= '&debug'; 166 if (isset($_GET['debug'])) $redirect .= '&debug';
203 if ($debug_mode) { 167 if ($debug_mode) {
204 debug('Redirecting to hide access key, follow URL below to continue'); 168 debug('Redirecting to hide access key, follow URL below to continue');
@@ -211,7 +175,7 @@ if (isset($_GET['key']) && ($key_index = array_search($_GET['key'], $options->ap
211 175
212/////////////////////////////////////////////// 176///////////////////////////////////////////////
213// Set timezone. 177// Set timezone.
214// Prevents warnings, but needs more testing - 178// Prevents warnings, but needs more testing -
215// perhaps if timezone is set in php.ini we 179// perhaps if timezone is set in php.ini we
216// don't need to set it at all... 180// don't need to set it at all...
217/////////////////////////////////////////////// 181///////////////////////////////////////////////
@@ -233,7 +197,7 @@ if (isset($_GET['key']) && isset($_GET['hash']) && isset($options->api_keys[(int
233} 197}
234$key_index = ($valid_key) ? (int)$_GET['key'] : 0; 198$key_index = ($valid_key) ? (int)$_GET['key'] : 0;
235if (!$valid_key && $options->key_required) { 199if (!$valid_key && $options->key_required) {
236 die('A valid key must be supplied'); 200 die('A valid key must be supplied');
237} 201}
238if (!$valid_key && isset($_GET['key']) && $_GET['key'] != '') { 202if (!$valid_key && isset($_GET['key']) && $_GET['key'] != '') {
239 die('The entered key is invalid'); 203 die('The entered key is invalid');
@@ -285,6 +249,28 @@ if ($options->favour_feed_titles == 'user') {
285} 249}
286 250
287/////////////////////////////////////////////// 251///////////////////////////////////////////////
252// Include full content in output?
253///////////////////////////////////////////////
254if ($options->content === 'user') {
255 if (isset($_GET['content']) && $_GET['content'] === '0') {
256 $options->content = false;
257 } else {
258 $options->content = true;
259 }
260}
261
262///////////////////////////////////////////////
263// Include summaries in output?
264///////////////////////////////////////////////
265if ($options->summary === 'user') {
266 if (isset($_GET['summary']) && $_GET['summary'] === '1') {
267 $options->summary = true;
268 } else {
269 $options->summary = false;
270 }
271}
272
273///////////////////////////////////////////////
288// Exclude items if extraction fails 274// Exclude items if extraction fails
289/////////////////////////////////////////////// 275///////////////////////////////////////////////
290if ($options->exclude_items_on_fail === 'user') { 276if ($options->exclude_items_on_fail === 'user') {
@@ -306,15 +292,6 @@ if ($options->detect_language === 'user') {
306 $detect_language = $options->detect_language; 292 $detect_language = $options->detect_language;
307} 293}
308 294
309if ($detect_language >= 2) {
310 $language_codes = array('albanian' => 'sq','arabic' => 'ar','azeri' => 'az','bengali' => 'bn','bulgarian' => 'bg',
311 'cebuano' => 'ceb', // ISO 639-2
312 'croatian' => 'hr','czech' => 'cs','danish' => 'da','dutch' => 'nl','english' => 'en','estonian' => 'et','farsi' => 'fa','finnish' => 'fi','french' => 'fr','german' => 'de','hausa' => 'ha',
313 'hawaiian' => 'haw', // ISO 639-2
314 'hindi' => 'hi','hungarian' => 'hu','icelandic' => 'is','indonesian' => 'id','italian' => 'it','kazakh' => 'kk','kyrgyz' => 'ky','latin' => 'la','latvian' => 'lv','lithuanian' => 'lt','macedonian' => 'mk','mongolian' => 'mn','nepali' => 'ne','norwegian' => 'no','pashto' => 'ps',
315 'pidgin' => 'cpe', // ISO 639-2
316 'polish' => 'pl','portuguese' => 'pt','romanian' => 'ro','russian' => 'ru','serbian' => 'sr','slovak' => 'sk','slovene' => 'sl','somali' => 'so','spanish' => 'es','swahili' => 'sw','swedish' => 'sv','tagalog' => 'tl','turkish' => 'tr','ukrainian' => 'uk','urdu' => 'ur','uzbek' => 'uz','vietnamese' => 'vi','welsh' => 'cy');
317}
318$use_cld = extension_loaded('cld') && (version_compare(PHP_VERSION, '5.3.0') >= 0); 295$use_cld = extension_loaded('cld') && (version_compare(PHP_VERSION, '5.3.0') >= 0);
319 296
320///////////////////////////////////// 297/////////////////////////////////////
@@ -364,7 +341,7 @@ if ($options->cors) header('Access-Control-Allow-Origin: *');
364////////////////////////////////// 341//////////////////////////////////
365if ($options->caching) { 342if ($options->caching) {
366 debug('Caching is enabled...'); 343 debug('Caching is enabled...');
367 $cache_id = md5($max.$url.$valid_key.$links.$favour_feed_titles.$xss_filter.$exclude_on_fail.$format.$detect_language.(int)isset($_GET['pubsub'])); 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']));
368 $check_cache = true; 345 $check_cache = true;
369 if ($options->apc && $options->smart_cache) { 346 if ($options->apc && $options->smart_cache) {
370 apc_add("cache.$cache_id", 0, 10*60); 347 apc_add("cache.$cache_id", 0, 10*60);
@@ -415,6 +392,7 @@ if (!$debug_mode) {
