includes/media/IPTC.php

   1 <?php
   2 /**
   3 *Class for some IPTC functions.
   4
   5 */
   6 class IPTC {
   7
   8         /**
   9         * This takes the results of iptcparse() and puts it into a
  10         * form that can be handled by mediawiki. Generally called from
  11         * BitmapMetadataHandler::doApp13.
  12         *
  13         * @see http://www.iptc.org/std/IIM/4.1/specification/IIMV4.1.pdf
  14         *
  15         * @param $rawData String app13 block from jpeg containing iptc/iim data
  16         * @return Array iptc metadata array
  17         */
  18         static function parse( $rawData ) {
  19                 $parsed = iptcparse( $rawData );
  20                 $data = Array();
  21                 if (!is_array($parsed)) {
  22                                 return $data;
  23                 }
  24
  25                 $c = '';
  26                 //charset info contained in tag 1:90.
  27                 if (isset($parsed['1#090']) && isset($parsed['1#090'][0])) {
  28                         $c = self::getCharset($parsed['1#090'][0]);
  29                         if ($c === false) {
  30                                 //Unknown charset. refuse to parse.
  31                                 //note: There is a different between
  32                                 //unknown and no charset specified.
  33                                 return array();
  34                         }
  35                         unset( $parsed['1#090'] );
  36                 }
  37
  38                 foreach ( $parsed as $tag => $val ) {
  39                         if ( isset( $val[0] ) && trim($val[0]) == '' ) {
  40                                 wfDebugLog('iptc', "IPTC tag $tag had only whitespace as its value.");
  41                                 continue;
  42                         }
  43                         switch( $tag ) {
  44                                 case '2#120': /*IPTC caption. mapped with exif ImageDescription*/
  45                                         $data['ImageDescription'] = self::convIPTC( $val, $c );
  46                                         break;
  47                                 case '2#116': /* copyright. Mapped with exif copyright */
  48                                         $data['Copyright'] = self::convIPTC( $val, $c );
  49                                         break;
  50                                 case '2#080': /* byline. Mapped with exif Artist */
  51                                         /* merge with byline title (2:85)
  52                                          * like how exif does it with
  53                                          * Title, person. Not sure if this is best
  54                                          * approach since we no longer have the two fields
  55                                          * separate. each byline title entry corresponds to a
  56                                          * specific byline.                          */
  57
  58                                         $bylines = self::convIPTC( $val, $c );
  59                                         if ( isset( $parsed['2#085'] ) ) {
  60                                                 $titles = self::convIPTC( $parsed['2#085'], $c );
  61                                         } else {
  62                                                 $titles = array();
  63                                         }
  64
  65                                         for ( $i = 0; $i < count( $titles ); $i++ ) {
  66                                                 if ( isset( $bylines[$i] ) ) {
  67                                                         // theoretically this should always be set
  68                                                         // but doesn't hurt to be careful.
