includes/media/JpegMetadataExtractor.php

   1 <?php
   2 /**
   3 * Class for reading jpegs and extracting metadata.
   4 * see also BitmapMetadataHandler.
   5 *
   6 * Based somewhat on GIFMetadataExtrator.
   7 */
   8 class JpegMetadataExtractor {
   9
  10         const MAX_JPEG_SEGMENTS = 200;
  11         // the max segment is a sanity check.
  12         // A jpeg file should never even remotely have
  13         // that many segments. Your average file has about 10.
  14
  15         /** Function to extract metadata segments of interest from jpeg files
  16         * based on GIFMetadataExtractor.
  17         *
  18         * we can almost use getimagesize to do this
  19         * but gis doesn't support having multiple app1 segments
  20         * and those can't extract xmp on files containing both exif and xmp data
  21         *
  22         * @param String $filename name of jpeg file
  23         * @return Array of interesting segments.
  24         * @throws MWException if given invalid file.
  25         */
  26         static function segmentSplitter ( $filename ) {
  27                 $showXMP = function_exists( 'xml_parser_create_ns' );
  28
  29                 $segmentCount = 0;
  30
  31                 $segments = array(
  32                         'XMP_ext' => array(),
  33                         'COM' => array(),
  34                         'PSIR' => array(),
  35                 );
  36
  37                 if ( !$filename ) {
  38                         throw new MWException( "No filename specified for " . __METHOD__ );
  39                 }
  40                 if ( !file_exists( $filename ) || is_dir( $filename ) ) {
  41                         throw new MWException( "Invalid file $filename passed to " . __METHOD__ );
  42                 }
  43
  44                 $fh = fopen( $filename, "rb" );
  45
  46                 if ( !$fh ) {
  47                         throw new MWException( "Could not open file $filename" );
  48                 }
  49
  50                 $buffer = fread( $fh, 2 );
  51                 if ( $buffer !== "\xFF\xD8" ) {
  52                         throw new MWException( "Not a jpeg, no SOI" );
  53                 }
  54                 while ( !feof( $fh ) ) {
  55                         $buffer = fread( $fh, 1 );
  56                         $segmentCount++;
  57                         if ( $segmentCount > self::MAX_JPEG_SEGMENTS ) {
  58                                 // this is just a sanity check
  59                                 throw new MWException( 'Too many jpeg segments. Aborting' );
  60                         }
  61                         if ( $buffer !== "\xFF" ) {
  62                                 throw new MWException( "Error reading jpeg file marker. Expected 0xFF but got " . bin2hex( $buffer ) );
  63                         }
  64
  65                         $buffer = fread( $fh, 1 );
  66                         while( $buffer === "\xFF" && !feof( $fh ) ) {
  67                                 // Skip through any 0xFF padding bytes.
  68                                 $buffer = fread( $fh, 1 );
  69                         }
  70                         if ( $buffer === "\xFE" ) {
  71
  72                                 // COM section -- file comment
  73                                 // First see if valid utf-8,
  74                                 // if not try to convert it to windows-1252.
  75                                 $com = $oldCom = trim( self::jpegExtractMarker( $fh ) );
  76                                 UtfNormal::quickIsNFCVerify( $com );
  77                                 // turns $com to valid utf-8.
  78                                 // thus if no change, its utf-8, otherwise its something else.
  79                                 if ( $com !== $oldCom ) {
  80                                         wfSuppressWarnings();
  81                                         $com = $oldCom = iconv( 'windows-1252', 'UTF-8//IGNORE', $oldCom );
  82                                         wfRestoreWarnings();
  83                                 }
  84                                 // Try it again, if its still not a valid string, then probably
  85                                 // binary junk or some really weird encoding, so don't extract.
  86                                 UtfNormal::quickIsNFCVerify( $com );
  87                                 if ( $com === $oldCom ) {
  88                                         $segments["COM"][] = $oldCom;
  89                                 } else {
  90                                         wfDebug( __METHOD__ . ' Ignoring JPEG comment as is garbage.' );
  91                                 }
  92
  93                         } elseif ( $buffer === "\xE1" ) {
  94                                 // APP1 section (Exif, XMP, and XMP extended)
  95                                 // only extract if XMP is enabled.
