X-Git-Url: https://git.mxchange.org/?a=blobdiff_plain;f=include%2FScrape.php;h=95a3c0221f51ac32ca89017e4eeb52fceaa24101;hb=b0548018d82cf6ecc717b3382ca5e699631458e9;hp=c2917d589355ce9e318226e6cb89a1c054bf8842;hpb=caa21e6021dc2623aba8294f55279a31c94aab02;p=friendica.git diff --git a/include/Scrape.php b/include/Scrape.php index c2917d5893..95a3c0221f 100644 --- a/include/Scrape.php +++ b/include/Scrape.php @@ -2,9 +2,10 @@ require_once('library/HTML5/Parser.php'); require_once('include/crypto.php'); +require_once('include/feed.php'); if(! function_exists('scrape_dfrn')) { -function scrape_dfrn($url) { +function scrape_dfrn($url, $dont_probe = false) { $a = get_app(); @@ -17,6 +18,13 @@ function scrape_dfrn($url) { if(! $s) return $ret; + if (!$dont_probe) { + $probe = probe_url($url); + + if (isset($probe["addr"])) + $ret["addr"] = $probe["addr"]; + } + $headers = $a->get_curl_headers(); logger('scrape_dfrn: headers=' . $headers, LOGGER_DEBUG); @@ -228,7 +236,9 @@ function scrape_feed($url) { $a = get_app(); $ret = array(); - $s = fetch_url($url); + $cookiejar = tempnam(get_temppath(), 'cookiejar-scrape-feed-'); + $s = fetch_url($url, false, $redirects, 0, Null, $cookiejar); + unlink($cookiejar); $headers = $a->get_curl_headers(); $code = $a->get_curl_code(); @@ -249,7 +259,7 @@ function scrape_feed($url) { $ret['feed_atom'] = $url; return $ret; } - if(stristr($line,'application/rss+xml') || stristr($s,'loadHTML($s); + $xpath = new DomXPath($doc); + $base = $xpath->query("//base"); + foreach ($base as $node) { + $attr = array(); - $head = $dom->getElementsByTagName('base'); - if($head) { - foreach($head as $head0) { - $basename = $head0->getAttribute('href'); - break; - } + if ($node->attributes->length) + foreach ($node->attributes as $attribute) + $attr[$attribute->name] = $attribute->value; + + if ($attr["href"] != "") + $basename = $attr["href"] ; } - if(! $basename) - $basename = implode('/', array_slice(explode('/',$url),0,3)) . '/'; - $items = $dom->getElementsByTagName('link'); + $list = $xpath->query("//link"); + foreach ($list as $node) { + $attr = array(); - // get Atom/RSS link elements, take the first one of either. + if ($node->attributes->length) + foreach ($node->attributes as $attribute) + $attr[$attribute->name] = $attribute->value; - if($items) { - foreach($items as $item) { - $x = $item->getAttribute('rel'); - if(($x === 'alternate') && ($item->getAttribute('type') === 'application/atom+xml')) { - if(! x($ret,'feed_atom')) - $ret['feed_atom'] = $item->getAttribute('href'); - } - if(($x === 'alternate') && ($item->getAttribute('type') === 'application/rss+xml')) { - if(! x($ret,'feed_rss')) - $ret['feed_rss'] = $item->getAttribute('href'); - } - } + if (($attr["rel"] == "alternate") AND ($attr["type"] == "application/atom+xml")) + $ret["feed_atom"] = $attr["href"]; + + if (($attr["rel"] == "alternate") AND ($attr["type"] == "application/rss+xml")) + $ret["feed_rss"] = $attr["href"]; } - // Drupal and perhaps others only provide relative URL's. Turn them into absolute. + // Drupal and perhaps others only provide relative URLs. Turn them into absolute. if(x($ret,'feed_atom') && (! strstr($ret['feed_atom'],'://'))) $ret['feed_atom'] = $basename . $ret['feed_atom']; @@ -327,7 +330,7 @@ function scrape_feed($url) { * PROBE_DIASPORA has a bias towards returning Diaspora information * while PROBE_NORMAL has a bias towards dfrn/zot - in the case where * an address (such as a Friendica address) supports more than one type - * of network. + * of network. * */ @@ -335,7 +338,7 @@ function scrape_feed($url) { define ( 'PROBE_NORMAL', 0); define ( 'PROBE_DIASPORA', 1); -function probe_url($url, $mode = PROBE_NORMAL) { +function probe_url($url, $mode = PROBE_NORMAL, $level = 1) { require_once('include/email.php'); $result = array(); @@ -364,19 +367,6 @@ function probe_url($url, $mode = PROBE_NORMAL) { $network = NETWORK_TWITTER; } - if (strpos($url,'www.facebook.com')) { - $connectornetworks = true; - $network = NETWORK_FACEBOOK; - } - - if (strpos($url,'alpha.app.net')) { - $appnet = true; - $network = NETWORK_APPNET; - } - - - // Twitter is deactivated since twitter closed its old API - //$twitter = ((strpos($url,'twitter.com') !