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