  96                                 $temp = self::jpegExtractMarker( $fh );
  97                                 // check what type of app segment this is.
  98                                 if ( substr( $temp, 0, 29 ) === "http://ns.adobe.com/xap/1.0/\x00" && $showXMP ) {
  99                                         $segments["XMP"] = substr( $temp, 29 );
 100                                 } elseif ( substr( $temp, 0, 35 ) === "http://ns.adobe.com/xmp/extension/\x00" && $showXMP ) {
 101                                         $segments["XMP_ext"][] = substr( $temp, 35 );
 102                                 } elseif ( substr( $temp, 0, 29 ) === "XMP\x00://ns.adobe.com/xap/1.0/\x00" && $showXMP ) {
 103                                         // Some images (especially flickr images) seem to have this.
 104                                         // I really have no idea what the deal is with them, but
 105                                         // whatever...
 106                                         $segments["XMP"] = substr( $temp, 29 );
 107                                         wfDebug( __METHOD__ . ' Found XMP section with wrong app identifier '
 108                                                 . "Using anyways.\n" );
 109                                 } elseif ( substr( $temp, 0, 6 ) === "Exif\0\0" ) {
 110                                         // Just need to find out what the byte order is.
 111                                         // because php's exif plugin sucks...
 112                                         // This is a II for little Endian, MM for big. Not a unicode BOM.
 113                                         $byteOrderMarker = substr( $temp, 6, 2 );
 114                                         if ( $byteOrderMarker === 'MM' ) {
 115                                                 $segments['byteOrder'] = 'BE';
 116                                         } elseif ( $byteOrderMarker === 'II' ) {
 117                                                 $segments['byteOrder'] = 'LE';
 118                                         } else {
 119                                                 wfDebug( __METHOD__ . ' Invalid byte ordering?!' );
 120                                         }
 121                                 }
 122                         } elseif ( $buffer === "\xED" ) {
 123                                 // APP13 - PSIR. IPTC and some photoshop stuff
 124                                 $temp = self::jpegExtractMarker( $fh );
 125                                 if ( substr( $temp, 0, 14 ) === "Photoshop 3.0\x00" ) {
 126                                         $segments["PSIR"][] = $temp;
 127                                 }
 128                         } elseif ( $buffer === "\xD9" || $buffer === "\xDA" ) {
 129                                 // EOI - end of image or SOS - start of scan. either way we're past any interesting segments
 130                                 return $segments;
 131                         } else {
 132                                 // segment we don't care about, so skip
 133                                 $size = wfUnpack( "nint", fread( $fh, 2 ), 2 );
 134                                 if ( $size['int'] <= 2 ) throw new MWException( "invalid marker size in jpeg" );
 135                                 fseek( $fh, $size['int'] - 2, SEEK_CUR );
 136                         }
 137
 138                 }
 139                 // shouldn't get here.
 140                 throw new MWException( "Reached end of jpeg file unexpectedly" );
 141         }
 142
 143         /**
 144         * Helper function for jpegSegmentSplitter
 145         * @param &$fh FileHandle for jpeg file
 146         * @return string data content of segment.
 147         */
 148         private static function jpegExtractMarker( &$fh ) {
 149                 $size = wfUnpack( "nint", fread( $fh, 2 ), 2 );
 150                 if ( $size['int'] <= 2 ) {
 151                         throw new MWException( "invalid marker size in jpeg" );
 152                 }
 153                 $segment = fread( $fh, $size['int'] - 2 );
 154                 if ( strlen( $segment ) !== $size['int'] - 2 ) {
 155                         throw new MWException( "Segment shorter than expected" );
 156                 }
 157                 return $segment;
 158         }
 159
 160         /**
 161         * This reads the photoshop image resource.
 162         * Currently it only compares the iptc/iim hash
 163         * with the stored hash, which is used to determine the precedence
 164         * of the iptc data. In future it may extract some other info, like
 165         * url of copyright license.
