src/Wallabag/CoreBundle/Helper/ContentProxy.php

   1 <?php
   2
   3 namespace Wallabag\CoreBundle\Helper;
   4
   5 use Graby\Graby;
   6 use Psr\Log\LoggerInterface;
   7 use Symfony\Component\HttpFoundation\File\MimeType\MimeTypeExtensionGuesser;
   8 use Symfony\Component\Validator\Constraints\Locale as LocaleConstraint;
   9 use Symfony\Component\Validator\Constraints\Url as UrlConstraint;
  10 use Symfony\Component\Validator\Validator\ValidatorInterface;
  11 use Wallabag\CoreBundle\Entity\Entry;
  12 use Wallabag\CoreBundle\Tools\Utils;
  13
  14 /**
  15  * This kind of proxy class take care of getting the content from an url
  16  * and update the entry with what it found.
  17  */
  18 class ContentProxy
  19 {
  20     protected $graby;
  21     protected $tagger;
  22     protected $validator;
  23     protected $logger;
  24     protected $mimeGuesser;
  25     protected $fetchingErrorMessage;
  26     protected $eventDispatcher;
  27     protected $storeArticleHeaders;
  28
  29     public function __construct(Graby $graby, RuleBasedTagger $tagger, ValidatorInterface $validator, LoggerInterface $logger, $fetchingErrorMessage, $storeArticleHeaders = false)
  30     {
  31         $this->graby = $graby;
  32         $this->tagger = $tagger;
  33         $this->validator = $validator;
  34         $this->logger = $logger;
  35         $this->mimeGuesser = new MimeTypeExtensionGuesser();
  36         $this->fetchingErrorMessage = $fetchingErrorMessage;
  37         $this->storeArticleHeaders = $storeArticleHeaders;
  38     }
  39
  40     /**
  41      * Update entry using either fetched or provided content.
  42      *
  43      * @param Entry  $entry                Entry to update
  44      * @param string $url                  Url of the content
  45      * @param array  $content              Array with content provided for import with AT LEAST keys title, html, url to skip the fetchContent from the url
  46      * @param bool   $disableContentUpdate Whether to skip trying to fetch content using Graby
  47      */
  48     public function updateEntry(Entry $entry, $url, array $content = [], $disableContentUpdate = false)
  49     {
  50         if (!empty($content['html'])) {
  51             $content['html'] = $this->graby->cleanupHtml($content['html'], $url);
  52         }
  53
  54         if ((empty($content) || false === $this->validateContent($content)) && false === $disableContentUpdate) {
  55             $fetchedContent = $this->graby->fetchContent($url);
  56             $fetchedContent['title'] = $this->sanitizeContentTitle($fetchedContent['title'], $fetchedContent['content_type']);
  57
  58             // when content is imported, we have information in $content
  59             // in case fetching content goes bad, we'll keep the imported information instead of overriding them
  60             if (empty($content) || $fetchedContent['html'] !== $this->fetchingErrorMessage) {
  61                 $content = $fetchedContent;
  62             }
  63         }
  64
  65         // be sure to keep the url in case of error
  66         // so we'll be able to refetch it in the future
  67         $content['url'] = !empty($content['url']) ? $content['url'] : $url;
  68
  69         $this->stockEntry($entry, $content);
  70     }
  71
  72     /**
  73      * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
  74      * @param $title
  75      * @param $contentType
  76      * @return string
  77      */
  78     private function sanitizeContentTitle($title, $contentType) {
  79         if ('application/pdf' === $contentType) {
  80             $convertedTitle = $this->convertPdfEncodingToUTF8($title);
  81             return $this->sanitizeUTF8Text($convertedTitle);
  82         }
  83         return $this->sanitizeUTF8Text($title);
  84     }
  85
  86     /**
  87      * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
  88      * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
  89      * @param $title
  90      * @return string (maybe contains invalid UTF-8 character)
  91      */
  92     private function convertPdfEncodingToUTF8($title) {
  93         // first try UTF-8 because its easier to detect its present/absence
  94         foreach (array('UTF-8', 'UTF-16BE', 'WINDOWS-1252') as $encoding) {
  95             if (mb_check_encoding($title, $encoding)) {
  96                 return mb_convert_encoding($title, 'UTF-8', $encoding);
  97             }
  98         }
  99         return $title;
 100     }
 101
 102     /**
 103      * Remove invalid UTF-8 characters from the given string.
 104      * @param String $rawText
 105      * @return string
 106      */
 107     private function sanitizeUTF8Text($rawText) {
 108         if (mb_check_encoding($rawText, 'UTF-8')) {
 109             return $rawText;
 110         }
 111         return iconv("UTF-8", "UTF-8//IGNORE", $rawText);
 112     }
 113
 114     /**
 115      * Use a Symfony validator to ensure the language is well formatted.
 116      *
 117      * @param Entry  $entry
 118      * @param string $value Language to validate and save
 119      */
 120     public function updateLanguage(Entry $entry, $value)
 121     {
 122         // some lang are defined as fr-FR, es-ES.
