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