415////////////////////////////////// 392//////////////////////////////////
416// Set up HTTP agent 393// Set up HTTP agent
417////////////////////////////////// 394//////////////////////////////////
395global $http;
418$http = new HumbleHttpAgent(); 396$http = new HumbleHttpAgent();
419$http->debug = $debug_mode; 397$http->debug = $debug_mode;
420$http->userAgentMap = $options->user_agents; 398$http->userAgentMap = $options->user_agents;
@@ -478,29 +456,6 @@ if ($html_only || !$result) {
478 $isDummyFeed = true; 456 $isDummyFeed = true;
479 unset($feed, $result); 457 unset($feed, $result);
480 // create single item dummy feed object 458 // create single item dummy feed object
481 class DummySingleItemFeed {
482 public $item;
483 function __construct($url) { $this->item = new DummySingleItem($url); }
484 public function get_title() { return ''; }
485 public function get_description() { return 'Content extracted from '.$this->item->url; }
486 public function get_link() { return $this->item->url; }
487 public function get_language() { return false; }
488 public function get_image_url() { return false; }
489 public function get_items($start=0, $max=1) { return array(0=>$this->item); }
490 }
491 class DummySingleItem {
492 public $url;
493 function __construct($url) { $this->url = $url; }
494 public function get_permalink() { return $this->url; }
495 public function get_title() { return null; }
496 public function get_date($format='') { return false; }
497 public function get_author($key=0) { return null; }
498 public function get_authors() { return null; }
499 public function get_description() { return ''; }
500 public function get_enclosure($key=0, $prefer=null) { return null; }
501 public function get_enclosures() { return null; }
502 public function get_categories() { return null; }
503 }
504 $feed = new DummySingleItemFeed($url); 459 $feed = new DummySingleItemFeed($url);
505} 460}
506 461
@@ -524,7 +479,7 @@ if ($img_url = $feed->get_image_url()) {
524//////////////////////////////////////////// 479////////////////////////////////////////////
525// Loop through feed items 480// Loop through feed items
526//////////////////////////////////////////// 481////////////////////////////////////////////
527$items = $feed->get_items(0, $max); 482$items = $feed->get_items(0, $max);
528// Request all feed items in parallel (if supported) 483// Request all feed items in parallel (if supported)
529$urls_sanitized = array(); 484$urls_sanitized = array();
530$urls = array(); 485$urls = array();
@@ -606,24 +561,43 @@ foreach ($items as $key => $item) {
606 $is_single_page = false; 561 $is_single_page = false;
607 if ($single_page_response = getSinglePage($item, $html, $effective_url)) { 562 if ($single_page_response = getSinglePage($item, $html, $effective_url)) {
608 $is_single_page = true; 563 $is_single_page = true;
609 $html = $single_page_response['body'];
610 // remove strange things
611 $html = str_replace('</[>', '', $html);
612 $html = convert_to_utf8($html, $single_page_response['headers']);
613 $effective_url = $single_page_response['effective_url']; 564 $effective_url = $single_page_response['effective_url'];
614 debug("Retrieved single-page view from $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 $effective_url");
586 }
615 unset($single_page_response); 587 unset($single_page_response);
616 } 588 }
589 }
590 if ($do_content_extraction) {
617 debug('--------'); 591 debug('--------');
618 debug('Attempting to extract content'); 592 debug('Attempting to extract content');
619 $extract_result = $extractor->process($html, $effective_url); 593 $extract_result = $extractor->process($html, $effective_url);
620 $readability = $extractor->readability; 594 $readability = $extractor->readability;
621 $content_block = ($extract_result) ? $extractor->getContent() : null; 595 $content_block = ($extract_result) ? $extractor->getContent() : null;
622 $extracted_title = ($extract_result) ? $extractor->getTitle() : ''; 596 $extracted_title = ($extract_result) ? $extractor->getTitle() : '';
623 // Deal with multi-page articles 597 // Deal with multi-page articles
624 //die('Next: '.$extractor->getNextPageUrl()); 598 //die('Next: '.$extractor->getNextPageUrl());
625 $is_multi_page = (!$is_single_page && $extract_result && $extractor->getNextPageUrl()); 599 $is_multi_page = (!$is_single_page && $extract_result && $extractor->getNextPageUrl());
626 if ($options->multipage && $is_multi_page) { 600 if ($options->multipage && $is_multi_page && $options->content) {
627 debug('--------'); 601 debug('--------');
628 debug('Attempting to process multi-page article'); 602 debug('Attempting to process multi-page article');
629 $multi_page_urls = array(); 603 $multi_page_urls = array();
@@ -636,7 +610,7 @@ foreach ($items as $key => $item) {
636 // check it's not what we have already! 610 // check it's not what we have already!