  69                                                         $bylines[$i] = $titles[$i] . ', ' . $bylines[$i];
  70                                                 }
  71                                         }
  72                                         $data['Artist'] = $bylines;
  73                                         break;
  74                                 case '2#025': /* keywords */
  75                                         $data['Keywords'] = self::convIPTC( $val, $c );
  76                                         break;
  77                                 case '2#101': /* Country (shown)*/
  78                                         $data['CountryDest'] = self::convIPTC( $val, $c );
  79                                         break;
  80                                 case '2#095': /* state/province (shown) */
  81                                         $data['ProvinceOrStateDest'] = self::convIPTC( $val, $c );
  82                                         break;
  83                                 case '2#090': /* city (Shown) */
  84                                         $data['CityDest'] = self::convIPTC( $val, $c );
  85                                         break;
  86                                 case '2#092': /* sublocation (shown) */
  87                                         $data['SublocationDest'] = self::convIPTC( $val, $c );
  88                                         break;
  89                                 case '2#005': /* object name/title */
  90                                         $data['ObjectName'] = self::convIPTC( $val, $c );
  91                                         break;
  92                                 case '2#040': /* special instructions */
  93                                         $data['SpecialInstructions'] = self::convIPTC( $val, $c );
  94                                         break;
  95                                 case '2#105': /* headline*/
  96                                         $data['Headline'] = self::convIPTC( $val, $c );
  97                                         break;
  98                                 case '2#110': /* credit */
  99                                         /*"Identifies the provider of the objectdata,
 100                                          * not necessarily the owner/creator". */
 101                                         $data['Credit'] = self::convIPTC( $val, $c );
 102                                         break;
 103                                 case '2#115': /* source */
 104                                         /* "Identifies the original owner of the intellectual content of the
 105                                          *objectdata. This could be an agency, a member of an agency or
 106                                          *an individual." */
 107                                         $data['Source'] = self::convIPTC( $val, $c );
 108                                         break;
 109
 110                                 case '2#007': /* edit status (lead, correction, etc) */
 111                                         $data['EditStatus'] = self::convIPTC( $val, $c );
 112                                         break;
 113                                 case '2#015': /* category. deprecated. max 3 letters in theory, often more */
 114                                         $data['iimCategory'] = self::convIPTC( $val, $c );
 115                                         break;
 116                                 case '2#020': /* category. deprecated. */
 117                                         $data['iimSupplementalCategory'] = self::convIPTC( $val, $c );
 118                                         break;
 119                                 case '2#010': /*urgency (1-8. 1 most, 5 normal, 8 low priority)*/
 120                                         $data['Urgency'] = self::convIPTC( $val, $c );
 121                                         break;
 122                                 case '2#022':
 123                                         /* "Identifies objectdata that recurs often and predictably...
 124                                          * Example: Euroweather" */
 125                                         $data['FixtureIdentifier'] = self::convIPTC( $val, $c );
 126                                         break;
 127                                 case '2#026':
 128                                         /* Content location code (iso 3166 + some custom things)
 129                                          * ex: TUR (for turkey), XUN (for UN), XSP (outer space)
 130                                          * See wikipedia article on iso 3166 and appendix D of iim std. */
 131                                         $data['LocationDestCode'] = self::convIPTC( $val, $c );
 132                                         break;
 133                                 case '2#027':
 134                                         /* Content location name. Full printable name
 135                                          * of location of photo. */
 136                                         $data['LocationDest'] = self::convIPTC( $val, $c );
 137                                         break;
 138                                 case '2#065':
 139                                         /* Originating Program.
 140                                          * Combine with Program version (2:70) if present.
 141                                          */
 142                                         $software = self::convIPTC( $val, $c );
 143
 144                                         if ( count( $software ) !== 1 ) {
 145                                                 //according to iim standard this cannot have multiple values
 146                                                 //so if there is more than one, something weird is happening,
 147                                                 //and we skip it.
 148                                                 wfDebugLog( 'iptc', 'IPTC: Wrong count on 2:65 Software field' );
 149                                                 break;
 150                                         }
 151
 152                                         if ( isset( $parsed['2#070'] ) ) {
 153                                                 //if a version is set for the software.
 154                                                 $softwareVersion = self::convIPTC( $parsed['2#070'], $c );
 155                                                 unset($parsed['2#070']);
 156                                                 $data['Software'] = array( array( $software[0], $softwareVersion[0] ) );
 157                                         } else {
 158                                                 $data['Software'] = $software;
 159                                         }
 160                                         break;
 161                                 case '2#075':
 162                                         /* Object cycle.
 163                                          * a for morning (am), p for evening, b for both */
 164                                         $data['ObjectCycle'] = self::convIPTC( $val, $c );
 165                                         break;
 166                                 case '2#100':
 167                                         /* Country/Primary location code.
 168                                          * "Indicates the code of the country/primary location where the
 169                                          * intellectual property of the objectdata was created"
 170                                          * unclear how this differs from 2#026
 171                                          */
 172                                         $data['CountryCodeDest'] = self::convIPTC( $val, $c );
 173                                         break;
 174                                 case '2#103':
 175                                         /* original transmission ref.
 176                                          * "A code representing the location of original transmission ac-
 177                                          * cording to practises of the provider."