== false) ? true : false); $lastfm = ((strpos($url,'last.fm/user') !== false) ? true : false); $at_addr = ((strpos($url,'@') !== false) ? true : false); @@ -425,7 +415,7 @@ function probe_url($url, $mode = PROBE_NORMAL) { $pubkey = $diaspora_key; $diaspora = true; } - if($link['@attributes']['rel'] === 'http://ostatus.org/schema/1.0/subscribe') { + if(($link['@attributes']['rel'] === 'http://ostatus.org/schema/1.0/subscribe') AND ($mode == PROBE_NORMAL)) { $diaspora = false; } } @@ -520,6 +510,7 @@ function probe_url($url, $mode = PROBE_NORMAL) { } if($mode == PROBE_NORMAL) { + if(strlen($zot)) { $s = fetch_url($zot); if($s) { @@ -539,8 +530,9 @@ function probe_url($url, $mode = PROBE_NORMAL) { } } + if(strlen($dfrn)) { - $ret = scrape_dfrn(($hcard) ? $hcard : $dfrn); + $ret = scrape_dfrn(($hcard) ? $hcard : $dfrn, true); if(is_array($ret) && x($ret,'dfrn-request')) { $network = NETWORK_DFRN; $request = $ret['dfrn-request']; @@ -556,9 +548,23 @@ function probe_url($url, $mode = PROBE_NORMAL) { } } + // Scrape the public key from the hcard. + // Diaspora will remove it from the webfinger somewhere in the future. + if (($hcard != "") AND ($pubkey == "")) { + $ret = scrape_dfrn(($hcard) ? $hcard : $dfrn, true); + if (isset($ret["key"])) { + $hcard_key = $ret["key"]; + if(strstr($hcard_key,'RSA ')) + $pubkey = rsatopem($hcard_key); + else + $pubkey = $hcard_key; + } + } if($diaspora && $diaspora_base && $diaspora_guid) { - if($mode == PROBE_DIASPORA || ! $notify) { - $notify = $diaspora_base . 'receive/users/' . $diaspora_guid; + $diaspora_notify = $diaspora_base.'receive/users/'.$diaspora_guid; + + if($mode == PROBE_DIASPORA || ! $notify || ($notify == $diaspora_notify)) { + $notify = $diaspora_notify; $batch = $diaspora_base . 'receive/public' ; } if(strpos($url,'@')) @@ -568,7 +574,7 @@ function probe_url($url, $mode = PROBE_NORMAL) { if($network !== NETWORK_ZOT && $network !== NETWORK_DFRN && $network !== NETWORK_MAIL) { if($diaspora) $network = NETWORK_DIASPORA; - elseif($has_lrdd) + elseif($has_lrdd AND ($notify)) $network = NETWORK_OSTATUS; if(strpos($url,'@')) @@ -597,21 +603,6 @@ function probe_url($url, $mode = PROBE_NORMAL) { $vcard['nick'] = $addr_parts[0]; } - /* if($twitter) { - logger('twitter: setup'); - $tid = basename($url); - $tapi = 'https://api.twitter.com/1/statuses/user_timeline.rss'; - if(intval($tid)) - $poll = $tapi . '?user_id=' . $tid; - else - $poll = $tapi . '?screen_name=' . $tid; - $profile = 'http://twitter.com/#!/' . $tid; - //$vcard['photo'] = 'https://api.twitter.com/1/users/profile_image/' . $tid; - $vcard['photo'] = 'https://api.twitter.com/1/users/profile_image?screen_name=' . $tid . '&size=bigger'; - $vcard['nick'] = $tid; - $vcard['fn'] = $tid; - } */ - if($lastfm) { $profile = $url; $poll = str_replace(array('www.','last.fm/'),array('','ws.audioscrobbler.com/1.0/'),$url) . '/recenttracks.rss'; @@ -645,6 +636,7 @@ function probe_url($url, $mode = PROBE_NORMAL) { if($check_feed) { $feedret = scrape_feed(($poll) ? $poll : $url); + logger('probe_url: scrape_feed ' . (($poll)? $poll : $url) . ' returns: ' . print_r($feedret,true), LOGGER_DATA); if(count($feedret) && ($feedret['feed_atom'] || $feedret['feed_rss'])) { $poll = ((x($feedret,'feed_atom')) ? unamp($feedret['feed_atom']) : unamp($feedret['feed_rss'])); @@ -654,96 +646,79 @@ function probe_url($url, $mode = PROBE_NORMAL) { if(x($feedret,'photo') && (! x($vcard,'photo'))) $vcard['photo'] = $feedret['photo']; - require_once('library/simplepie/simplepie.inc'); - $feed = new SimplePie(); - $xml = fetch_url($poll); + + $cookiejar = tempnam(get_temppath(), 'cookiejar-scrape-feed-'); + $xml = fetch_url($poll, false, $redirects, 0, Null, $cookiejar); + unlink($cookiejar); logger('probe_url: fetch feed: ' . $poll . ' returns: ' . $xml, LOGGER_DATA); - $a = get_app(); - - logger('probe_url: scrape_feed: headers: ' . $a->get_curl_headers(), LOGGER_DATA); - - // Don't try and parse an empty string - $feed->set_raw_data(($xml) ? $xml : ''); - - $feed->init(); - if($feed->error()) - logger('probe_url: scrape_feed: Error parsing XML: ' . $feed->error()); - - - if(! x($vcard,'photo')) - $vcard['photo'] = $feed->get_image_url(); - $author = $feed->get_author(); - - if($author) { - $vcard['fn'] = unxmlify(trim($author->get_name())); - if(! $vcard['fn']) - $vcard['fn'] = trim(unxmlify($author->get_email())); - if(strpos($vcard['fn'],'@') !== false) - $vcard['fn'] = substr($vcard['fn'],0,strpos($vcard['fn'],'@')); - $email = unxmlify($author->get_email()); - if(! $profile && $author->get_link()) - $profile = trim(unxmlify($author->get_link())); - if(! $vcard['photo']) { - $rawtags = $feed->get_feed_tags( SIMPLEPIE_NAMESPACE_ATOM_10, 'author'); - if($rawtags) { - $elems = $rawtags[0]['child'][SIMPLEPIE_NAMESPACE_ATOM_10]; - if((x($elems,'link')) && ($elems['link'][0]['attribs']['']['rel'] === 'photo')) - $vcard['photo'] = $elems['link'][0]['attribs']['']['href']; - } + + if ($xml == "") { + logger("scrape_feed: XML is empty for feed ".$poll); + $network = NETWORK_PHANTOM; + } else { + $data = feed_import($xml,$dummy1,$dummy2, $dummy3, true); + + if (!is_array($data)) { + logger("scrape_feed: This doesn't seem to be a feed: ".$poll); + $network = NETWORK_PHANTOM; + } else { + if (($vcard["photo"] == "") AND ($data["header"]["author-avatar"] != "")) + $vcard["photo"] = $data["header"]["author-avatar"]; + + if (($vcard["fn"] == "") AND ($data["header"]["author-name"] != "")) + $vcard["fn"] = $data["header"]["author-name"]; + + if (($vcard["nick"] == "") AND ($data["header"]["author-nick"] != "")) + $vcard["nick"] = $data["header"]["author-nick"]; + + if(!$profile AND ($data["header"]["author-link"] != "") AND !in_array($network, array("", NETWORK_FEED))) + $profile = $data["header"]["author-link"]; } } - else { - $item = $feed->get_item(0); - if($item) { - $author = $item->get_author(); - if($author) { - $vcard['fn'] = trim(unxmlify($author->get_name())); - if(! $vcard['fn']) - $vcard['fn'] = trim(unxmlify($author->get_email())); - if(strpos($vcard['fn'],'@') !== false) - $vcard['fn'] = substr($vcard['fn'],0,strpos($vcard['fn'],'@')); - $email = unxmlify($author->get_email()); - if(! $profile && $author->get_link()) - $profile = trim(unxmlify($author->get_link())); - } - if(! $vcard['photo']) { - $rawmedia = $item->get_item_tags('http://search.yahoo.com/mrss/','thumbnail'); - if($rawmedia && $rawmedia[0]['attribs']['']['url']) - $vcard['photo'] = unxmlify($rawmedia[0]['attribs']['']['url']); - } - if(! $vcard['photo']) { - $rawtags = $item->get_item_tags( SIMPLEPIE_NAMESPACE_ATOM_10, 'author'); - if($rawtags) { - $elems = $rawtags[0]['child'][SIMPLEPIE_NAMESPACE_ATOM_10]; - if((x($elems,'link')) && ($elems['link'][0]['attribs']['']['rel'] === 'photo')) - $vcard['photo'] = $elems['link'][0]['attribs']['']['href']; - } - } + + // Workaround for misconfigured Friendica servers + if (($network == "") AND (strstr($url, "/profile/"))) { + $noscrape = str_replace("/profile/", "/noscrape/", $url); + $noscrapejson = fetch_url($noscrape); + if ($noscrapejson) { + + $network = NETWORK_DFRN; + + $poco = str_replace("/profile/", "/poco/", $url); + + $noscrapedata = json_decode($noscrapejson, true); + + if (isset($noscrapedata["addr"])) + $addr = $noscrapedata["addr"]; + + if (isset($noscrapedata["fn"])) + $vcard["fn"] = $noscrapedata["fn"]; + + if (isset($noscrapedata["key"])) + $pubkey = $noscrapedata["key"]; + + if (isset($noscrapedata["photo"])) + $vcard["photo"] = $noscrapedata["photo"]; + + if (isset($noscrapedata["dfrn-request"])) + $request = $noscrapedata["dfrn-request"]; + + if (isset($noscrapedata["dfrn-confirm"])) + $confirm = $noscrapedata["dfrn-confirm"]; + + if (isset($noscrapedata["dfrn-notify"])) + $notify = $noscrapedata["dfrn-notify"]; + + if (isset($noscrapedata["dfrn-poll"])) + $poll = $noscrapedata["dfrn-poll"]; + } } - if((! $vcard['photo']) && strlen($email)) - $vcard['photo'] = avatar_img($email); - if($poll === $profile) - $lnk = $feed->get_permalink(); - if(isset($lnk) && strlen($lnk)) - $profile = $lnk; - - if(! $network) { + if(! $network) $network = NETWORK_FEED; - // If it is a feed, don't take the author name as feed name - unset($vcard['fn']); - } - if(! (x($vcard,'fn'))) - $vcard['fn'] = notags($feed->get_title()); - if(! (x($vcard,'fn'))) - $vcard['fn'] = notags($feed->get_description()); - - if(strpos($vcard['fn'],'Twitter / ') !== false) { - $vcard['fn'] = substr($vcard['fn'],strpos($vcard['fn'],'/')+1); - $vcard['fn'] = trim($vcard['fn']); - } + if(! x($vcard,'nick')) { $vcard['nick'] = strtolower(notags(unxmlify($vcard['fn']))); if(strpos($vcard['nick'],' ')) @@ -756,7 +731,7 @@ function probe_url($url, $mode = PROBE_NORMAL) { if(! x($vcard,'photo')) { $a = get_app(); - $vcard['photo'] = $a->get_baseurl() . '/images/person-175.jpg' ; + $vcard['photo'] = App::get_baseurl() . '/images/person-175.jpg' ; } if(! $profile) @@ -768,21 +743,24 @@ function probe_url($url, $mode = PROBE_NORMAL) { $vcard['fn'] = $url; if (($notify != "") AND ($poll != "")) { - $baseurl = matching($notify, $poll); + $baseurl = matching(normalise_link($notify), normalise_link($poll)); - $baseurl2 = matching($baseurl, $profile); + $baseurl2 = matching($baseurl, normalise_link($profile)); if ($baseurl2 != "") $baseurl = $baseurl2; } if (($baseurl == "") AND ($notify != "")) - $baseurl = matching($profile, $notify); + $baseurl = matching(normalise_link($profile), normalise_link($notify)); if (($baseurl == "") AND ($poll != "")) - $baseurl = matching($profile, $poll); + $baseurl = matching(normalise_link($profile), normalise_link($poll)); $baseurl = rtrim($baseurl, "/"); + if(strpos($url,'@') AND ($addr == "") AND ($network == NETWORK_DFRN)) + $addr = str_replace('acct:', '', $url); + $vcard['fn'] = notags($vcard['fn']); $vcard['nick'] = str_replace(' ','',notags($vcard['nick'])); @@ -805,16 +783,28 @@ function probe_url($url, $mode = PROBE_NORMAL) { logger('probe_url: ' . print_r($result,true), LOGGER_DEBUG); - // Trying if it maybe a diaspora account - if (($result['network'] == NETWORK_FEED) OR ($result['addr'] == "")) { - require_once('include/bbcode.php'); - $address = GetProfileUsername($url, "", true); - $result2 = probe_url($address, $mode); - if ($result2['network'] != "") - $result = $result2; + if ($level == 1) { + // Trying if it maybe a diaspora account + if (($result['network'] == NETWORK_FEED) OR ($result['addr'] == "")) { + require_once('include/bbcode.php'); + $address = GetProfileUsername($url, "", true); + $result2 = probe_url($address, $mode, ++$level); + if ($result2['network'] != "") + $result = $result2; + } + + // Maybe it's some non standard GNU Social installation (Single user, subfolder or no uri rewrite) + if (($result['network'] == NETWORK_FEED) AND ($result['baseurl'] != "") AND ($result['nick'] != "")) { + $addr = $result['nick'].'@'.str_replace("http://", "", $result['baseurl']); + $result2 = probe_url($addr, $mode, ++$level); + if (($result2['network'] != "") AND ($result2['network'] != NETWORK_FEED)) + $result = $result2; + } } - Cache::set("probe_url:".$mode.":".$url,serialize($result)); + // Only store into the cache if the value seems to be valid + if ($result['network'] != NETWORK_PHANTOM) + Cache::set("probe_url:".$mode.":".$url,serialize($result), CACHE_DAY); return $result; }