Improved scraping and poco discovery.

diff --git a/include/Scrape.php b/include/Scrape.php
index 90fb1a9e3dd99e992d391a6f09f1ff61618c9f9c..83c099769d456d851175ec9dcffba9f2e37a7c5f 100644 (file)
--- a/include/Scrape.php
+++ b/include/Scrape.php
@@ -249,7 +249,7 @@ function scrape_feed($url) {
                                        $ret['feed_atom'] = $url;
                                        return $ret;
                                }
-                               if(stristr($line,'application/rss+xml') || stristr($s,'<rss')) {
+                               if(stristr($line,'application/rss+xml') || stristr($s,'<rss')) {
                                        $ret['feed_rss'] = $url;
                                        return $ret;
                                }
@@ -262,47 +262,40 @@ function scrape_feed($url) {
                }
        }
 
-       try {
-               $dom = HTML5_Parser::parse($s);
-       } catch (DOMException $e) {
-               logger('scrape_feed: parse error: ' . $e);
-       }
+       $basename = implode('/', array_slice(explode('/',$url),0,3)) . '/';
 
-       if(! $dom) {
-               logger('scrape_feed: failed to parse.');
-               return $ret;
-       }
+       $doc = new DOMDocument();
+       @$doc->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 . $ret['feed_atom'];
@@ -509,6 +502,7 @@ function probe_url($url, $mode = PROBE_NORMAL, $level = 1) {
        }
 
        if($mode == PROBE_NORMAL) {
+
                if(strlen($zot)) {
                        $s = fetch_url($zot);
                        if($s) {
@@ -528,6 +522,7 @@ function probe_url($url, $mode = PROBE_NORMAL, $level = 1) {
                        }
                }
 
+
                if(strlen($dfrn)) {
                        $ret = scrape_dfrn(($hcard) ? $hcard : $dfrn);
                        if(is_array($ret) && x($ret,'dfrn-request')) {
@@ -634,6 +629,7 @@ function probe_url($url, $mode = PROBE_NORMAL, $level = 1) {
                if($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']));
@@ -653,7 +649,7 @@ function probe_url($url, $mode = PROBE_NORMAL, $level = 1) {
                        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 : '<?xml version="1.0" encoding="utf-8" ?><xml></xml>');
+                       $feed->set_raw_data(($xml) ? $xml : '<?xml version="1.0" encoding="utf-8" ?><xml></xml>');
 
                        $feed->init();
                        if($feed->error())

diff --git a/include/socgraph.php b/include/socgraph.php
index 45cf3a650a30ca6bc615878bb517b72c0bed68ce..9348cb67b56bba34c6351723f16707e7198824d8 100644 (file)
--- a/include/socgraph.php
+++ b/include/socgraph.php
@@ -4,6 +4,11 @@ require_once('include/datetime.php');
 require_once("include/Scrape.php");
 require_once("include/html2bbcode.php");
 
+/*
+ To-Do:
+ - Move GNU Social URL schemata (http://server.tld/user/number) to http://server.tld/username
+*/
+
 /*
  * poco_load
  *
@@ -167,6 +172,14 @@ function poco_check($profile_url, $name, $network, $profile_photo, $about, $loca
        if ($profile_url == "")
                return $gcid;
 
+       $urlparts = parse_url($profile_url);
+       if (!isset($urlparts["scheme"]))
+               return $gcid;
+
+       if (in_array($urlparts["host"], array("facebook.com", "twitter.com", "www.twitter-rss.com",
+                                               "identi.ca", "alpha.app.net")))
+               return $gcid;
+
        $orig_updated = $updated;
 
        // Don't store the statusnet connector as network
@@ -220,7 +233,8 @@ function poco_check($profile_url, $name, $network, $profile_photo, $about, $loca
                $nick = end(explode("/", $urlparts["path"]));
        }
 
-       if (($network == "") OR ($name == "") OR ($profile_photo == "") OR ($server_url == "")) {
+       if ((($network == "") OR ($name == "") OR ($profile_photo == "") OR ($server_url == ""))
+               AND poco_reachable($profile_url, $server_url, $network)) {
                $data = probe_url($profile_url);
                $network = $data["network"];
                $name = $data["name"];
@@ -248,7 +262,9 @@ function poco_check($profile_url, $name, $network, $profile_photo, $about, $loca
        poco_check_server($server_url, $network);
 
        // Only fetch last update manually if it wasn't provided and enabled in the system
-       if (get_config('system','poco_completion') AND ($orig_updated == "0000-00-00 00:00:00") AND poco_do_update($created, $updated, $last_failure, $last_contact)) {
+       if (get_config('system','poco_completion') AND ($orig_updated == "0000-00-00 00:00:00")
+               AND poco_do_update($created, $updated, $last_failure, $last_contact)
+               AND poco_reachable($profile_url, $server_url, $network)) {
                $last_updated = poco_last_updated($profile_url);
                if ($last_updated) {
                        $updated = $last_updated;
@@ -361,6 +377,49 @@ function poco_check($profile_url, $name, $network, $profile_photo, $about, $loca
        return $gcid;
 }
 
