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 takes care of getting the content from an url
  16  * and updates 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         $this->graby->toggleImgNoReferrer(true);
  51         if (!empty($content['html'])) {
  52             $content['html'] = $this->graby->cleanupHtml($content['html'], $url);
  53         }
  54
  55         if ((empty($content) || false === $this->validateContent($content)) && false === $disableContentUpdate) {
  56             $fetchedContent = $this->graby->fetchContent($url);
  57
  58             $fetchedContent['title'] = $this->sanitizeContentTitle(
  59                 $fetchedContent['title'],
  60                 isset($fetchedContent['headers']['content-type']) ? $fetchedContent['headers']['content-type'] : ''
  61             );
  62
  63             // when content is imported, we have information in $content
  64             // in case fetching content goes bad, we'll keep the imported information instead of overriding them
  65             if (empty($content) || $fetchedContent['html'] !== $this->fetchingErrorMessage) {
  66                 $content = $fetchedContent;
  67             }
  68         }
  69
  70         // be sure to keep the url in case of error
  71         // so we'll be able to refetch it in the future
  72         $content['url'] = !empty($content['url']) ? $content['url'] : $url;
  73
  74         // In one case (at least in tests), url is empty here
  75         // so we set it using $url provided in the updateEntry call.
  76         // Not sure what are the other possible cases where this property is empty
  77         if (empty($entry->getUrl()) && !empty($url)) {
  78             $entry->setUrl($url);
  79         }
  80
  81         $this->stockEntry($entry, $content);
  82     }
  83
  84     /**
  85      * Use a Symfony validator to ensure the language is well formatted.
  86      *
  87      * @param Entry  $entry
  88      * @param string $value Language to validate and save
  89      */
  90     public function updateLanguage(Entry $entry, $value)
  91     {
  92         // some lang are defined as fr-FR, es-ES.
  93         // replacing - by _ might increase language support
  94         $value = str_replace('-', '_', $value);
  95
  96         $errors = $this->validator->validate(
  97             $value,
  98             (new LocaleConstraint())
  99         );
 100
 101         if (0 === \count($errors)) {
 102             $entry->setLanguage($value);
 103
 104             return;
 105         }
 106
 107         $this->logger->warning('Language validation failed. ' . (string) $errors);
 108     }
 109
 110     /**
 111      * Use a Symfony validator to ensure the preview picture is a real url.
 112      *
 113      * @param Entry  $entry
 114      * @param string $value URL to validate and save
 115      */
 116     public function updatePreviewPicture(Entry $entry, $value)
 117     {
 118         $errors = $this->validator->validate(
 119             $value,
 120             (new UrlConstraint())
 121         );
 122
 123         if (0 === \count($errors)) {
 124             $entry->setPreviewPicture($value);
 125
 126             return;
 127         }
 128
 129         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 130     }
 131
 132     /**
 133      * Update date.
 134      *
 135      * @param Entry  $entry
 136      * @param string $value Date to validate and save
 137      */
 138     public function updatePublishedAt(Entry $entry, $value)
 139     {
 140         $date = $value;
 141
 142         // is it a timestamp?
 143         if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
 144             $date = '@' . $date;
 145         }
 146
 147         try {
 148             // is it already a DateTime?
 149             // (it's inside the try/catch in case of fail to be parse time string)
 150             if (!$date instanceof \DateTime) {
 151                 $date = new \DateTime($date);
 152             }
 153
 154             $entry->setPublishedAt($date);
 155         } catch (\Exception $e) {
 156             $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
 157         }
 158     }
 159
 160     /**
 161      * Helper to extract and save host from entry url.
 162      *
 163      * @param Entry $entry
 164      */
 165     public function setEntryDomainName(Entry $entry)
 166     {
 167         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 168         if (false !== $domainName) {
 169             $entry->setDomainName($domainName);
 170         }
 171     }
 172
 173     /**
 174      * Helper to set a default title using:
 175      * - url basename, if applicable
 176      * - hostname.
