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 $ignoreOriginProcessor;
  23     protected $validator;
  24     protected $logger;
  25     protected $mimeGuesser;
  26     protected $fetchingErrorMessage;
  27     protected $eventDispatcher;
  28     protected $storeArticleHeaders;
  29
  30     public function __construct(Graby $graby, RuleBasedTagger $tagger, RuleBasedIgnoreOriginProcessor $ignoreOriginProcessor, ValidatorInterface $validator, LoggerInterface $logger, $fetchingErrorMessage, $storeArticleHeaders = false)
  31     {
  32         $this->graby = $graby;
  33         $this->tagger = $tagger;
  34         $this->ignoreOriginProcessor = $ignoreOriginProcessor;
  35         $this->validator = $validator;
  36         $this->logger = $logger;
  37         $this->mimeGuesser = new MimeTypeExtensionGuesser();
  38         $this->fetchingErrorMessage = $fetchingErrorMessage;
  39         $this->storeArticleHeaders = $storeArticleHeaders;
  40     }
  41
  42     /**
  43      * Update entry using either fetched or provided content.
  44      *
  45      * @param Entry  $entry                Entry to update
  46      * @param string $url                  Url of the content
  47      * @param array  $content              Array with content provided for import with AT LEAST keys title, html, url to skip the fetchContent from the url
  48      * @param bool   $disableContentUpdate Whether to skip trying to fetch content using Graby
  49      */
  50     public function updateEntry(Entry $entry, $url, array $content = [], $disableContentUpdate = false)
  51     {
  52         $this->graby->toggleImgNoReferrer(true);
  53         if (!empty($content['html'])) {
  54             $content['html'] = $this->graby->cleanupHtml($content['html'], $url);
  55         }
  56
  57         if ((empty($content) || false === $this->validateContent($content)) && false === $disableContentUpdate) {
  58             $fetchedContent = $this->graby->fetchContent($url);
  59
  60             $fetchedContent['title'] = $this->sanitizeContentTitle(
  61                 $fetchedContent['title'],
  62                 isset($fetchedContent['headers']['content-type']) ? $fetchedContent['headers']['content-type'] : ''
  63             );
  64
  65             // when content is imported, we have information in $content
  66             // in case fetching content goes bad, we'll keep the imported information instead of overriding them
  67             if (empty($content) || $fetchedContent['html'] !== $this->fetchingErrorMessage) {
  68                 $content = $fetchedContent;
  69             }
  70         }
  71
  72         // be sure to keep the url in case of error
  73         // so we'll be able to refetch it in the future
  74         $content['url'] = !empty($content['url']) ? $content['url'] : $url;
  75
  76         // In one case (at least in tests), url is empty here
  77         // so we set it using $url provided in the updateEntry call.
  78         // Not sure what are the other possible cases where this property is empty
  79         if (empty($entry->getUrl()) && !empty($url)) {
  80             $entry->setUrl($url);
  81         }
  82
  83         $entry->setGivenUrl($url);
  84
  85         $this->stockEntry($entry, $content);
  86     }
  87
  88     /**
  89      * Use a Symfony validator to ensure the language is well formatted.
  90      *
  91      * @param string $value Language to validate and save
  92      */
  93     public function updateLanguage(Entry $entry, $value)
  94     {
  95         // some lang are defined as fr-FR, es-ES.
  96         // replacing - by _ might increase language support
  97         $value = str_replace('-', '_', $value);
  98
  99         $errors = $this->validator->validate(
 100             $value,
 101             (new LocaleConstraint())
 102         );
 103
 104         if (0 === \count($errors)) {
 105             $entry->setLanguage($value);
 106
 107             return;
 108         }
 109
 110         $this->logger->warning('Language validation failed. ' . (string) $errors);
 111     }
 112
 113     /**
 114      * Use a Symfony validator to ensure the preview picture is a real url.
 115      *
 116      * @param string $value URL to validate and save
 117      */
 118     public function updatePreviewPicture(Entry $entry, $value)
 119     {
 120         $errors = $this->validator->validate(
 121             $value,
 122             (new UrlConstraint())
 123         );
 124
 125         if (0 === \count($errors)) {
 126             $entry->setPreviewPicture($value);
 127
 128             return;
 129         }
 130
 131         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 132     }
 133
 134     /**
 135      * Update date.