+function poco_reachable($profile, $server = "", $network = "") {
+
+       if ($server == "")
+               $server = poco_detect_server($profile);
+
+       if ($server == "")
+               return true;
+
+       return poco_check_server($server, $network);
+}
+
+function poco_detect_server($profile) {
+
+       // Try to detect the server path based upon some known standard paths
+       $server_url = "";
+
+       if ($server_url == "") {
+               $friendica = preg_replace("=(https?://)(.*)/profile/(.*)=ism", "$1$2", $profile);
+               if ($friendica != $profile) {
+                       $server_url = $friendica;
+                       $network = NETWORK_DFRN;
+               }
+       }
+
+       if ($server_url == "") {
+               $diaspora = preg_replace("=(https?://)(.*)/u/(.*)=ism", "$1$2", $profile);
+               if ($diaspora != $profile) {
+                       $server_url = $diaspora;
+                       $network = NETWORK_DIASPORA;
+               }
+       }
+
+       if ($server_url == "") {
+               $red = preg_replace("=(https?://)(.*)/channel/(.*)=ism", "$1$2", $profile);
+               if ($red != $profile) {
+                       $server_url = $red;
+                       $network = NETWORK_DIASPORA;
+               }
+       }
+
+       return $server_url;
+}
+
 function poco_last_updated($profile) {
 
        $gcontacts = q("SELECT * FROM `gcontact` WHERE `nurl` = '%s'",
@@ -437,7 +496,7 @@ function poco_last_updated($profile) {
 
        // If we only can poll the feed, then we only do this once a while
        if (!poco_do_update($gcontacts[0]["created"], $gcontacts[0]["updated"], $gcontacts[0]["last_failure"],  $gcontacts[0]["last_contact"]))
-/              return $gcontacts[0]["updated"];
+               return $gcontacts[0]["updated"];
 
        $data = probe_url($profile);
 
@@ -558,7 +617,7 @@ function poco_to_boolean($val) {
        return ($val);
 }
 
-function poco_check_server($server_url, $network = "") {
+function poco_check_server($server_url, $network = "", $force = false) {
 
        if ($server_url == "")
                return false;
@@ -584,8 +643,10 @@ function poco_check_server($server_url, $network = "") {
                $info = $servers[0]["info"];
                $register_policy = $servers[0]["register_policy"];
 
-               if (!poco_do_update($servers[0]["created"], "", $last_failure, $last_contact))
+               if (!$force AND !poco_do_update($servers[0]["created"], "", $last_failure, $last_contact)) {
+                       logger("Use cached data for server ".$server_url, LOGGER_DEBUG);
                        return ($last_contact >= $last_failure);
+               }
        } else {
                $poco = "";
                $noscrape = "";
@@ -598,6 +659,7 @@ function poco_check_server($server_url, $network = "") {
                $last_contact = "0000-00-00 00:00:00";
                $last_failure = "0000-00-00 00:00:00";
        }
+       logger("Server ".$server_url." is unknown. Start discovery.", LOGGER_DEBUG);
 
        $failure = false;
        $orig_last_failure = $last_failure;
@@ -752,6 +814,9 @@ function poco_check_server($server_url, $network = "") {
                }
        }
 
+       // Check again if the server exists
+       $servers = q("SELECT `nurl` FROM `gserver` WHERE `nurl` = '%s'", dbesc(normalise_link($server_url)));
+
        if ($servers)
                 q("UPDATE `gserver` SET `url` = '%s', `version` = '%s', `site_name` = '%s', `info` = '%s', `register_policy` = %d, `poco` = '%s', `noscrape` = '%s',
                        `network` = '%s', `platform` = '%s', `last_contact` = '%s', `last_failure` = '%s' WHERE `nurl` = '%s'",
@@ -770,7 +835,7 @@ function poco_check_server($server_url, $network = "") {
                );
        else
                q("INSERT INTO `gserver` (`url`, `nurl`, `version`, `site_name`, `info`, `register_policy`, `poco`, `noscrape`, `network`, `platform`, `created`, `last_contact`, `last_failure`)
-                                       VALUES ('%s', '%s', '%s', '%s', '%s', %d, '%s', '%s', '%s', '%s', '%s', '%s')",
+                                       VALUES ('%s', '%s', '%s', '%s', '%s', %d, '%s', '%s', '%s', '%s', '%s', '%s', '%s')",
                                dbesc($server_url),
                                dbesc(normalise_link($server_url)),
                                dbesc($version),
@@ -786,6 +851,9 @@ function poco_check_server($server_url, $network = "") {
                                dbesc($last_failure),
                                dbesc(datetime_convert())
                );
+
+       logger("End discovery for server ".$server_url, LOGGER_DEBUG);
+
        return !$failure;
 }
 
