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             $entry->setGivenUrl($url);
  80         }
  81
  82         $this->stockEntry($entry, $content);
  83     }
  84
  85     /**
  86      * Use a Symfony validator to ensure the language is well formatted.
  87      *
  88      * @param Entry  $entry
  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 Entry  $entry
 115      * @param string $value URL to validate and save
 116      */
 117     public function updatePreviewPicture(Entry $entry, $value)
 118     {
 119         $errors = $this->validator->validate(
 120             $value,
 121             (new UrlConstraint())
 122         );
 123
 124         if (0 === \count($errors)) {
 125             $entry->setPreviewPicture($value);
 126
 127             return;
 128         }
 129
 130         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 131     }
 132
 133     /**
 134      * Update date.
 135      *
 136      * @param Entry  $entry
 137      * @param string $value Date to validate and save
 138      */
 139     public function updatePublishedAt(Entry $entry, $value)
 140     {
 141         $date = $value;
 142
 143         // is it a timestamp?
 144         if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
 145             $date = '@' . $date;
 146         }
 147
 148         try {
 149             // is it already a DateTime?
 150             // (it's inside the try/catch in case of fail to be parse time string)
 151             if (!$date instanceof \DateTime) {
 152                 $date = new \DateTime($date);
 153             }
 154
 155             $entry->setPublishedAt($date);
 156         } catch (\Exception $e) {
 157             $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
 158         }
 159     }
 160
 161     /**
 162      * Helper to extract and save host from entry url.
 163      *
 164      * @param Entry $entry
 165      */
 166     public function setEntryDomainName(Entry $entry)
 167     {
 168         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 169         if (false !== $domainName) {
 170             $entry->setDomainName($domainName);
 171         }
 172     }
 173
 174     /**
 175      * Helper to set a default title using:
 176      * - url basename, if applicable
 177      * - hostname.
 178      *
 179      * @param Entry $entry
 180      */
 181     public function setDefaultEntryTitle(Entry $entry)
 182     {
 183         $url = parse_url($entry->getUrl());
 184         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 185
 186         if (empty($path)) {
 187             $path = $url['host'];
 188         }
 189
 190         $entry->setTitle($path);
 191     }
 192
 193     /**
 194      * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
 195      *
 196      * @param string $title
 197      * @param string $contentType
 198      *
 199      * @return string
 200      */
 201     private function sanitizeContentTitle($title, $contentType)
 202     {
 203         if ('application/pdf' === $contentType) {
 204             $title = $this->convertPdfEncodingToUTF8($title);
 205         }
 206
 207         return $this->sanitizeUTF8Text($title);
 208     }
 209
 210     /**
 211      * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
 212      * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
 213      *
 214      * @param $title
 215      *
 216      * @return string (maybe contains invalid UTF-8 character)
 217      */
 218     private function convertPdfEncodingToUTF8($title)
 219     {
 220         // first try UTF-8 because its easier to detect its present/absence
 221         foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
 222             if (mb_check_encoding($title, $encoding)) {
 223                 return mb_convert_encoding($title, 'UTF-8', $encoding);
 224             }
 225         }
 226
 227         return $title;
 228     }
 229
 230     /**
 231      * Remove invalid UTF-8 characters from the given string.
 232      *
 233      * @param string $rawText
 234      *
 235      * @return string
 236      */
 237     private function sanitizeUTF8Text($rawText)
 238     {
 239         if (mb_check_encoding($rawText, 'UTF-8')) {
 240             return $rawText;
 241         }
 242
 243         return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
 244     }
 245
 246     /**
 247      * Stock entry with fetched or imported content.
 248      * Will fall back to OpenGraph data if available.
 249      *
 250      * @param Entry $entry   Entry to stock
 251      * @param array $content Array with at least title, url & html
 252      */
 253     private function stockEntry(Entry $entry, array $content)
 254     {
 255         $this->updateOriginUrl($entry, $content['url']);
 256
 257         $this->setEntryDomainName($entry);
 258
 259         if (!empty($content['title'])) {
 260             $entry->setTitle($content['title']);
 261         }
 262
 263         if (empty($content['html'])) {
 264             $content['html'] = $this->fetchingErrorMessage;
 265
 266             if (!empty($content['description'])) {
 267                 $content['html'] .= '<p><i>But we found a short description: </i></p>';
 268                 $content['html'] .= $content['description'];
 269             }
 270         }
 271
 272         $entry->setContent($content['html']);
 273         $entry->setReadingTime(Utils::getReadingTime($content['html']));
 274
 275         if (!empty($content['status'])) {
 276             $entry->setHttpStatus($content['status']);
 277         }
 278
 279         if (!empty($content['authors']) && \is_array($content['authors'])) {
 280             $entry->setPublishedBy($content['authors']);
 281         }
 282
 283         if (!empty($content['headers'])) {
 284             $entry->setHeaders($content['headers']);
 285         }
 286
 287         if (!empty($content['date'])) {
 288             $this->updatePublishedAt($entry, $content['date']);
 289         }
 290
 291         if (!empty($content['language'])) {
 292             $this->updateLanguage($entry, $content['language']);
 293         }
 294
 295         $previewPictureUrl = '';
 296         if (!empty($content['image'])) {
 297             $previewPictureUrl = $content['image'];
 298         }
 299
 300         // if content is an image, define it as a preview too
 301         if (!empty($content['headers']['content-type']) && \in_array($this->mimeGuesser->guess($content['headers']['content-type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 302             $previewPictureUrl = $content['url'];
 303         } elseif (empty($previewPictureUrl)) {
 304             $this->logger->debug('Extracting images from content to provide a default preview picture');
 305             $imagesUrls = DownloadImages::extractImagesUrlsFromHtml($content['html']);
 306             $this->logger->debug(\count($imagesUrls) . ' pictures found');
 307
 308             if (!empty($imagesUrls)) {
 309                 $previewPictureUrl = $imagesUrls[0];
 310             }
 311         }
 312
 313         if (!empty($content['headers']['content-type'])) {
 314             $entry->setMimetype($content['headers']['content-type']);
 315         }
 316
 317         if (!empty($previewPictureUrl)) {
 318             $this->updatePreviewPicture($entry, $previewPictureUrl);
 319         }
 320
 321         try {
 322             $this->tagger->tag($entry);
 323         } catch (\Exception $e) {
 324             $this->logger->error('Error while trying to automatically tag an entry.', [
 325                 'entry_url' => $content['url'],
 326                 'error_msg' => $e->getMessage(),
 327             ]);
 328         }
 329     }
 330
 331     /**
 332      * Update the origin_url field when a redirection occurs
 333      * This field is set if it is empty and new url does not match ignore list.