 136      *
 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     public function setEntryDomainName(Entry $entry)
 165     {
 166         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 167         if (false !== $domainName) {
 168             $entry->setDomainName($domainName);
 169         }
 170     }
 171
 172     /**
 173      * Helper to set a default title using:
 174      * - url basename, if applicable
 175      * - hostname.
 176      */
 177     public function setDefaultEntryTitle(Entry $entry)
 178     {
 179         $url = parse_url($entry->getUrl());
 180         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 181
 182         if (empty($path)) {
 183             $path = $url['host'];
 184         }
 185
 186         $entry->setTitle($path);
 187     }
 188
 189     /**
 190      * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
 191      *
 192      * @param string $title
 193      * @param string $contentType
 194      *
 195      * @return string
 196      */
 197     private function sanitizeContentTitle($title, $contentType)
 198     {
 199         if ('application/pdf' === $contentType) {
 200             $title = $this->convertPdfEncodingToUTF8($title);
 201         }
 202
 203         return $this->sanitizeUTF8Text($title);
 204     }
 205
 206     /**
 207      * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
 208      * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
 209      *
 210      * @param $title
 211      *
 212      * @return string (maybe contains invalid UTF-8 character)
 213      */
 214     private function convertPdfEncodingToUTF8($title)
 215     {
 216         // first try UTF-8 because its easier to detect its present/absence
 217         foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
 218             if (mb_check_encoding($title, $encoding)) {
 219                 return mb_convert_encoding($title, 'UTF-8', $encoding);
 220             }
 221         }
 222
 223         return $title;
 224     }
 225
 226     /**
 227      * Remove invalid UTF-8 characters from the given string.
 228      *
 229      * @param string $rawText
 230      *
 231      * @return string
 232      */
 233     private function sanitizeUTF8Text($rawText)
 234     {
 235         if (mb_check_encoding($rawText, 'UTF-8')) {
 236             return $rawText;
 237         }
 238
 239         return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
 240     }
 241
 242     /**
 243      * Stock entry with fetched or imported content.
 244      * Will fall back to OpenGraph data if available.
 245      *
 246      * @param Entry $entry   Entry to stock
 247      * @param array $content Array with at least title, url & html
 248      */
 249     private function stockEntry(Entry $entry, array $content)
 250     {
 251         $this->updateOriginUrl($entry, $content['url']);
 252
 253         $this->setEntryDomainName($entry);
 254
 255         if (!empty($content['title'])) {
 256             $entry->setTitle($content['title']);
 257         }
 258
 259         if (empty($content['html'])) {
 260             $content['html'] = $this->fetchingErrorMessage;
 261
 262             if (!empty($content['description'])) {
 263                 $content['html'] .= '<p><i>But we found a short description: </i></p>';
 264                 $content['html'] .= $content['description'];
 265             }
 266         }
 267
 268         $entry->setContent($content['html']);
 269         $entry->setReadingTime(Utils::getReadingTime($content['html']));
 270
 271         if (!empty($content['status'])) {
 272             $entry->setHttpStatus($content['status']);
 273         }
 274
 275         if (!empty($content['authors']) && \is_array($content['authors'])) {
 276             $entry->setPublishedBy($content['authors']);
 277         }
 278
 279         if (!empty($content['headers'])) {
 280             $entry->setHeaders($content['headers']);
 281         }
 282
 283         if (!empty($content['date'])) {
 284             $this->updatePublishedAt($entry, $content['date']);
 285         }
 286
 287         if (!empty($content['language'])) {
 288             $this->updateLanguage($entry, $content['language']);
 289         }
 290
 291         $previewPictureUrl = '';
 292         if (!empty($content['image'])) {
 293             $previewPictureUrl = $content['image'];
 294         }
 295
 296         // if content is an image, define it as a preview too
 297         if (!empty($content['headers']['content-type']) && \in_array($this->mimeGuesser->guess($content['headers']['content-type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 298             $previewPictureUrl = $content['url'];