 177      *
 178      * @param Entry $entry
 179      */
 180     public function setDefaultEntryTitle(Entry $entry)
 181     {
 182         $url = parse_url($entry->getUrl());
 183         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 184
 185         if (empty($path)) {
 186             $path = $url['host'];
 187         }
 188
 189         $entry->setTitle($path);
 190     }
 191
 192     /**
 193      * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
 194      *
 195      * @param string $title
 196      * @param string $contentType
 197      *
 198      * @return string
 199      */
 200     private function sanitizeContentTitle($title, $contentType)
 201     {
 202         if ('application/pdf' === $contentType) {
 203             $title = $this->convertPdfEncodingToUTF8($title);
 204         }
 205
 206         return $this->sanitizeUTF8Text($title);
 207     }
 208
 209     /**
 210      * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
 211      * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
 212      *
 213      * @param $title
 214      *
 215      * @return string (maybe contains invalid UTF-8 character)
 216      */
 217     private function convertPdfEncodingToUTF8($title)
 218     {
 219         // first try UTF-8 because its easier to detect its present/absence
 220         foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
 221             if (mb_check_encoding($title, $encoding)) {
 222                 return mb_convert_encoding($title, 'UTF-8', $encoding);
 223             }
 224         }
 225
 226         return $title;
 227     }
 228
 229     /**
 230      * Remove invalid UTF-8 characters from the given string.
 231      *
 232      * @param string $rawText
 233      *
 234      * @return string
 235      */
 236     private function sanitizeUTF8Text($rawText)
 237     {
 238         if (mb_check_encoding($rawText, 'UTF-8')) {
 239             return $rawText;
 240         }
 241
 242         return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
 243     }
 244
 245     /**
 246      * Stock entry with fetched or imported content.
 247      * Will fall back to OpenGraph data if available.
 248      *
 249      * @param Entry $entry   Entry to stock
 250      * @param array $content Array with at least title, url & html
 251      */
 252     private function stockEntry(Entry $entry, array $content)
 253     {
 254         $this->updateOriginUrl($entry, $content['url']);
 255
 256         $this->setEntryDomainName($entry);
 257
 258         if (!empty($content['title'])) {
 259             $entry->setTitle($content['title']);
 260         }
 261
 262         if (empty($content['html'])) {
 263             $content['html'] = $this->fetchingErrorMessage;
 264
 265             if (!empty($content['description'])) {
 266                 $content['html'] .= '<p><i>But we found a short description: </i></p>';
 267                 $content['html'] .= $content['description'];
 268             }
 269         }
 270
 271         $entry->setContent($content['html']);
 272         $entry->setReadingTime(Utils::getReadingTime($content['html']));
 273
 274         if (!empty($content['status'])) {
 275             $entry->setHttpStatus($content['status']);
 276         }
 277
 278         if (!empty($content['authors']) && \is_array($content['authors'])) {
 279             $entry->setPublishedBy($content['authors']);
 280         }
 281
 282         if (!empty($content['headers'])) {
 283             $entry->setHeaders($content['headers']);
 284         }
 285
 286         if (!empty($content['date'])) {
 287             $this->updatePublishedAt($entry, $content['date']);
 288         }
 289
 290         if (!empty($content['language'])) {
 291             $this->updateLanguage($entry, $content['language']);
 292         }
 293
 294         $previewPictureUrl = '';
 295         if (!empty($content['image'])) {
 296             $previewPictureUrl = $content['image'];
 297         }
 298
 299         // if content is an image, define it as a preview too
 300         if (!empty($content['headers']['content-type']) && \in_array($this->mimeGuesser->guess($content['headers']['content-type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 301             $previewPictureUrl = $content['url'];
 302
 303             $entry->setMimetype($content['headers']['content-type']);
 304         } elseif (empty($previewPictureUrl)) {
 305             $this->logger->debug('Extracting images from content to provide a default preview picture');
 306             $imagesUrls = DownloadImages::extractImagesUrlsFromHtml($content['html']);
 307             $this->logger->debug(\count($imagesUrls) . ' pictures found');
 308
 309             if (!empty($imagesUrls)) {
 310                 $previewPictureUrl = $imagesUrls[0];
 311             }
 312         }
 313
 314         if (!empty($previewPictureUrl)) {
 315             $this->updatePreviewPicture($entry, $previewPictureUrl);
 316         }
 317
 318         try {
 319             $this->tagger->tag($entry);
 320         } catch (\Exception $e) {
 321             $this->logger->error('Error while trying to automatically tag an entry.', [
 322                 'entry_url' => $content['url'],
 323                 'error_msg' => $e->getMessage(),
 324             ]);
 325         }
 326     }
 327
 328     /**
 329      * Update the origin_url field when a redirection occurs
 330      * This field is set if it is empty and new url does not match ignore list.
