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      * Use a Symfony validator to ensure the language is well formatted.
  74      *
  75      * @param Entry  $entry
  76      * @param string $value Language to validate and save
  77      */
  78     public function updateLanguage(Entry $entry, $value)
  79     {
  80         // some lang are defined as fr-FR, es-ES.
  81         // replacing - by _ might increase language support
  82         $value = str_replace('-', '_', $value);
  83
  84         $errors = $this->validator->validate(
  85             $value,
  86             (new LocaleConstraint())
  87         );
  88
  89         if (0 === \count($errors)) {
  90             $entry->setLanguage($value);
  91
  92             return;
  93         }
  94
  95         $this->logger->warning('Language validation failed. ' . (string) $errors);
  96     }
  97
  98     /**
  99      * Use a Symfony validator to ensure the preview picture is a real url.
 100      *
 101      * @param Entry  $entry
 102      * @param string $value URL to validate and save
 103      */
 104     public function updatePreviewPicture(Entry $entry, $value)
 105     {
 106         $errors = $this->validator->validate(
 107             $value,
 108             (new UrlConstraint())
 109         );
 110
 111         if (0 === \count($errors)) {
 112             $entry->setPreviewPicture($value);
 113
 114             return;
 115         }
 116
 117         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 118     }
 119
 120     /**
 121      * Update date.
 122      *
 123      * @param Entry  $entry
 124      * @param string $value Date to validate and save
 125      */
 126     public function updatePublishedAt(Entry $entry, $value)
 127     {
 128         $date = $value;
 129
 130         // is it a timestamp?
 131         if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
 132             $date = '@' . $date;
 133         }
 134
 135         try {
 136             // is it already a DateTime?
 137             // (it's inside the try/catch in case of fail to be parse time string)
 138             if (!$date instanceof \DateTime) {
 139                 $date = new \DateTime($date);
 140             }
 141
 142             $entry->setPublishedAt($date);
 143         } catch (\Exception $e) {
 144             $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
 145         }
 146     }
 147
 148     /**
 149      * Helper to extract and save host from entry url.
 150      *
 151      * @param Entry $entry
 152      */
 153     public function setEntryDomainName(Entry $entry)
 154     {
 155         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 156         if (false !== $domainName) {
 157             $entry->setDomainName($domainName);
 158         }
 159     }
 160
 161     /**
 162      * Helper to set a default title using:
 163      * - url basename, if applicable
 164      * - hostname.
 165      *
 166      * @param Entry $entry
 167      */
 168     public function setDefaultEntryTitle(Entry $entry)
 169     {
 170         $url = parse_url($entry->getUrl());
 171         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 172
 173         if (empty($path)) {
 174             $path = $url['host'];
 175         }
 176
 177         $entry->setTitle($path);
 178     }
 179
 180     /**
 181      * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
 182      *
 183      * @param $title
 184      * @param $contentType
 185      *
 186      * @return string
 187      */
 188     private function sanitizeContentTitle($title, $contentType)
 189     {
 190         if ('application/pdf' === $contentType) {
 191             $title = $this->convertPdfEncodingToUTF8($title);
 192         }
 193
 194         return $this->sanitizeUTF8Text($title);
 195     }
 196
 197     /**
 198      * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
 199      * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
 200      *
 201      * @param $title
 202      *
 203      * @return string (maybe contains invalid UTF-8 character)
 204      */
 205     private function convertPdfEncodingToUTF8($title)
 206     {
 207         // first try UTF-8 because its easier to detect its present/absence
 208         foreach (['UTF-8', 'UTF-16BE', 'WINDOWS-1252'] as $encoding) {
 209             if (mb_check_encoding($title, $encoding)) {
 210                 return mb_convert_encoding($title, 'UTF-8', $encoding);
 211             }
 212         }
 213
 214         return $title;
 215     }
 216
 217     /**
 218      * Remove invalid UTF-8 characters from the given string.
 219      *
 220      * @param string $rawText
 221      *
 222      * @return string
 223      */
 224     private function sanitizeUTF8Text($rawText)
 225     {
 226         if (mb_check_encoding($rawText, 'UTF-8')) {
 227             return $rawText;
 228         }
 229
 230         return iconv('UTF-8', 'UTF-8//IGNORE', $rawText);
 231     }
 232
 233     /**
 234      * Stock entry with fetched or imported content.
 235      * Will fall back to OpenGraph data if available.
 236      *
 237      * @param Entry $entry   Entry to stock
 238      * @param array $content Array with at least title, url & html
 239      */
 240     private function stockEntry(Entry $entry, array $content)
 241     {
 242         $entry->setUrl($content['url']);
 243
 244         $this->setEntryDomainName($entry);
 245
 246         if (!empty($content['title'])) {
 247             $entry->setTitle($content['title']);
 248         } elseif (!empty($content['open_graph']['og_title'])) {
 249             $entry->setTitle($content['open_graph']['og_title']);
 250         }
 251
 252         $html = $content['html'];
 253         if (false === $html) {
 254             $html = $this->fetchingErrorMessage;
 255
 256             if (!empty($content['open_graph']['og_description'])) {
 257                 $html .= '<p><i>But we found a short description: </i></p>';
 258                 $html .= $content['open_graph']['og_description'];
 259             }
 260         }
 261
 262         $entry->setContent($html);
 263         $entry->setReadingTime(Utils::getReadingTime($html));
 264
 265         if (!empty($content['status'])) {
 266             $entry->setHttpStatus($content['status']);
 267         }
 268
 269         if (!empty($content['authors']) && \is_array($content['authors'])) {
 270             $entry->setPublishedBy($content['authors']);
 271         }
 272
 273         if (!empty($content['all_headers']) && $this->storeArticleHeaders) {
 274             $entry->setHeaders($content['all_headers']);
 275         }
 276
 277         if (!empty($content['date'])) {
 278             $this->updatePublishedAt($entry, $content['date']);
 279         }
 280
 281         if (!empty($content['language'])) {
 282             $this->updateLanguage($entry, $content['language']);
 283         }
 284
 285         if (!empty($content['open_graph']['og_image'])) {
 286             $this->updatePreviewPicture($entry, $content['open_graph']['og_image']);
 287         }
 288
 289         // if content is an image, define it as a preview too
 290         if (!empty($content['content_type']) && \in_array($this->mimeGuesser->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 291             $this->updatePreviewPicture($entry, $content['url']);
 292         }
 293
 294         if (!empty($content['content_type'])) {
 295             $entry->setMimetype($content['content_type']);
 296         }
 297
 298         try {
 299             $this->tagger->tag($entry);
 300         } catch (\Exception $e) {
 301             $this->logger->error('Error while trying to automatically tag an entry.', [
 302                 'entry_url' => $content['url'],
 303                 'error_msg' => $e->getMessage(),
 304             ]);
 305         }
 306     }
 307
 308     /**
 309      * Validate that the given content has at least a title, an html and a url.
 310      *
 311      * @param array $content
 312      *
 313      * @return bool true if valid otherwise false
 314      */
 315     private function validateContent(array $content)
 316     {
 317         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 318     }
 319 }