@@ -1096,11 +1164,17 @@ function update_suggestions() {
 
 function poco_discover($complete = false) {
 
-       $last_update = date("c", time() - (60 * 60 * 24));
+       $no_of_queries = 5;
+
+       $last_update = date("c", time() - (60 * 60 * 6)); // 24
 
-       $r = q("SELECT `poco`, `nurl` FROM `gserver` WHERE `last_contact` > `last_failure` AND `poco` != '' AND `last_poco_query` < '%s' ORDER BY RAND()", dbesc($last_update));
+       $r = q("SELECT `poco`, `nurl`, `url`, `network` FROM `gserver` WHERE `last_contact` > `last_failure` AND `poco` != '' AND `last_poco_query` < '%s' ORDER BY RAND()", dbesc($last_update));
        if ($r)
                foreach ($r AS $server) {
+
+                       if (!poco_check_server($server["url"], $server["network"]))
+                               continue;
+
                        // Fetch all users from the other server
                        $url = $server["poco"]."/?fields=displayName,urls,photos,updated,network,aboutMe,currentLocation,tags,gender,generation";
 
@@ -1109,6 +1183,7 @@ function poco_discover($complete = false) {
                        $retdata = z_fetch_url($url);
                        if ($retdata["success"]) {
                                $data = json_decode($retdata["body"]);
+
                                poco_discover_server($data, 2);
 
                                if (get_config('system','poco_discovery') > 1) {
@@ -1122,24 +1197,29 @@ function poco_discover($complete = false) {
                                        // Fetch all global contacts from the other server (Not working with Redmatrix and Friendica versions before 3.3)
                                        $url = $server["poco"]."/@global?updatedSince=".$updatedSince."&fields=displayName,urls,photos,updated,network,aboutMe,currentLocation,tags,gender,generation";
 
+                                       $success = false;
+
                                        $retdata = z_fetch_url($url);
                                        if ($retdata["success"]) {
                                                logger("Fetch all global contacts from the server ".$server["nurl"], LOGGER_DEBUG);
-                                               poco_discover_server(json_decode($retdata["body"]));
-                                       } elseif (get_config('system','poco_discovery') > 2) {
+                                               $success = poco_discover_server(json_decode($retdata["body"]));
+                                       }
+
+                                       if (!$success AND (get_config('system','poco_discovery') > 2)) {
                                                logger("Fetch contacts from users of the server ".$server["nurl"], LOGGER_DEBUG);
-                                               poco_discover_server_users($data);
+                                               poco_discover_server_users($data, $server);
                                        }
                                }
 
                                q("UPDATE `gserver` SET `last_poco_query` = '%s' WHERE `nurl` = '%s'", dbesc(datetime_convert()), dbesc($server["nurl"]));
-                               if (!$complete)
+                               if (!$complete AND (--$no_of_queries == 0))
                                        break;
-                       }
+                       } else  // If the server hadn't replied correctly, then force a sanity check
+                               poco_check_server($server["url"], $server["network"], true);
                }
 }
 
-function poco_discover_server_users($data) {
+function poco_discover_server_users($data, $server) {
        foreach ($data->entry AS $entry) {
                $username = "";
                if (isset($entry->urls)) {
@@ -1166,7 +1246,9 @@ function poco_discover_server_users($data) {
 function poco_discover_server($data, $default_generation = 0) {
 
        if (!isset($data->entry) OR !count($data->entry))
-               return;
+               return false;
+
+       $success = false;
 
        foreach ($data->entry AS $entry) {
                $profile_url = '';
@@ -1195,6 +1277,7 @@ function poco_discover_server($data, $default_generation = 0) {
                                }
                        }
                }
+
                if(isset($entry->photos)) {
                        foreach($entry->photos as $photo) {
                                if($photo->type == 'profile') {
@@ -1227,10 +1310,13 @@ function poco_discover_server($data, $default_generation = 0) {
                                $keywords = implode(", ", $tag);
 
                if ($generation > 0) {
+                       $success = true;
+
                        logger("Store profile ".$profile_url, LOGGER_DEBUG);
-                       poco_check($profile_url, $name, $network, $profile_photo, $about, $location, $gender, $keywords, $connect_url, $updated, $generation);
+                       poco_check($profile_url, $name, $network, $profile_photo, $about, $location, $gender, $keywords, $connect_url, $updated, $generation, 0, 0, 0);
                        logger("Done for profile ".$profile_url, LOGGER_DEBUG);
                }
        }
+       return $success;
 }
 ?>