Tests

[friendica.git] / src / Util / ParseUrl.php

diff --git a/src/Util/ParseUrl.php b/src/Util/ParseUrl.php
index d2986a362ac9cc87477095f55731aca7cd5d2124..d5c9f02ad4a4eaf22f1879c4323a698054fc3e63 100644 (file)
--- a/src/Util/ParseUrl.php
+++ b/src/Util/ParseUrl.php
@@ -1,6 +1,6 @@
 <?php
 /**
- * @copyright Copyright (C) 2020, Friendica
+ * @copyright Copyright (C) 2010-2021, the Friendica project
  *
  * @license GNU AGPL version 3 or any later version
  *
@@ -54,7 +54,7 @@ class ParseUrl
        /**
         * Fetch the content type of the given url
         * @param string $url URL of the page
-        * @return array content type 
+        * @return array content type
         */
        public static function getContentType(string $url)
        {
@@ -197,7 +197,7 @@ class ParseUrl
                ];
 
                if ($count > 10) {
-                       Logger::log('Endless loop detected for ' . $url, Logger::DEBUG);
+                       Logger::notice('Endless loop detected', ['url' => $url]);
                        return $siteinfo;
                }
 
@@ -214,7 +214,7 @@ class ParseUrl
                }
 
                $curlResult = DI::httpRequest()->get($url);
-               if (!$curlResult->isSuccess()) {
+               if (!$curlResult->isSuccess() || empty($curlResult->getBody())) {
                        return $siteinfo;
                }
 
@@ -297,7 +297,7 @@ class ParseUrl
                        // See https://github.com/friendica/friendica/issues/5470#issuecomment-418351211
                        $charset = str_ireplace('latin-1', 'latin1', $charset);
 
-                       Logger::log('detected charset ' . $charset, Logger::DEBUG);
+                       Logger::info('detected charset', ['charset' => $charset]);
                        $body = iconv($charset, 'UTF-8//TRANSLIT', $body);
                }
 
@@ -349,6 +349,9 @@ class ParseUrl
                        $siteinfo['title'] = trim($list->item(0)->nodeValue);
                }
 
+               $twitter_card = false;
+               $twitter_image = false;
+
                $list = $xpath->query('//meta[@name]');
                foreach ($list as $node) {
                        $meta_tag = [];
@@ -376,6 +379,7 @@ class ParseUrl
                                        break;
                                case 'twitter:image':
                                        $siteinfo['image'] = $meta_tag['content'];
+                                       $twitter_image = true;
                                        break;
                                case 'twitter:image:src':
                                        $siteinfo['image'] = $meta_tag['content'];
@@ -383,7 +387,7 @@ class ParseUrl
                                case 'twitter:card':
                                        // Detect photo pages
                                        if ($meta_tag['content'] == 'summary_large_image') {
-                                               $siteinfo['type'] = 'photo';
+                                               $twitter_card = true;
                                        }
                                        break;
                                case 'twitter:description':
@@ -458,6 +462,7 @@ class ParseUrl
                                                break;
                                        case 'twitter:image':
                                                $siteinfo['image'] = $meta_tag['content'];
+                                               $twitter_image = true;
                                                break;
                                }
                        }
@@ -472,10 +477,11 @@ class ParseUrl
                        }
                }
 
+// Currently deactivated, see https://github.com/friendica/friendica/pull/10148#issuecomment-821512658
                // Prevent to have a photo type without an image
-               if ((empty($siteinfo['image']) || !empty($siteinfo['text'])) && ($siteinfo['type'] == 'photo')) {
-                       $siteinfo['type'] = 'link';
-               }
+//             if ($twitter_card && $twitter_image && !empty($siteinfo['image'])) {
+//                     $siteinfo['type'] = 'photo';
+//             }
 
