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'] = $this->sanitizeUTF8Text($fetchedContent['title']);
  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      * Remove invalid UTF-8 characters from the given string in following steps:
  74      * - try to interpret the given string as ISO-8859-1, convert it to UTF-8 and return it (if its valid)
  75      * - simply remove every invalid UTF-8 character and return the result (https://stackoverflow.com/a/1433665)
  76      * @param String $rawText
  77      * @return string
  78      */
  79     private function sanitizeUTF8Text(String $rawText) {
  80         if (mb_check_encoding($rawText, 'utf-8')) {
  81             return $rawText; // return because its valid utf-8 text
  82         }
  83
  84         // we assume that $text is encoded in ISO-8859-1 (and not the similar Windows-1252 or other encoding)
  85         $convertedText = utf8_encode($rawText);
  86         if (mb_check_encoding($convertedText, 'utf-8')) {
  87             return $convertedText;
  88         }
  89
  90         // last resort: simply remove invalid UTF-8 character because $rawText can have some every exotic encoding
  91         return iconv("UTF-8", "UTF-8//IGNORE", $rawText);
  92     }
  93
  94     /**
  95      * Use a Symfony validator to ensure the language is well formatted.
  96      *
  97      * @param Entry  $entry
  98      * @param string $value Language to validate and save
  99      */
 100     public function updateLanguage(Entry $entry, $value)
 101     {
 102         // some lang are defined as fr-FR, es-ES.
 103         // replacing - by _ might increase language support
 104         $value = str_replace('-', '_', $value);
 105
 106         $errors = $this->validator->validate(
 107             $value,
 108             (new LocaleConstraint())
 109         );
 110
 111         if (0 === \count($errors)) {
 112             $entry->setLanguage($value);
 113
 114             return;
 115         }
 116
 117         $this->logger->warning('Language validation failed. ' . (string) $errors);
 118     }
 119
 120     /**
 121      * Use a Symfony validator to ensure the preview picture is a real url.
 122      *
 123      * @param Entry  $entry
 124      * @param string $value URL to validate and save
 125      */
 126     public function updatePreviewPicture(Entry $entry, $value)
 127     {
 128         $errors = $this->validator->validate(
 129             $value,
 130             (new UrlConstraint())
 131         );
 132
 133         if (0 === \count($errors)) {
 134             $entry->setPreviewPicture($value);
 135
 136             return;
 137         }
 138
 139         $this->logger->warning('PreviewPicture validation failed. ' . (string) $errors);
 140     }
 141
 142     /**
 143      * Update date.
 144      *
 145      * @param Entry  $entry
 146      * @param string $value Date to validate and save
 147      */
 148     public function updatePublishedAt(Entry $entry, $value)
 149     {
 150         $date = $value;
 151
 152         // is it a timestamp?
 153         if (false !== filter_var($date, FILTER_VALIDATE_INT)) {
 154             $date = '@' . $date;
 155         }
 156
 157         try {
 158             // is it already a DateTime?
 159             // (it's inside the try/catch in case of fail to be parse time string)
 160             if (!$date instanceof \DateTime) {
 161                 $date = new \DateTime($date);
 162             }
 163
 164             $entry->setPublishedAt($date);
 165         } catch (\Exception $e) {
 166             $this->logger->warning('Error while defining date', ['e' => $e, 'url' => $entry->getUrl(), 'date' => $value]);
 167         }
 168     }
 169
 170     /**
 171      * Helper to extract and save host from entry url.
 172      *
 173      * @param Entry $entry
 174      */
 175     public function setEntryDomainName(Entry $entry)
 176     {
 177         $domainName = parse_url($entry->getUrl(), PHP_URL_HOST);
 178         if (false !== $domainName) {
 179             $entry->setDomainName($domainName);
 180         }
 181     }
 182
 183     /**
 184      * Helper to set a default title using:
 185      * - url basename, if applicable
 186      * - hostname.
 187      *
 188      * @param Entry $entry
 189      */
 190     public function setDefaultEntryTitle(Entry $entry)
 191     {
 192         $url = parse_url($entry->getUrl());
 193         $path = pathinfo($url['path'], PATHINFO_BASENAME);
 194
 195         if (empty($path)) {
 196             $path = $url['host'];
 197         }
 198
 199         $entry->setTitle($path);
 200     }
 201
 202     /**
 203      * Stock entry with fetched or imported content.
 204      * Will fall back to OpenGraph data if available.
 205      *
 206      * @param Entry $entry   Entry to stock
 207      * @param array $content Array with at least title, url & html
 208      */
 209     private function stockEntry(Entry $entry, array $content)
 210     {
 211         $entry->setUrl($content['url']);
 212
 213         $this->setEntryDomainName($entry);
 214
 215         if (!empty($content['title'])) {
 216             $entry->setTitle($content['title']);
 217         } elseif (!empty($content['open_graph']['og_title'])) {
 218             $entry->setTitle($content['open_graph']['og_title']);
 219         }
 220
 221         $html = $content['html'];
 222         if (false === $html) {
 223             $html = $this->fetchingErrorMessage;
 224
 225             if (!empty($content['open_graph']['og_description'])) {
 226                 $html .= '<p><i>But we found a short description: </i></p>';
 227                 $html .= $content['open_graph']['og_description'];
 228             }
 229         }
 230
 231         $entry->setContent($html);
 232         $entry->setReadingTime(Utils::getReadingTime($html));
 233
 234         if (!empty($content['status'])) {
 235             $entry->setHttpStatus($content['status']);
 236         }
 237
 238         if (!empty($content['authors']) && \is_array($content['authors'])) {
 239             $entry->setPublishedBy($content['authors']);
 240         }
 241
 242         if (!empty($content['all_headers']) && $this->storeArticleHeaders) {
 243             $entry->setHeaders($content['all_headers']);
 244         }
 245
 246         if (!empty($content['date'])) {
 247             $this->updatePublishedAt($entry, $content['date']);
 248         }
 249
 250         if (!empty($content['language'])) {
 251             $this->updateLanguage($entry, $content['language']);
 252         }
 253
 254         if (!empty($content['open_graph']['og_image'])) {
 255             $this->updatePreviewPicture($entry, $content['open_graph']['og_image']);
 256         }
 257
 258         // if content is an image, define it as a preview too
 259         if (!empty($content['content_type']) && \in_array($this->mimeGuesser->guess($content['content_type']), ['jpeg', 'jpg', 'gif', 'png'], true)) {
 260             $this->updatePreviewPicture($entry, $content['url']);
 261         }
 262
 263         if (!empty($content['content_type'])) {
 264             $entry->setMimetype($content['content_type']);
 265         }
 266
 267         try {
 268             $this->tagger->tag($entry);
 269         } catch (\Exception $e) {
 270             $this->logger->error('Error while trying to automatically tag an entry.', [
 271                 'entry_url' => $content['url'],
 272                 'error_msg' => $e->getMessage(),
 273             ]);
 274         }
 275     }
 276
 277     /**
 278      * Validate that the given content has at least a title, an html and a url.
 279      *
 280      * @param array $content
 281      *
 282      * @return bool true if valid otherwise false
 283      */
 284     private function validateContent(array $content)
 285     {
 286         return !empty($content['title']) && !empty($content['html']) && !empty($content['url']);
 287     }
 288 }