637 if (!in_array($next_page_url, $multi_page_urls)) { 611 if (!in_array($next_page_url, $multi_page_urls)) {
638 // it's not, so let's attempt to fetch it 612 // it's not, so let's attempt to fetch it
639 $multi_page_urls[] = $next_page_url; 613 $multi_page_urls[] = $next_page_url;
640 $_prev_ref = $http->referer; 614 $_prev_ref = $http->referer;
641 if (($response = $http->get($next_page_url, true)) && $response['status_code'] < 300) { 615 if (($response = $http->get($next_page_url, true)) && $response['status_code'] < 300) {
642 // make sure mime type is not something with a different action associated 616 // make sure mime type is not something with a different action associated
@@ -661,13 +635,15 @@ foreach ($items as $key => $item) {
661 // did we successfully deal with this multi-page article? 635 // did we successfully deal with this multi-page article?
662 if (empty($multi_page_content)) { 636 if (empty($multi_page_content)) {
663 debug('Failed to extract all parts of multi-page article, so not going to include them'); 637 debug('Failed to extract all parts of multi-page article, so not going to include them');
664 $multi_page_content[] = $readability->dom->createElement('p')->innerHTML = '<em>This article appears to continue on subsequent pages which we could not extract</em>'; 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[] = $_page;
665 } 641 }
666 foreach ($multi_page_content as $_page) { 642 foreach ($multi_page_content as $_page) {
667 $_page = $content_block->ownerDocument->importNode($_page, true); 643 $_page = $content_block->ownerDocument->importNode($_page, true);
668 $content_block->appendChild($_page); 644 $content_block->appendChild($_page);
669 } 645 }
670 unset($multi_page_urls, $multi_page_content, $page_mime_info, $next_page_url); 646 unset($multi_page_urls, $multi_page_content, $page_mime_info, $next_page_url, $_page);
671 } 647 }
672 } 648 }
673 // use extracted title for both feed and item title if we're using single-item dummy feed 649 // use extracted title for both feed and item title if we're using single-item dummy feed
@@ -695,7 +671,11 @@ foreach ($items as $key => $item) {
695 $html .= $item->get_description(); 671 $html .= $item->get_description();
696 } else { 672 } else {
697 $readability->clean($content_block, 'select'); 673 $readability->clean($content_block, 'select');
698 if ($options->rewrite_relative_urls) makeAbsolute($effective_url, $content_block); 674 // get base URL
675 $base_url = get_base_url($readability->dom);
676 if (!$base_url) $base_url = $effective_url;
677 // rewrite URLs
678 if ($options->rewrite_relative_urls) makeAbsolute($base_url, $content_block);
699 // footnotes 679 // footnotes
700 if (($links == 'footnotes') && (strpos($effective_url, 'wikipedia.org') === false)) { 680 if (($links == 'footnotes') && (strpos($effective_url, 'wikipedia.org') === false)) {
701 $readability->addFootnotes($content_block); 681 $readability->addFootnotes($content_block);
@@ -714,7 +694,7 @@ foreach ($items as $key => $item) {
714 } else { 694 } else {
715 $html = $content_block->ownerDocument->saveXML($content_block); // essentially outerHTML 695 $html = $content_block->ownerDocument->saveXML($content_block); // essentially outerHTML
716 } 696 }
717 unset($content_block); 697 //unset($content_block);
718 // post-processing cleanup 698 // post-processing cleanup
719 $html = preg_replace('!<p>[\s\h\v]*</p>!u', '', $html); 699 $html = preg_replace('!<p>[\s\h\v]*</p>!u', '', $html);
720 if ($links == 'remove') { 700 if ($links == 'remove') {
@@ -727,130 +707,155 @@ foreach ($items as $key => $item) {
727 } 707 }
728 } 708 }
729 709
730 if ($valid_key && isset($_GET['pubsub'])) { // used only on fivefilters.org at the moment 710 if ($valid_key && isset($_GET['pubsub'])) { // used only on fivefilters.org at the moment
731 $newitem->addElement('guid', 'http://fivefilters.org/content-only/redirect.php?url='.urlencode($item->get_permalink()), array('isPermaLink'=>'false')); 711 $newitem->addElement('guid', 'http://fivefilters.org/content-only/redirect.php?url='.urlencode($item->get_permalink()), array('isPermaLink'=>'false'));
712 } else {
713 $newitem->addElement('guid', $item->get_permalink(), array('isPermaLink'=>'true'));
714 }
715 // filter xss?
