src/Wallabag/CoreBundle/Helper/DownloadImages.php

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