src/Wallabag/CoreBundle/Helper/DownloadImages.php

   1 <?php
   2
   3 namespace Wallabag\CoreBundle\Helper;
   4
   5 use Http\Client\Common\HttpMethodsClient;
   6 use Http\Client\Common\Plugin\ErrorPlugin;
   7 use Http\Client\Common\PluginClient;
   8 use Http\Client\HttpClient;
   9 use Http\Discovery\MessageFactoryDiscovery;
  10 use Http\Message\MessageFactory;
  11 use Psr\Http\Message\ResponseInterface;
  12 use Psr\Log\LoggerInterface;
  13 use Symfony\Component\DomCrawler\Crawler;
  14 use Symfony\Component\Finder\Finder;
  15 use Symfony\Component\HttpFoundation\File\MimeType\MimeTypeExtensionGuesser;
  16
  17 class DownloadImages
  18 {
  19     const REGENERATE_PICTURES_QUALITY = 80;
  20
  21     private $client;
  22     private $baseFolder;
  23     private $logger;
  24     private $mimeGuesser;
  25     private $wallabagUrl;
  26
  27     public function __construct(HttpClient $client, $baseFolder, $wallabagUrl, LoggerInterface $logger, MessageFactory $messageFactory = null)
  28     {
  29         $this->client = new HttpMethodsClient(new PluginClient($client, [new ErrorPlugin()]), $messageFactory ?: MessageFactoryDiscovery::find());
  30         $this->baseFolder = $baseFolder;
  31         $this->wallabagUrl = rtrim($wallabagUrl, '/');
  32         $this->logger = $logger;
  33         $this->mimeGuesser = new MimeTypeExtensionGuesser();
  34
  35         $this->setFolder();
  36     }
  37
  38     /**
  39      * Process the html and extract images URLs from it.
  40      *
  41      * @param string $html
  42      *
  43      * @return string[]
  44      */
  45     public static function extractImagesUrlsFromHtml($html)
  46     {
  47         $crawler = new Crawler($html);
  48         $imagesCrawler = $crawler
  49             ->filterXpath('//img');
  50         $imagesUrls = $imagesCrawler
  51             ->extract(['src']);
  52         $imagesSrcsetUrls = self::getSrcsetUrls($imagesCrawler);
  53
  54         return array_unique(array_merge($imagesUrls, $imagesSrcsetUrls));
  55     }
  56
  57     /**
  58      * Process the html and extract image from it, save them to local and return the updated html.
  59      *
  60      * @param int    $entryId ID of the entry
  61      * @param string $html
  62      * @param string $url     Used as a base path for relative image and folder
  63      *
  64      * @return string
  65      */
  66     public function processHtml($entryId, $html, $url)
  67     {
  68         $imagesUrls = self::extractImagesUrlsFromHtml($html);
  69
  70         $relativePath = $this->getRelativePath($entryId);
  71
  72         // download and save the image to the folder
  73         foreach ($imagesUrls as $image) {
  74             $imagePath = $this->processSingleImage($entryId, $image, $url, $relativePath);
  75
  76             if (false === $imagePath) {
  77                 continue;
  78             }
  79
  80             // if image contains "&" and we can't find it in the html it might be because it's encoded as &amp;
  81             if (false !== stripos($image, '&') && false === stripos($html, $image)) {
  82                 $image = str_replace('&', '&amp;', $image);
  83             }
  84
  85             $html = str_replace($image, $imagePath, $html);
  86         }
  87
  88         return $html;
  89     }
  90
  91     /**
  92      * Process a single image:
  93      *     - retrieve it
  94      *     - re-saved it (for security reason)
  95      *     - return the new local path.
