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