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