716 if ($xss_filter) {
717 debug('Filtering HTML to remove XSS');
718 $html = htmLawed::hl($html, array('safe'=>1, 'deny_attribute'=>'style', 'comment'=>1, 'cdata'=>1));
719 }
720
721 // add content
722 if ($options->summary === true) {
723 // get summary
724 $summary = '';
725 if (!$do_content_extraction) {
726 $summary = $html;
732 } else { 727 } else {
733 $newitem->addElement('guid', $item->get_permalink(), array('isPermaLink'=>'true')); 728 // Try to get first few paragraphs
734 } 729 if (isset($content_block) && ($content_block instanceof DOMElement)) {
735 // filter xss? 730 $_paras = $content_block->getElementsByTagName('p');
736 if ($xss_filter) { 731 foreach ($_paras as $_para) {
737 debug('Filtering HTML to remove XSS'); 732 $summary .= preg_replace("/[\n\r\t ]+/", ' ', $_para->textContent).' ';
738 $html = htmLawed::hl($html, array('safe'=>1, 'deny_attribute'=>'style', 'comment'=>1, 'cdata'=>1)); 733 if (strlen($summary) > 200) break;
739 }
740 $newitem->setDescription($html);
741
742 // set date
743 if ((int)$item->get_date('U') > 0) {
744 $newitem->setDate((int)$item->get_date('U'));
745 } elseif ($extractor->getDate()) {
746 $newitem->setDate($extractor->getDate());
747 }
748
749 // add authors
750 if ($authors = $item->get_authors()) {
751 foreach ($authors as $author) {
752 // for some feeds, SimplePie stores author's name as email, e.g. http://feeds.feedburner.com/nymag/intel
753 if ($author->get_name() !== null) {
754 $newitem->addElement('dc:creator', $author->get_name());
755 } elseif ($author->get_email() !== null) {
756 $newitem->addElement('dc:creator', $author->get_email());
757 } 734 }
735 } else {
736 $summary = $html;
758 } 737 }
759 } elseif ($authors = $extractor->getAuthors()) { 738 }
760 //TODO: make sure the list size is reasonable 739 unset($_paras, $_para);
761 foreach ($authors as $author) { 740 $summary = get_excerpt($summary);
762 // TODO: xpath often selects authors from other articles linked from the page. 741 $newitem->setDescription($summary);
763 // for now choose first item 742 if ($options->content) $newitem->setElement('content:encoded', $html);
764 $newitem->addElement('dc:creator', $author); 743 } else {
765 break; 744 if ($options->content) $newitem->setDescription($html);
745 }
746
747 // set date
748 if ((int)$item->get_date('U') > 0) {
749 $newitem->setDate((int)$item->get_date('U'));
750 } elseif ($extractor->getDate()) {
751 $newitem->setDate($extractor->getDate());
752 }
753
754 // add authors
755 if ($authors = $item->get_authors()) {
756 foreach ($authors as $author) {
757 // for some feeds, SimplePie stores author's name as email, e.g. http://feeds.feedburner.com/nymag/intel
758 if ($author->get_name() !== null) {
759 $newitem->addElement('dc:creator', $author->get_name());
760 } elseif ($author->get_email() !== null) {
761 $newitem->addElement('dc:creator', $author->get_email());
766 } 762 }
767 } 763 }
768 764 } elseif ($authors = $extractor->getAuthors()) {
769 // add language 765 //TODO: make sure the list size is reasonable
770 if ($detect_language) { 766 foreach ($authors as $author) {
771 $language = $extractor->getLanguage(); 767 // TODO: xpath often selects authors from other articles linked from the page.
772 if (!$language) $language = $feed->get_language(); 768 // for now choose first item
773 if (($detect_language == 3 || (!$language && $detect_language == 2)) && $text_sample) { 769 $newitem->addElement('dc:creator', $author);
774 try { 770 break;
775 if ($use_cld) { 771 }
776 // Use PHP-CLD extension 772 }
777 $php_cld = 'CLD\detect'; // in quotes to prevent PHP 5.2 parse error 773
778 $res = $php_cld($text_sample); 774 // add language
779 if (is_array($res) && count($res) > 0) { 775 if ($detect_language) {
780 $language = $res[0]['code']; 776 $language = $extractor->getLanguage();
781 } 777 if (!$language) $language = $feed->get_language();
782 } else { 778 if (($detect_language == 3 || (!$language && $detect_language == 2)) && $text_sample) {
783 //die('what'); 779 try {
784 // Use PEAR's Text_LanguageDetect 780 if ($use_cld) {
785 if (!isset($l)) { 781 // Use PHP-CLD extension
786 $l = new Text_LanguageDetect('libraries/language-detect/lang.dat', 'libraries/language-detect/unicode_blocks.dat'); 782 $php_cld = 'CLD\detect'; // in quotes to prevent PHP 5.2 parse error
787 } 783 $res = $php_cld($text_sample);
788 $l_result = $l->detect($text_sample, 1); 784 if (is_array($res) && count($res) > 0) {
789 if (count($l_result) > 0) { 785 $language = $res[0]['code'];
790 $language = $language_codes[key($l_result)]; 786 }
791 } 787 } else {
788 //die('what');
789 // Use PEAR's Text_LanguageDetect
790 if (!isset($l)) {
791 $l = new Text_LanguageDetect();
792 $l->setNameMode(2); // return ISO 639-1 codes (e.g. "en")
793 }
794 $l_result = $l->detect($text_sample, 1);
795 if (count($l_result) > 0) {
796 $language = key($l_result);
792 } 797 }
793 } catch (Exception $e) {
794 //die('error: '.$e);
795 // do nothing
796 } 798 }
797 } 799 } catch (Exception $e) {
798 if ($language && (strlen($language) < 7)) { 800 //die('error: '.$e);
799 $newitem->addElement('dc:language', $language); 801 // do nothing
800 } 802 }
801 } 803 }
802 804 if ($language && (strlen($language) < 7)) {
803 // add MIME type (if it appeared in our exclusions lists) 805 $newitem->addElement('dc:language', $language);
804 if (isset($mime_info['mime'])) $newitem->addElement('dc:format', $mime_info['mime']);
805 // add effective URL (URL after redirects)
806 if (isset($effective_url)) {
807 //TODO: ensure $effective_url is valid witout - sometimes it causes problems, e.g.