 331      *
 332      * @param Entry  $entry
 333      * @param string $url
 334      */
 335     private function updateOriginUrl(Entry $entry, $url)
 336     {
 337         if (empty($url) || $entry->getUrl() === $url) {
 338             return false;
 339         }
 340
 341         $parsed_entry_url = parse_url($entry->getUrl());
 342         $parsed_content_url = parse_url($url);
 343
 344         /**
 345          * The following part computes the list of part changes between two
 346          * parse_url arrays.
 347          *
 348          * As array_diff_assoc only computes changes to go from the left array
 349          * to the right one, we make two differents arrays to have both
 350          * directions. We merge these two arrays and sort keys before passing
 351          * the result to the switch.
 352          *
 353          * The resulting array gives us all changing parts between the two
 354          * urls: scheme, host, path, query and/or fragment.
 355          */
 356         $diff_ec = array_diff_assoc($parsed_entry_url, $parsed_content_url);
 357         $diff_ce = array_diff_assoc($parsed_content_url, $parsed_entry_url);
 358
 359         $diff = array_merge($diff_ec, $diff_ce);
 360         $diff_keys = array_keys($diff);
 361         sort($diff_keys);
 362
 363         if ($this->ignoreUrl($entry->getUrl())) {
 364             $entry->setUrl($url);
 365
 366             return false;
 367         }
 368
 369         /**
 370          * This switch case lets us apply different behaviors according to
 371          * changing parts of urls.
 372          *
 373          * As $diff_keys is an array, we provide arrays as cases. ['path'] means
 374          * 'only the path is different between the two urls' whereas
 375          * ['fragment', 'query'] means 'only fragment and query string parts are
 376          * different between the two urls'.
 377          *
 378          * Note that values in $diff_keys are sorted.
 379          */
 380         switch ($diff_keys) {
 381             case ['path']:
 382                 if (($parsed_entry_url['path'] . '/' === $parsed_content_url['path']) // diff is trailing slash, we only replace the url of the entry
 383                     || ($url === urldecode($entry->getUrl()))) { // we update entry url if new url is a decoded version of it, see EntryRepository#findByUrlAndUserId
 384                     $entry->setUrl($url);
 385                 }
 386                 break;
 387             case ['scheme']:
 388                 $entry->setUrl($url);
 389                 break;
 390             case ['fragment']:
 391                 // noop
 392                 break;
 393             default:
 394                 if (empty($entry->getOriginUrl())) {
 395                     $entry->setOriginUrl($entry->getUrl());
 396                 }
 397                 $entry->setUrl($url);
 398                 break;
 399         }
 400     }
 401
 402     /**
 403      * Check entry url against an ignore list to replace with content url.
 404      *
 405      * XXX: move the ignore list in the database to let users handle it
 406      *
 407      * @param string $url url to test
 408      *
 409      * @return bool true if url matches ignore list otherwise false
 410      */
 411     private function ignoreUrl($url)
 412     {
 413         $ignored_hosts = ['feedproxy.google.com', 'feeds.reuters.com'];
 414         $ignored_patterns = ['https?://www\.lemonde\.fr/tiny.*'];
 415
 416         $parsed_url = parse_url($url);
 417
 418         $filtered = array_filter($ignored_hosts, function ($var) use ($parsed_url) {
 419             return $var === $parsed_url['host'];
 420         });
 421
 422         if ([] !== $filtered) {
 423             return true;
 424         }
 425
 426         $filtered = array_filter($ignored_patterns, function ($var) use ($url) {
 427             return preg_match("`$var`i", $url);
 428         });
 429
 430         if ([] !== $filtered) {
 431             return true;
 432         }
 433
 434         return false;
 435     }
 436
 437     /**
 438      * Validate that the given content has at least a title, an html and a url.
 439      *
 440      * @param array $content
 441      *
 442      * @return bool true if valid otherwise false
 443      */
 444     private function validateContent(array $content)
 445     {
 446         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 447     }
 448 }