 299         } elseif (empty($previewPictureUrl)) {
 300             $this->logger->debug('Extracting images from content to provide a default preview picture');
 301             $imagesUrls = DownloadImages::extractImagesUrlsFromHtml($content['html']);
 302             $this->logger->debug(\count($imagesUrls) . ' pictures found');
 303
 304             if (!empty($imagesUrls)) {
 305                 $previewPictureUrl = $imagesUrls[0];
 306             }
 307         }
 308
 309         if (!empty($content['headers']['content-type'])) {
 310             $entry->setMimetype($content['headers']['content-type']);
 311         }
 312
 313         if (!empty($previewPictureUrl)) {
 314             $this->updatePreviewPicture($entry, $previewPictureUrl);
 315         }
 316
 317         try {
 318             $this->tagger->tag($entry);
 319         } catch (\Exception $e) {
 320             $this->logger->error('Error while trying to automatically tag an entry.', [
 321                 'entry_url' => $content['url'],
 322                 'error_msg' => $e->getMessage(),
 323             ]);
 324         }
 325     }
 326
 327     /**
 328      * Update the origin_url field when a redirection occurs
 329      * This field is set if it is empty and new url does not match ignore list.
 330      *
 331      * @param string $url
 332      */
 333     private function updateOriginUrl(Entry $entry, $url)
 334     {
 335         if (empty($url) || $entry->getUrl() === $url) {
 336             return false;
 337         }
 338
 339         $parsed_entry_url = parse_url($entry->getUrl());
 340         $parsed_content_url = parse_url($url);
 341
 342         /**
 343          * The following part computes the list of part changes between two
 344          * parse_url arrays.
 345          *
 346          * As array_diff_assoc only computes changes to go from the left array
 347          * to the right one, we make two differents arrays to have both
 348          * directions. We merge these two arrays and sort keys before passing
 349          * the result to the switch.
 350          *
 351          * The resulting array gives us all changing parts between the two
 352          * urls: scheme, host, path, query and/or fragment.
 353          */
 354         $diff_ec = array_diff_assoc($parsed_entry_url, $parsed_content_url);
 355         $diff_ce = array_diff_assoc($parsed_content_url, $parsed_entry_url);
 356
 357         $diff = array_merge($diff_ec, $diff_ce);
 358         $diff_keys = array_keys($diff);
 359         sort($diff_keys);
 360
 361         if ($this->ignoreOriginProcessor->process($entry)) {
 362             $entry->setUrl($url);
 363
 364             return false;
 365         }
 366
 367         /**
 368          * This switch case lets us apply different behaviors according to
 369          * changing parts of urls.
 370          *
 371          * As $diff_keys is an array, we provide arrays as cases. ['path'] means
 372          * 'only the path is different between the two urls' whereas
 373          * ['fragment', 'query'] means 'only fragment and query string parts are
 374          * different between the two urls'.
 375          *
 376          * Note that values in $diff_keys are sorted.
 377          */
 378         switch ($diff_keys) {
 379             case ['path']:
 380                 if (($parsed_entry_url['path'] . '/' === $parsed_content_url['path']) // diff is trailing slash, we only replace the url of the entry
 381                     || ($url === urldecode($entry->getUrl()))) { // we update entry url if new url is a decoded version of it, see EntryRepository#findByUrlAndUserId
 382                     $entry->setUrl($url);
 383                 }
 384                 break;
 385             case ['scheme']:
 386                 $entry->setUrl($url);
 387                 break;
 388             case ['fragment']:
 389                 // noop
 390                 break;
 391             default:
 392                 if (empty($entry->getOriginUrl())) {
 393                     $entry->setOriginUrl($entry->getUrl());
 394                 }
 395                 $entry->setUrl($url);
 396                 break;
 397         }
 398     }
 399
 400     /**
 401      * Validate that the given content has at least a title, an html and a url.
 402      *
 403      * @return bool true if valid otherwise false
 404      */
 405     private function validateContent(array $content)
 406     {
 407         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 408     }
 409 }