808 //http://www.siasat.pk/forum/showthread.php?108883-Pakistan-Chowk-by-Rana-Mubashir-–-25th-March-2012-Special-Program-from-Liari-(Karachi)
809 //temporary measure: use utf8_encode()
810 $newitem->addElement('dc:identifier', remove_url_cruft(utf8_encode($effective_url)));
811 } else {
812 $newitem->addElement('dc:identifier', remove_url_cruft($item->get_permalink()));
813 } 806 }
814 807 }
815 // add categories 808
816 if ($categories = $item->get_categories()) { 809 // add MIME type (if it appeared in our exclusions lists)
817 foreach ($categories as $category) { 810 if (isset($mime_info['mime'])) $newitem->addElement('dc:format', $mime_info['mime']);
818 if ($category->get_label() !== null) { 811 // add effective URL (URL after redirects)
819 $newitem->addElement('category', $category->get_label()); 812 if (isset($effective_url)) {
820 } 813 //TODO: ensure $effective_url is valid witout - sometimes it causes problems, e.g.
814 //http://www.siasat.pk/forum/showthread.php?108883-Pakistan-Chowk-by-Rana-Mubashir-�-25th-March-2012-Special-Program-from-Liari-(Karachi)
815 //temporary measure: use utf8_encode()
816 $newitem->addElement('dc:identifier', remove_url_cruft(utf8_encode($effective_url)));
817 } else {
818 $newitem->addElement('dc:identifier', remove_url_cruft($item->get_permalink()));
819 }
820
821 // add categories
822 if ($categories = $item->get_categories()) {
823 foreach ($categories as $category) {
824 if ($category->get_label() !== null) {
825 $newitem->addElement('category', $category->get_label());
821 } 826 }
822 } 827 }
823 828 }
824 // check for enclosures 829
825 if ($options->keep_enclosures) { 830 // check for enclosures
826 if ($enclosures = $item->get_enclosures()) { 831 if ($options->keep_enclosures) {
827 foreach ($enclosures as $enclosure) { 832 if ($enclosures = $item->get_enclosures()) {
828 // thumbnails 833 foreach ($enclosures as $enclosure) {
829 foreach ((array)$enclosure->get_thumbnails() as $thumbnail) { 834 // thumbnails
830 $newitem->addElement('media:thumbnail', '', array('url'=>$thumbnail)); 835 foreach ((array)$enclosure->get_thumbnails() as $thumbnail) {
831 } 836 $newitem->addElement('media:thumbnail', '', array('url'=>$thumbnail));
832 if (!$enclosure->get_link()) continue;
833 $enc = array();
834 // Media RSS spec ($enc): http://search.yahoo.com/mrss
835 // SimplePie methods ($enclosure): http://simplepie.org/wiki/reference/start#methods4
836 $enc['url'] = $enclosure->get_link();
837 if ($enclosure->get_length()) $enc['fileSize'] = $enclosure->get_length();
838 if ($enclosure->get_type()) $enc['type'] = $enclosure->get_type();
839 if ($enclosure->get_medium()) $enc['medium'] = $enclosure->get_medium();
840 if ($enclosure->get_expression()) $enc['expression'] = $enclosure->get_expression();
841 if ($enclosure->get_bitrate()) $enc['bitrate'] = $enclosure->get_bitrate();
842 if ($enclosure->get_framerate()) $enc['framerate'] = $enclosure->get_framerate();
843 if ($enclosure->get_sampling_rate()) $enc['samplingrate'] = $enclosure->get_sampling_rate();
844 if ($enclosure->get_channels()) $enc['channels'] = $enclosure->get_channels();
845 if ($enclosure->get_duration()) $enc['duration'] = $enclosure->get_duration();
846 if ($enclosure->get_height()) $enc['height'] = $enclosure->get_height();
847 if ($enclosure->get_width()) $enc['width'] = $enclosure->get_width();
848 if ($enclosure->get_language()) $enc['lang'] = $enclosure->get_language();
849 $newitem->addElement('media:content', '', $enc);
850 } 837 }
838 if (!$enclosure->get_link()) continue;
839 $enc = array();
840 // Media RSS spec ($enc): http://search.yahoo.com/mrss
841 // SimplePie methods ($enclosure): http://simplepie.org/wiki/reference/start#methods4
842 $enc['url'] = $enclosure->get_link();
843 if ($enclosure->get_length()) $enc['fileSize'] = $enclosure->get_length();
844 if ($enclosure->get_type()) $enc['type'] = $enclosure->get_type();
845 if ($enclosure->get_medium()) $enc['medium'] = $enclosure->get_medium();
846 if ($enclosure->get_expression()) $enc['expression'] = $enclosure->get_expression();
847 if ($enclosure->get_bitrate()) $enc['bitrate'] = $enclosure->get_bitrate();