  96      *
  97      * @param int    $entryId      ID of the entry
  98      * @param string $imagePath    Path to the image to retrieve
  99      * @param string $url          Url from where the image were found
 100      * @param string $relativePath Relative local path to saved the image
 101      *
 102      * @return string Relative url to access the image from the web
 103      */
 104     public function processSingleImage($entryId, $imagePath, $url, $relativePath = null)
 105     {
 106         if (null === $imagePath) {
 107             return false;
 108         }
 109
 110         if (null === $relativePath) {
 111             $relativePath = $this->getRelativePath($entryId);
 112         }
 113
 114         $this->logger->debug('DownloadImages: working on image: ' . $imagePath);
 115
 116         $folderPath = $this->baseFolder . '/' . $relativePath;
 117
 118         // build image path
 119         $absolutePath = $this->getAbsoluteLink($url, $imagePath);
 120         if (false === $absolutePath) {
 121             $this->logger->error('DownloadImages: Can not determine the absolute path for that image, skipping.');
 122
 123             return false;
 124         }
 125
 126         try {
 127             $res = $this->client->get($absolutePath);
 128         } catch (\Exception $e) {
 129             $this->logger->error('DownloadImages: Can not retrieve image, skipping.', ['exception' => $e]);
 130
 131             return false;
 132         }
 133
 134         $ext = $this->getExtensionFromResponse($res, $imagePath);
 135         if (false === $res) {
 136             return false;
 137         }
 138
 139         $hashImage = hash('crc32', $absolutePath);
 140         $localPath = $folderPath . '/' . $hashImage . '.' . $ext;
 141
 142         try {
 143             $im = imagecreatefromstring((string) $res->getBody());
 144         } catch (\Exception $e) {
 145             $im = false;
 146         }
 147
 148         if (false === $im) {
 149             $this->logger->error('DownloadImages: Error while regenerating image', ['path' => $localPath]);
 150
 151             return false;
 152         }
 153
 154         switch ($ext) {
 155             case 'gif':
 156                 // use Imagick if available to keep GIF animation
 157                 if (class_exists('\\Imagick')) {
 158                     try {
 159                         $imagick = new \Imagick();
 160                         $imagick->readImageBlob($res->getBody());
 161                         $imagick->setImageFormat('gif');
 162                         $imagick->writeImages($localPath, true);
 163                     } catch (\Exception $e) {
 164                         // if Imagick fail, fallback to the default solution
 165                         imagegif($im, $localPath);
 166                     }
 167                 } else {
 168                     imagegif($im, $localPath);
 169                 }
 170
 171                 $this->logger->debug('DownloadImages: Re-creating gif');
 172                 break;
 173             case 'jpeg':
 174             case 'jpg':
 175                 imagejpeg($im, $localPath, self::REGENERATE_PICTURES_QUALITY);
 176                 $this->logger->debug('DownloadImages: Re-creating jpg');
 177                 break;
 178             case 'png':
 179                 imagealphablending($im, false);
 180                 imagesavealpha($im, true);
 181                 imagepng($im, $localPath, ceil(self::REGENERATE_PICTURES_QUALITY / 100 * 9));
 182                 $this->logger->debug('DownloadImages: Re-creating png');
 183         }
 184
 185         imagedestroy($im);
 186
 187         return $this->wallabagUrl . '/assets/images/' . $relativePath . '/' . $hashImage . '.' . $ext;
 188     }
 189
 190     /**
 191      * Remove all images for the given entry id.
 192      *
 193      * @param int $entryId ID of the entry
 194      */
 195     public function removeImages($entryId)
 196     {
 197         $relativePath = $this->getRelativePath($entryId);
 198         $folderPath = $this->baseFolder . '/' . $relativePath;
 199
 200         $finder = new Finder();
 201         $finder
 202             ->files()
 203             ->ignoreDotFiles(true)
 204             ->in($folderPath);
 205
 206         foreach ($finder as $file) {
 207             @unlink($file->getRealPath());
 208         }
 209
 210         @rmdir($folderPath);
 211     }
 212
 213     /**
 214      * Get images urls from the srcset image attribute.
 215      *
 216      * @param Crawler $imagesCrawler
 217      *
 218      * @return array An array of urls
 219      */
 220     private static function getSrcsetUrls(Crawler $imagesCrawler)
 221     {
 222         $urls = [];
 223         $iterator = $imagesCrawler
 224             ->getIterator();
 225         while ($iterator->valid()) {
 226             $srcsetAttribute = $iterator->current()->getAttribute('srcset');
 227             if ('' !== $srcsetAttribute) {
 228                 // Couldn't start with " OR ' OR a white space
 229                 // Could be one or more white space
 230                 // Must be one or more digits followed by w OR x
 231                 $pattern = "/(?:[^\"'\s]+\s*(?:\d+[wx])+)/";
 232                 preg_match_all($pattern, $srcsetAttribute, $matches);
 233                 $srcset = \call_user_func_array('array_merge', $matches);
 234                 $srcsetUrls = array_map(function ($src) {
 235                     return trim(explode(' ', $src, 2)[0]);
 236                 }, $srcset);
 237                 $urls = array_merge($srcsetUrls, $urls);
 238             }
 239             $iterator->next();
 240         }
 241
 242         return $urls;
 243     }
 244
 245     /**
 246      * Setup base folder where all images are going to be saved.