 178                                         */
 179                                         $data['OriginalTransmissionRef'] = self::convIPTC( $val, $c );
 180                                         break;
 181                                 case '2#118': /*contact*/
 182                                         $data['Contact'] = self::convIPTC( $val, $c );
 183                                         break;
 184                                 case '2#122':
 185                                         /* Writer/Editor
 186                                          * "Identification of the name of the person involved in the writing,
 187                                          * editing or correcting the objectdata or caption/abstract."
 188                                          */
 189                                         $data['Writer'] = self::convIPTC( $val, $c );
 190                                         break;
 191                                 case '2#135': /* lang code */
 192                                         $data['LanguageCode'] = self::convIPTC( $val, $c );
 193                                         break;
 194
 195                                 // Start date stuff.
 196                                 // It doesn't accept incomplete dates even though they are valid
 197                                 // according to spec.
 198                                 // Should potentially store timezone as well.
 199                                 case '2#055':
 200                                         //Date created (not date digitized).
 201                                         //Maps to exif DateTimeOriginal
 202                                         if ( isset( $parsed['2#060'] ) ) {
 203                                                 $time = $parsed['2#060'];
 204                                         } else {
 205                                                 $time = Array();
 206                                         }
 207                                         $timestamp =  self::timeHelper( $val, $time, $c );
 208                                         if ($timestamp) {
 209                                                 $data['DateTimeOriginal'] = $timestamp;
 210                                         }
 211                                         break;
 212
 213                                 case '2#062':
 214                                         //Date converted to digital representation.
 215                                         //Maps to exif DateTimeDigitized
 216                                         if ( isset( $parsed['2#063'] ) ) {
 217                                                 $time = $parsed['2#063'];
 218                                         } else {
 219                                                 $time = Array();
 220                                         }
 221                                         $timestamp =  self::timeHelper( $val, $time, $c );
 222                                         if ($timestamp) {
 223                                                 $data['DateTimeDigitized'] = $timestamp;
 224                                         }
 225                                         break;
 226
 227                                 case '2#030':
 228                                         //Date released.
 229                                         if ( isset( $parsed['2#035'] ) ) {
 230                                                 $time = $parsed['2#035'];
 231                                         } else {
 232                                                 $time = Array();
 233                                         }
 234                                         $timestamp =  self::timeHelper( $val, $time, $c );
 235                                         if ($timestamp) {
 236                                                 $data['DateTimeReleased'] = $timestamp;
 237                                         }
 238                                         break;
 239
 240                                 case '2#037':
 241                                         //Date expires.
 242                                         if ( isset( $parsed['2#038'] ) ) {
 243                                                 $time = $parsed['2#038'];
 244                                         } else {
 245                                                 $time = Array();
 246                                         }
 247                                         $timestamp =  self::timeHelper( $val, $time, $c );
 248                                         if ($timestamp) {
 249                                                 $data['DateTimeExpires'] = $timestamp;
 250                                         }
 251                                         break;
 252
 253                                 case '2#000': /* iim version */
 254                                         // unlike other tags, this is a 2-byte binary number.
 255                                         //technically this is required if there is iptc data
 256                                         //but in practise it isn't always there.
 257                                         if ( strlen( $val[0] ) == 2 ) {
 258                                                 //if is just to be paranoid.
 259                                                 $versionValue = ord( substr( $val[0], 0, 1 ) ) * 256;
 260                                                 $versionValue += ord( substr( $val[0], 1, 1 ) );
 261                                                 $data['iimVersion'] = $versionValue;
 262                                         }
 263                                         break;
 264
 265                                 case '2#004':
 266                                         // IntellectualGenere.
 267                                         // first 4 characters are an id code
 268                                         // That we're not really interested in.
 269
 270                                         // This prop is weird, since it's
 271                                         // allowed to have multiple values
 272                                         // in iim 4.1, but not in the XMP
 273                                         // stuff. We're going to just
 274                                         // extract the first value.