848 if ($enclosure->get_framerate()) $enc['framerate'] = $enclosure->get_framerate();
849 if ($enclosure->get_sampling_rate()) $enc['samplingrate'] = $enclosure->get_sampling_rate();
850 if ($enclosure->get_channels()) $enc['channels'] = $enclosure->get_channels();
851 if ($enclosure->get_duration()) $enc['duration'] = $enclosure->get_duration();
852 if ($enclosure->get_height()) $enc['height'] = $enclosure->get_height();
853 if ($enclosure->get_width()) $enc['width'] = $enclosure->get_width();
854 if ($enclosure->get_language()) $enc['lang'] = $enclosure->get_language();
855 $newitem->addElement('media:content', '', $enc);
851 } 856 }
852 } 857 }
853 /* } */ 858 }
854 $output->addItem($newitem); 859 $output->addItem($newitem);
855 unset($html); 860 unset($html);
856 $item_count++; 861 $item_count++;
@@ -887,7 +892,7 @@ if (!$debug_mode) {
887 } 892 }
888 if ($add_to_cache) { 893 if ($add_to_cache) {
889 ob_start(); 894 ob_start();
890 $output->genarateFeed(); 895 $output->genarateFeed(false);
891 $output = ob_get_contents(); 896 $output = ob_get_contents();
892 ob_end_clean(); 897 ob_end_clean();
893 if ($html_only && $item_count == 0) { 898 if ($html_only && $item_count == 0) {
@@ -898,299 +903,8 @@ if (!$debug_mode) {
898 } 903 }
899 echo $output; 904 echo $output;
900 } else { 905 } else {
901 $output->genarateFeed(); 906 $output->genarateFeed(false);
902 } 907 }
903 if ($callback) echo ');'; 908 if ($callback) echo ');';
904} 909}
905 910
906///////////////////////////////
907// HELPER FUNCTIONS
908///////////////////////////////
909
910function url_allowed($url) {
911 global $options;
912 if (!empty($options->allowed_urls)) {
913 $allowed = false;
914 foreach ($options->allowed_urls as $allowurl) {
915 if (stristr($url, $allowurl) !== false) {
916 $allowed = true;
917 break;
918 }
919 }
920 if (!$allowed) return false;
921 } else {
922 foreach ($options->blocked_urls as $blockurl) {
923 if (stristr($url, $blockurl) !== false) {
924 return false;
925 }
926 }
927 }
928 return true;
929}
930
931//////////////////////////////////////////////
932// Convert $html to UTF8
933// (uses HTTP headers and HTML to find encoding)
934// adapted from http://stackoverflow.com/questions/910793/php-detect-encoding-and-make-everything-utf-8
935//////////////////////////////////////////////
936function convert_to_utf8($html, $header=null)
937{
938 $encoding = null;
939 if ($html || $header) {
940 if (is_array($header)) $header = implode("\n", $header);
941 if (!$header || !preg_match_all('/^Content-Type:\s+([^;]+)(?:;\s*charset=["\']?([^;"\'\n]*))?/im', $header, $match, PREG_SET_ORDER)) {
942 // error parsing the response
943 debug('Could not find Content-Type header in HTTP response');
944 } else {
945 $match = end($match); // get last matched element (in case of redirects)
946 if (isset($match[2])) $encoding = trim($match[2], "\"' \r\n\0\x0B\t");
947 }
948 // TODO: check to see if encoding is supported (can we convert it?)
949 // If it's not, result will be empty string.
950 // For now we'll check for invalid encoding types returned by some sites, e.g. 'none'
951 // Problem URL: http://facta.co.jp/blog/archives/20111026001026.html
952 if (!$encoding || $encoding == 'none') {
953 // search for encoding in HTML - only look at the first 50000 characters
954 // Why 50000? See, for example, http://www.lemonde.fr/festival-de-cannes/article/2012/05/23/deux-cretes-en-goguette-sur-la-croisette_1705732_766360.html
955 // TODO: improve this so it looks at smaller chunks first
956 $html_head = substr($html, 0, 50000);
957 if (preg_match('/^<\?xml\s+version=(?:"[^"]*"|\'[^\']*\')\s+encoding=("[^"]*"|\'[^\']*\')/s', $html_head, $match)) {
958 $encoding = trim($match[1], '"\'');
959 } elseif (preg_match('/<meta\s+http-equiv=["\']?Content-Type["\']? content=["\'][^;]+;\s*charset=["\']?([^;"\'>]+)/i', $html_head, $match)) {
960 $encoding = trim($match[1]);
961 } elseif (preg_match_all('/<meta\s+([^>]+)>/i', $html_head, $match)) {
962 foreach ($match[1] as $_test) {
963 if (preg_match('/charset=["\']?([^"\']+)/i', $_test, $_m)) {
964 $encoding = trim($_m[1]);
965 break;
966 }
967 }
968 }
969 }
970 if (isset($encoding)) $encoding = trim($encoding);
971 // trim is important here!