 334      *
 335      * @param Entry  $entry
 336      * @param string $url
 337      */
 338     private function updateOriginUrl(Entry $entry, $url)
 339     {
 340         if (empty($url) || $entry->getUrl() === $url) {
 341             return false;
 342         }
 343
 344         $parsed_entry_url = parse_url($entry->getUrl());
 345         $parsed_content_url = parse_url($url);
 346
 347         /**
 348          * The following part computes the list of part changes between two
 349          * parse_url arrays.
 350          *
 351          * As array_diff_assoc only computes changes to go from the left array
 352          * to the right one, we make two differents arrays to have both
 353          * directions. We merge these two arrays and sort keys before passing
 354          * the result to the switch.
 355          *
 356          * The resulting array gives us all changing parts between the two
 357          * urls: scheme, host, path, query and/or fragment.
 358          */
 359         $diff_ec = array_diff_assoc($parsed_entry_url, $parsed_content_url);
 360         $diff_ce = array_diff_assoc($parsed_content_url, $parsed_entry_url);
 361
 362         $diff = array_merge($diff_ec, $diff_ce);
 363         $diff_keys = array_keys($diff);
 364         sort($diff_keys);
 365
 366         if ($this->ignoreUrl($entry->getUrl())) {
 367             $entry->setUrl($url);
 368
 369             return false;
 370         }
 371
 372         /**
 373          * This switch case lets us apply different behaviors according to
 374          * changing parts of urls.
 375          *
 376          * As $diff_keys is an array, we provide arrays as cases. ['path'] means
 377          * 'only the path is different between the two urls' whereas
 378          * ['fragment', 'query'] means 'only fragment and query string parts are
 379          * different between the two urls'.
 380          *
 381          * Note that values in $diff_keys are sorted.
 382          */
 383         switch ($diff_keys) {
 384             case ['path']:
 385                 if (($parsed_entry_url['path'] . '/' === $parsed_content_url['path']) // diff is trailing slash, we only replace the url of the entry
 386                     || ($url === urldecode($entry->getUrl()))) { // we update entry url if new url is a decoded version of it, see EntryRepository#findByUrlAndUserId
 387                     $entry->setUrl($url);
 388                 }
 389                 break;
 390             case ['scheme']:
 391                 $entry->setUrl($url);
 392                 break;
 393             case ['fragment']:
 394                 // noop
 395                 break;
 396             default:
 397                 if (empty($entry->getOriginUrl())) {
 398                     $entry->setOriginUrl($entry->getUrl());
 399                 }
 400                 $entry->setUrl($url);
 401                 break;
 402         }
 403     }
 404
 405     /**
 406      * Check entry url against an ignore list to replace with content url.
 407      *
 408      * XXX: move the ignore list in the database to let users handle it
 409      *
 410      * @param string $url url to test
 411      *
 412      * @return bool true if url matches ignore list otherwise false
 413      */
 414     private function ignoreUrl($url)
 415     {
 416         $ignored_hosts = ['feedproxy.google.com', 'feeds.reuters.com'];
 417         $ignored_patterns = ['https?://www\.lemonde\.fr/tiny.*'];
 418
 419         $parsed_url = parse_url($url);
 420
 421         $filtered = array_filter($ignored_hosts, function ($var) use ($parsed_url) {
 422             return $var === $parsed_url['host'];
 423         });
 424
 425         if ([] !== $filtered) {
 426             return true;
 427         }
 428
 429         $filtered = array_filter($ignored_patterns, function ($var) use ($url) {
 430             return preg_match("`$var`i", $url);
 431         });
 432
 433         if ([] !== $filtered) {
 434             return true;
 435         }
 436
 437         return false;
 438     }
 439
 440     /**
 441      * Validate that the given content has at least a title, an html and a url.
 442      *
 443      * @param array $content
 444      *
 445      * @return bool true if valid otherwise false
 446      */
 447     private function validateContent(array $content)
 448     {
 449         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 450     }
 451 }