                if (!empty($siteinfo['image'])) {
                        $siteinfo['images'] = $siteinfo['images'] ?? [];
@@ -514,15 +520,19 @@ class ParseUrl
                        array_walk($siteinfo['images'], function (&$image) use ($page_url) {
                                // According to the specifications someone could place a picture url into the content field as well.
                                // But this doesn't seem to happen in the wild, so we don't cover it here.
-                               $image['url'] = self::completeUrl($image['url'], $page_url);
-                               $photodata = Images::getInfoFromURLCached($image['url']);
-                               if (!empty($photodata) && ($photodata[0] > 50) && ($photodata[1] > 50)) {
-                                       $image['src'] = $image['url'];
-                                       $image['width'] = $photodata[0];
-                                       $image['height'] = $photodata[1];
-                                       $image['contenttype'] = $photodata['mime'];
-                                       unset($image['url']);
-                                       ksort($image);
+                               if (!empty($image['url'])) {
+                                       $image['url'] = self::completeUrl($image['url'], $page_url);
+                                       $photodata = Images::getInfoFromURLCached($image['url']);
+                                       if (!empty($photodata) && ($photodata[0] > 50) && ($photodata[1] > 50)) {
+                                               $image['src'] = $image['url'];
+                                               $image['width'] = $photodata[0];
+                                               $image['height'] = $photodata[1];
+                                               $image['contenttype'] = $photodata['mime'];
+                                               unset($image['url']);
+                                               ksort($image);
+                                       } else {
+                                               $image = [];
+                                       }
                                } else {
                                        $image = [];
                                }
@@ -542,7 +552,7 @@ class ParseUrl
                                                if (!empty($media[$field])) {
                                                        $media[$field] = self::completeUrl($media[$field], $page_url);
                                                        $type = self::getContentType($media[$field]);
-                                                       if ($type[0] == 'text') {
+                                                       if (($type[0] ?? '') == 'text') {
                                                                if ($field == 'embed') {
                                                                        $embed = $media[$field];
                                                                } else {
@@ -685,7 +695,9 @@ class ParseUrl
        {
                if (!empty($jsonld['@graph']) && is_array($jsonld['@graph'])) {
                        foreach ($jsonld['@graph'] as $part) {
-                               $siteinfo = self::parseParts($siteinfo, $part);
+                               if (!empty($part) && is_array($part)) {
+                                       $siteinfo = self::parseParts($siteinfo, $part);
+                               }
                        }
                } elseif (!empty($jsonld['@type'])) {
                        $siteinfo = self::parseJsonLd($siteinfo, $jsonld);
@@ -699,14 +711,16 @@ class ParseUrl
                        }
                        if ($numeric_keys) {
                                foreach ($jsonld as $part) {
-                                       $siteinfo = self::parseParts($siteinfo, $part);
+                                       if (!empty($part) && is_array($part)) {
+                                               $siteinfo = self::parseParts($siteinfo, $part);
+                                       }
                                }
                        }
                }
 
                array_walk_recursive($siteinfo, function (&$element) {
                        if (is_string($element)) {
-                               $element = html_entity_decode($element, ENT_COMPAT, 'UTF-8');
+                               $element = trim(strip_tags(html_entity_decode($element, ENT_COMPAT, 'UTF-8')));
                        }
                });
 
@@ -764,7 +778,7 @@ class ParseUrl
                        case 'QAPage':
                        case 'RealEstateListing':
                        case 'SearchResultsPage':
-                       case 'MediaGallery':                    
+                       case 'MediaGallery':
                        case 'ImageGallery':
                        case 'VideoGallery':
                        case 'RadioEpisode':
@@ -793,7 +807,7 @@ class ParseUrl
                        case 'PerformingGroup':
                        case 'DanceGroup';
                        case 'MusicGroup':
-                       case 'TheaterGroup':                    
+                       case 'TheaterGroup':
                                return self::parseJsonLdWebPerson($siteinfo, $jsonld);
                        case 'AudioObject':
                        case 'Audio':
@@ -937,7 +951,7 @@ class ParseUrl
                                        $siteinfo['keywords'][] = trim($keyword);
                                }
                        }
-               } else {
+               } elseif (!empty($jsonld['keywords'])) {
                        $content = JsonLD::fetchElementArray($jsonld, 'keywords');
                        if (!empty($content) && is_array($content)) {
                                $jsonldinfo['keywords'] = $content;