972 if (!$encoding || (strtolower($encoding) == 'iso-8859-1')) {
973 // replace MS Word smart qutoes
974 $trans = array();
975 $trans[chr(130)] = '&sbquo;'; // Single Low-9 Quotation Mark
976 $trans[chr(131)] = '&fnof;'; // Latin Small Letter F With Hook
977 $trans[chr(132)] = '&bdquo;'; // Double Low-9 Quotation Mark
978 $trans[chr(133)] = '&hellip;'; // Horizontal Ellipsis
979 $trans[chr(134)] = '&dagger;'; // Dagger
980 $trans[chr(135)] = '&Dagger;'; // Double Dagger
981 $trans[chr(136)] = '&circ;'; // Modifier Letter Circumflex Accent
982 $trans[chr(137)] = '&permil;'; // Per Mille Sign
983 $trans[chr(138)] = '&Scaron;'; // Latin Capital Letter S With Caron
984 $trans[chr(139)] = '&lsaquo;'; // Single Left-Pointing Angle Quotation Mark
985 $trans[chr(140)] = '&OElig;'; // Latin Capital Ligature OE
986 $trans[chr(145)] = '&lsquo;'; // Left Single Quotation Mark
987 $trans[chr(146)] = '&rsquo;'; // Right Single Quotation Mark
988 $trans[chr(147)] = '&ldquo;'; // Left Double Quotation Mark
989 $trans[chr(148)] = '&rdquo;'; // Right Double Quotation Mark
990 $trans[chr(149)] = '&bull;'; // Bullet
991 $trans[chr(150)] = '&ndash;'; // En Dash
992 $trans[chr(151)] = '&mdash;'; // Em Dash
993 $trans[chr(152)] = '&tilde;'; // Small Tilde
994 $trans[chr(153)] = '&trade;'; // Trade Mark Sign
995 $trans[chr(154)] = '&scaron;'; // Latin Small Letter S With Caron
996 $trans[chr(155)] = '&rsaquo;'; // Single Right-Pointing Angle Quotation Mark
997 $trans[chr(156)] = '&oelig;'; // Latin Small Ligature OE
998 $trans[chr(159)] = '&Yuml;'; // Latin Capital Letter Y With Diaeresis
999 $html = strtr($html, $trans);
1000 }
1001 if (!$encoding) {
1002 debug('No character encoding found, so treating as UTF-8');
1003 $encoding = 'utf-8';
1004 } else {
1005 debug('Character encoding: '.$encoding);
1006 if (strtolower($encoding) != 'utf-8') {
1007 debug('Converting to UTF-8');
1008 $html = SimplePie_Misc::change_encoding($html, $encoding, 'utf-8');
1009 /*
1010 if (function_exists('iconv')) {
1011 // iconv appears to handle certain character encodings better than mb_convert_encoding
1012 $html = iconv($encoding, 'utf-8', $html);
1013 } else {
1014 $html = mb_convert_encoding($html, 'utf-8', $encoding);
1015 }
1016 */
1017 }
1018 }
1019 }
1020 return $html;
1021}
1022
1023function makeAbsolute($base, $elem) {
1024 $base = new SimplePie_IRI($base);
1025 // remove '//' in URL path (used to prevent URLs from resolving properly)
1026 // TODO: check if this is still the case
1027 if (isset($base->path)) $base->path = preg_replace('!//+!', '/', $base->path);
1028 foreach(array('a'=>'href', 'img'=>'src') as $tag => $attr) {
1029 $elems = $elem->getElementsByTagName($tag);
1030 for ($i = $elems->length-1; $i >= 0; $i--) {
1031 $e = $elems->item($i);
1032 //$e->parentNode->replaceChild($articleContent->ownerDocument->createTextNode($e->textContent), $e);
1033 makeAbsoluteAttr($base, $e, $attr);
1034 }
1035 if (strtolower($elem->tagName) == $tag) makeAbsoluteAttr($base, $elem, $attr);
1036 }
1037}
1038function makeAbsoluteAttr($base, $e, $attr) {
1039 if ($e->hasAttribute($attr)) {
1040 // Trim leading and trailing white space. I don't really like this but
1041 // unfortunately it does appear on some sites. e.g. <img src=" /path/to/image.jpg" />
1042 $url = trim(str_replace('%20', ' ', $e->getAttribute($attr)));
1043 $url = str_replace(' ', '%20', $url);
1044 if (!preg_match('!https?://!i', $url)) {
1045 if ($absolute = SimplePie_IRI::absolutize($base, $url)) {
1046 $e->setAttribute($attr, $absolute);
1047 }
1048 }
1049 }
1050}
1051function makeAbsoluteStr($base, $url) {
1052 $base = new SimplePie_IRI($base);
1053 // remove '//' in URL path (causes URLs not to resolve properly)
1054 if (isset($base->path)) $base->path = preg_replace('!//+!', '/', $base->path);
1055 if (preg_match('!^https?://!i', $url)) {
1056 // already absolute
1057 return $url;
1058 } else {
1059 if ($absolute = SimplePie_IRI::absolutize($base, $url)) {
1060 return $absolute;
1061 }
1062 return false;
1063 }
1064}
1065// returns single page response, or false if not found
1066function getSinglePage($item, $html, $url) {
1067 global $http, $extractor;
1068 debug('Looking for site config files to see if single page link exists');
1069 $site_config = $extractor->buildSiteConfig($url, $html);
1070 $splink = null;
1071 if (!empty($site_config->single_page_link)) {
1072 $splink = $site_config->single_page_link;
1073 } elseif (!empty($site_config->single_page_link_in_feed)) {
1074 // single page link xpath is targeted at feed
1075 $splink = $site_config->single_page_link_in_feed;
1076 // so let's replace HTML with feed item description
1077 $html = $item->get_description();
1078 }
1079 if (isset($splink)) {
1080 // Build DOM tree from HTML
1081 $readability = new Readability($html, $url);
1082 $xpath = new DOMXPath($readability->dom);
1083 // Loop through single_page_link xpath expressions
1084 $single_page_url = null;
1085 foreach ($splink as $pattern) {
1086 $elems = @$xpath->evaluate($pattern, $readability->dom);
1087 if (is_string($elems)) {
1088 $single_page_url = trim($elems);
1089 break;
1090 } elseif ($elems instanceof DOMNodeList && $elems->length > 0) {
1091 foreach ($elems as $item) {
1092 if ($item instanceof DOMElement && $item->hasAttribute('href')) {
1093 $single_page_url = $item->getAttribute('href');
1094 break 2;
1095 } elseif ($item instanceof DOMAttr && $item->value) {
1096 $single_page_url = $item->value;
1097 break 2;
1098 }
1099 }
1100 }
1101 }
1102 // If we've got URL, resolve against $url
1103 if (isset($single_page_url) && ($single_page_url = makeAbsoluteStr($url, $single_page_url))) {
1104 // check it's not what we have already!