 166         *
 167         * This should generally be called by BitmapMetadataHandler::doApp13()
 168         *
 169         * @param String $app13 photoshop psir app13 block from jpg.
 170         * @throws MWException (It gets caught next level up though)
 171         * @return String if the iptc hash is good or not.
 172         */
 173         public static function doPSIR ( $app13 ) {
 174                 if ( !$app13 ) {
 175                         throw new MWException( "No App13 segment given" );
 176                 }
 177                 // First compare hash with real thing
 178                 // 0x404 contains IPTC, 0x425 has hash
 179                 // This is used to determine if the iptc is newer than
 180                 // the xmp data, as xmp programs update the hash,
 181                 // where non-xmp programs don't.
 182
 183                 $offset = 14; // skip past PHOTOSHOP 3.0 identifier. should already be checked.
 184                 $appLen = strlen( $app13 );
 185                 $realHash = "";
 186                 $recordedHash = "";
 187
 188                 // the +12 is the length of an empty item.
 189                 while ( $offset + 12 <= $appLen ) {
 190                         $valid = true;
 191                         if ( substr( $app13, $offset, 4 ) !== '8BIM' ) {
 192                                 // its supposed to be 8BIM
 193                                 // but apparently sometimes isn't esp. in
 194                                 // really old jpg's
 195                                 $valid = false;
 196                         }
 197                         $offset += 4;
 198                         $id = substr( $app13, $offset, 2 );
 199                         // id is a 2 byte id number which identifies
 200                         // the piece of info this record contains.
 201
 202                         $offset += 2;
 203
 204                         // some record types can contain a name, which
 205                         // is a pascal string 0-padded to be an even
 206                         // number of bytes. Most times (and any time
 207                         // we care) this is empty, making it two null bytes.
 208
 209                         $lenName = ord( substr( $app13, $offset, 1 ) ) + 1;
 210                         // we never use the name so skip it. +1 for length byte
 211                         if ( $lenName % 2 == 1 ) {
 212                                 $lenName++;
 213                         } // pad to even.
 214                         $offset += $lenName;
 215
 216                         // now length of data (unsigned long big endian)
 217                         $lenData = wfUnpack( 'Nlen', substr( $app13, $offset, 4 ), 4 );
 218                         // PHP can take issue with very large unsigned ints and make them negative.
 219                         // Which should never ever happen, as this has to be inside a segment
 220                         // which is limited to a 16 bit number.
 221                         if ( $lenData['len'] < 0 ) throw new MWException( "Too big PSIR (" . $lenData['len'] . ')' );
 222
 223                         $offset += 4; // 4bytes length field;
 224
 225                         // this should not happen, but check.
 226                         if ( $lenData['len'] + $offset > $appLen ) {
 227                                 throw new MWException( "PSIR data too long. (item length=" . $lenData['len']
 228                                         . "; offset=$offset; total length=$appLen)" );
 229                         }
 230
 231                         if ( $valid ) {
 232                                 switch ( $id ) {
 233                                         case "\x04\x04":
 234                                                 // IPTC block
 235                                                 $realHash = md5( substr( $app13, $offset, $lenData['len'] ), true );
 236                                                 break;
 237                                         case "\x04\x25":
 238                                                 $recordedHash = substr( $app13, $offset, $lenData['len'] );
 239                                                 break;
 240                                 }
 241                         }
 242
 243                         // if odd, add 1 to length to account for
 244                         // null pad byte.
 245                         if ( $lenData['len'] % 2 == 1 ) $lenData['len']++;
 246                         $offset += $lenData['len'];
 247
 248                 }
 249
 250                 if ( !$realHash || !$recordedHash ) {
 251                         return 'iptc-no-hash';
 252                 } elseif ( $realHash === $recordedHash ) {
 253                         return 'iptc-good-hash';
 254                 } else { /*$realHash !== $recordedHash */
 255                         return 'iptc-bad-hash';
 256                 }
 257         }
 258 }