 247      */
 248     private function setFolder()
 249     {
 250         // if folder doesn't exist, attempt to create one and store the folder name in property $folder
 251         if (!file_exists($this->baseFolder)) {
 252             mkdir($this->baseFolder, 0755, true);
 253         }
 254     }
 255
 256     /**
 257      * Generate the folder where we are going to save images based on the entry url.
 258      *
 259      * @param int $entryId ID of the entry
 260      *
 261      * @return string
 262      */
 263     private function getRelativePath($entryId)
 264     {
 265         $hashId = hash('crc32', $entryId);
 266         $relativePath = $hashId[0] . '/' . $hashId[1] . '/' . $hashId;
 267         $folderPath = $this->baseFolder . '/' . $relativePath;
 268
 269         if (!file_exists($folderPath)) {
 270             mkdir($folderPath, 0777, true);
 271         }
 272
 273         $this->logger->debug('DownloadImages: Folder used for that Entry id', ['folder' => $folderPath, 'entryId' => $entryId]);
 274
 275         return $relativePath;
 276     }
 277
 278     /**
 279      * Make an $url absolute based on the $base.
 280      *
 281      * @see Graby->makeAbsoluteStr
 282      *
 283      * @param string $base Base url
 284      * @param string $url  Url to make it absolute
 285      *
 286      * @return false|string
 287      */
 288     private function getAbsoluteLink($base, $url)
 289     {
 290         if (preg_match('!^https?://!i', $url)) {
 291             // already absolute
 292             return $url;
 293         }
 294
 295         $base = new \SimplePie_IRI($base);
 296
 297         // remove '//' in URL path (causes URLs not to resolve properly)
 298         if (isset($base->ipath)) {
 299             $base->ipath = preg_replace('!//+!', '/', $base->ipath);
 300         }
 301
 302         if ($absolute = \SimplePie_IRI::absolutize($base, $url)) {
 303             return $absolute->get_uri();
 304         }
 305
 306         $this->logger->error('DownloadImages: Can not make an absolute link', ['base' => $base, 'url' => $url]);
 307
 308         return false;
 309     }
 310
 311     /**
 312      * Retrieve and validate the extension from the response of the url of the image.
 313      *
 314      * @param ResponseInterface $res       Http Response
 315      * @param string   $imagePath Path from the src image from the content (used for log only)
 316      *
 317      * @return string|false Extension name or false if validation failed
 318      */
 319     private function getExtensionFromResponse(ResponseInterface $res, $imagePath)
 320     {
 321         $ext = $this->mimeGuesser->guess(current($res->getHeader('content-type')));
 322         $this->logger->debug('DownloadImages: Checking extension', ['ext' => $ext, 'header' => $res->getHeader('content-type')]);
 323
 324         // ok header doesn't have the extension, try a different way
 325         if (empty($ext)) {
 326             $types = [
 327                 'jpeg' => "\xFF\xD8\xFF",
 328                 'gif' => 'GIF',
 329                 'png' => "\x89\x50\x4e\x47\x0d\x0a",
 330             ];
 331             $bytes = substr((string) $res->getBody(), 0, 8);
 332
 333             foreach ($types as $type => $header) {
 334                 if (0 === strpos($bytes, $header)) {
 335                     $ext = $type;
 336                     break;
 337                 }
 338             }
 339
 340             $this->logger->debug('DownloadImages: Checking extension (alternative)', ['ext' => $ext]);
 341         }
 342
 343         if (!\in_array($ext, ['jpeg', 'jpg', 'gif', 'png'], true)) {
 344             $this->logger->error('DownloadImages: Processed image with not allowed extension. Skipping: ' . $imagePath);
 345
 346             return false;
 347         }
 348
 349         return $ext;
 350     }
 351 }