 123         // replacing - by _ might increase language support
 124         $value = str_replace('-', '_', $value);
 125
 126         $errors = $this->validator->validate(
 127             $value,
 128             (new LocaleConstraint())
 129         );
 130
 131         if (0 === \count($errors)) {
 132             $entry->setLanguage($value);
 133
 134             return;
 135         }
 136
 137         $this->logger->warning('Language validation failed. ' . (string) $errors);
 138     }
 139
 140     /**
 141      * Use a Symfony validator to ensure the preview picture is a real url.
 142      *
 143      * @param Entry  $entry
 144      * @param string $value URL to validate and save
 145      */
 146     public function updatePreviewPicture(Entry $entry, $value)
 147     {
 148         $errors = $this->validator->validate(
 149             $value,
 150             (new UrlConstraint())
 151         );
 152
 153         if (0 === \count($errors)) {
 154             $entry->setPreviewPicture($value);
 155
 156             return;
 157         }
 158
 159         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 160     }
 161
 162     /**
 163      * Update date.
 164      *
 165      * @param Entry  $entry
 166      * @param string $value Date to validate and save
 167      */
 168     public function updatePublishedAt(Entry $entry, $value)
 169     {
 170         $date = $value;
 171
 172         // is it a timestamp?
 173         if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
 174             $date = '@' . $date;
 175         }
 176
 177         try {
 178             // is it already a DateTime?
 179             // (it's inside the try/catch in case of fail to be parse time string)
 180             if (!$date instanceof \DateTime) {
 181                 $date = new \DateTime($date);
 182             }
 183
 184             $entry->setPublishedAt($date);
 185         } catch (\Exception $e) {
 186             $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
 187         }
 188     }
 189
 190     /**
 191      * Helper to extract and save host from entry url.
 192      *
 193      * @param Entry $entry
 194      */
 195     public function setEntryDomainName(Entry $entry)
 196     {
 197         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 198         if (false !== $domainName) {
 199             $entry->setDomainName($domainName);
 200         }
 201     }
 202
 203     /**
 204      * Helper to set a default title using:
 205      * - url basename, if applicable
 206      * - hostname.
 207      *
 208      * @param Entry $entry
 209      */
 210     public function setDefaultEntryTitle(Entry $entry)
 211     {
 212         $url = parse_url($entry->getUrl());
 213         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 214
 215         if (empty($path)) {
 216             $path = $url['host'];
 217         }
 218
 219         $entry->setTitle($path);
 220     }
 221
 222     /**
 223      * Stock entry with fetched or imported content.
 224      * Will fall back to OpenGraph data if available.
 225      *
 226      * @param Entry $entry   Entry to stock
 227      * @param array $content Array with at least title, url & html
 228      */
 229     private function stockEntry(Entry $entry, array $content)
 230     {
 231         $entry->setUrl($content['url']);
 232
 233         $this->setEntryDomainName($entry);
 234
 235         if (!empty($content['title'])) {
 236             $entry->setTitle($content['title']);
 237         } elseif (!empty($content['open_graph']['og_title'])) {
 238             $entry->setTitle($content['open_graph']['og_title']);
 239         }
 240
 241         $html = $content['html'];
 242         if (false === $html) {
 243             $html = $this->fetchingErrorMessage;
 244
 245             if (!empty($content['open_graph']['og_description'])) {
 246                 $html .= '<p><i>But we found a short description: </i></p>';
 247                 $html .= $content['open_graph']['og_description'];
 248             }
 249         }
 250
 251         $entry->setContent($html);
 252         $entry->setReadingTime(Utils::getReadingTime($html));
 253
 254         if (!empty($content['status'])) {
 255             $entry->setHttpStatus($content['status']);
 256         }
 257
 258         if (!empty($content['authors']) && \is_array($content['authors'])) {
 259             $entry->setPublishedBy($content['authors']);
 260         }
 261
 262         if (!empty($content['all_headers']) && $this->storeArticleHeaders) {
 263             $entry->setHeaders($content['all_headers']);
 264         }
 265
 266         if (!empty($content['date'])) {
 267             $this->updatePublishedAt($entry, $content['date']);
 268         }
 269
 270         if (!empty($content['language'])) {
 271             $this->updateLanguage($entry, $content['language']);
 272         }
 273
 274         if (!empty($content['open_graph']['og_image'])) {
 275             $this->updatePreviewPicture($entry, $content['open_graph']['og_image']);
 276         }
 277
 278         // if content is an image, define it as a preview too
 279         if (!empty($content['content_type']) && \in_array($this->mimeGuesser->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 280             $this->updatePreviewPicture($entry, $content['url']);
 281         }
 282
 283         if (!empty($content['content_type'])) {
 284             $entry->setMimetype($content['content_type']);
 285         }
 286
 287         try {
 288             $this->tagger->tag($entry);
 289         } catch (\Exception $e) {
 290             $this->logger->error('Error while trying to automatically tag an entry.', [
 291                 'entry_url' => $content['url'],
 292                 'error_msg' => $e->getMessage(),
 293             ]);
 294         }
 295     }
 296
 297     /**
 298      * Validate that the given content has at least a title, an html and a url.
 299      *
 300      * @param array $content
 301      *
 302      * @return bool true if valid otherwise false
 303      */
 304     private function validateContent(array $content)
 305     {
 306         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 307     }
 308 }