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