1105 if ($single_page_url != $url) {
1106 // it's not, so let's try to fetch it...
1107 $_prev_ref = $http->referer;
1108 $http->referer = $single_page_url;
1109 if (($response = $http->get($single_page_url, true)) && $response['status_code'] < 300) {
1110 $http->referer = $_prev_ref;
1111 return $response;
1112 }
1113 $http->referer = $_prev_ref;
1114 }
1115 }
1116 }
1117 return false;
1118}
1119
1120// based on content-type http header, decide what to do
1121// param: HTTP headers string
1122// return: array with keys: 'mime', 'type', 'subtype', 'action', 'name'
1123// e.g. array('mime'=>'image/jpeg', 'type'=>'image', 'subtype'=>'jpeg', 'action'=>'link', 'name'=>'Image')
1124function get_mime_action_info($headers) {
1125 global $options;
1126 // check if action defined for returned Content-Type
1127 $info = array();
1128 if (preg_match('!^Content-Type:\s*(([-\w]+)/([-\w\+]+))!im', $headers, $match)) {
1129 // look for full mime type (e.g. image/jpeg) or just type (e.g. image)
1130 // match[1] = full mime type, e.g. image/jpeg
1131 // match[2] = first part, e.g. image
1132 // match[3] = last part, e.g. jpeg
1133 $info['mime'] = strtolower(trim($match[1]));
1134 $info['type'] = strtolower(trim($match[2]));
1135 $info['subtype'] = strtolower(trim($match[3]));
1136 foreach (array($info['mime'], $info['type']) as $_mime) {
1137 if (isset($options->content_type_exc[$_mime])) {
1138 $info['action'] = $options->content_type_exc[$_mime]['action'];
1139 $info['name'] = $options->content_type_exc[$_mime]['name'];
1140 break;
1141 }
1142 }
1143 }
1144 return $info;
1145}
1146
1147function remove_url_cruft($url) {
1148 // remove google analytics for the time being
1149 // regex adapted from http://navitronic.co.uk/2010/12/removing-google-analytics-cruft-from-urls/
1150 // https://gist.github.com/758177
1151 return preg_replace('/(\?|\&)utm_[a-z]+=[^\&]+/', '', $url);
1152}
1153
1154function make_substitutions($string) {
1155 if ($string == '') return $string;
1156 global $item, $effective_url;
1157 $string = str_replace('{url}', htmlspecialchars($item->get_permalink()), $string);
1158 $string = str_replace('{effective-url}', htmlspecialchars($effective_url), $string);
1159 return $string;
1160}
1161
1162function get_cache() {
1163 global $options, $valid_key;
1164 static $cache = null;
1165 if ($cache === null) {
1166 $frontendOptions = array(
1167 'lifetime' => 10*60, // cache lifetime of 10 minutes
1168 'automatic_serialization' => false,
1169 'write_control' => false,
1170 'automatic_cleaning_factor' => $options->cache_cleanup,
1171 'ignore_user_abort' => false
1172 );
1173 $backendOptions = array(
1174 'cache_dir' => ($valid_key) ? $options->cache_dir.'/rss-with-key/' : $options->cache_dir.'/rss/', // directory where to put the cache files
1175 'file_locking' => false,
1176 'read_control' => true,
1177 'read_control_type' => 'strlen',
1178 'hashed_directory_level' => $options->cache_directory_level,
1179 'hashed_directory_perm' => 0777,
1180 'cache_file_perm' => 0664,
1181 'file_name_prefix' => 'ff'
1182 );
1183 // getting a Zend_Cache_Core object
1184 $cache = Zend_Cache::factory('Core', 'File', $frontendOptions, $backendOptions);
1185 }
1186 return $cache;
1187}
1188
1189function debug($msg) {
1190 global $debug_mode;
1191 if ($debug_mode) {
1192 echo '* ',$msg,"\n";
1193 ob_flush();
1194 flush();
1195 }
1196} \ No newline at end of file
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-06 16:18:15 +0000