 275                                         $con = self::ConvIPTC( $val, $c );
 276                                         if ( strlen( $con[0] ) < 5 ) {
 277                                                 wfDebugLog( 'iptc', 'IPTC: '
 278                                                         . '2:04 too short. '
 279                                                         . 'Ignoring.' );
 280                                                         break;
 281                                         }
 282                                         $extracted = substr( $con[0], 4 );
 283                                         $data['IntellectualGenre'] = $extracted;
 284                                         break;
 285
 286                                 case '2#012':
 287                                         // Subject News code - this is a compound field
 288                                         // at the moment we only extract the subject news
 289                                         // code, which is an 8 digit (ascii) number
 290                                         // describing the subject matter of the content.
 291                                         $codes = self::convIPTC( $val, $c );
 292                                         foreach ( $codes as $ic ) {
 293                                                 $fields = explode(':', $ic, 3 );
 294
 295                                                 if ( count( $fields ) < 2 ||
 296                                                         $fields[0] !== 'IPTC' )
 297                                                 {
 298                                                         wfDebugLog( 'IPTC', 'IPTC: '
 299                                                                 . 'Invalid 2:12 - ' . $ic );
 300                                                         break;
 301                                                 }
 302                                                 $data['SubjectNewsCode'] = $fields[1];
 303                                         }
 304                                         break;
 305
 306                                 // purposely does not do 2:125, 2:130, 2:131,
 307                                 // 2:47, 2:50, 2:45, 2:42, 2:8, 2:3
 308                                 // 2:200, 2:201, 2:202
 309                                 // or the audio stuff (2:150 to 2:154)
 310
 311                                 case '2#070':
 312                                 case '2#060':
 313                                 case '2#063':
 314                                 case '2#085':
 315                                 case '2#038':
 316                                 case '2#035':
 317                                         //ignore. Handled elsewhere.
 318                                         break;
 319
 320                                 default:
 321                                         wfDebugLog( 'iptc', "Unsupported iptc tag: $tag. Value: " . implode( ',', $val ));
 322                                         break;
 323                         }
 324
 325                 }
 326                 return $data;
 327         }
 328
 329         /**
 330         * Convert an iptc date and time tags into the exif format
 331         *
 332         * @todo Potentially this should also capture the timezone offset.
 333         * @param Array $date The date tag
 334         * @param Array $time The time tag
 335         * @param $c
 336         * @return String Date in exif format.
 337         */
 338         private static function timeHelper( $date, $time, $c ) {
 339                 if ( count( $date ) === 1 ) {
 340                         //the standard says this should always be 1
 341                         //just double checking.
 342                         list($date) = self::convIPTC( $date, $c );
 343                 } else {
 344                         return null;
 345                 }
 346
 347                 if ( count( $time ) === 1 ) {
 348                         list($time) = self::convIPTC( $time, $c );
 349                         $dateOnly = false;
 350                 } else {
 351                         $time = '000000+0000'; //placeholder
 352                         $dateOnly = true;
 353                 }
 354
 355                 if ( ! ( preg_match('/\d\d\d\d\d\d[-+]\d\d\d\d/', $time)
 356                         && preg_match('/\d\d\d\d\d\d\d\d/', $date)
 357                         && substr($date, 0, 4) !== '0000'
 358                         && substr($date, 4, 2) !== '00'
 359                         && substr($date, 6, 2) !== '00'
 360                  ) ) {
 361                         //something wrong.
 362                         // Note, this rejects some valid dates according to iptc spec
 363                         // for example: the date 00000400 means the photo was taken in
 364                         // April, but the year and day is unknown. We don't process these
 365                         // types of incomplete dates atm.
 366                         wfDebugLog( 'iptc', "IPTC: invalid time ( $time ) or date ( $date )");
 367                         return null;
 368                 }
 369
 370                 $unixTS = wfTimestamp( TS_UNIX, $date . substr( $time, 0, 6 ));
 371                 if ( $unixTS === false ) {
 372                         wfDebugLog( 'iptc', "IPTC: can't convert date to TS_UNIX: $date $time." );
 373                         return null;
 374                 }
 375
 376                 $tz = ( intval( substr( $time, 7, 2 ) ) *60*60 )
 377                         + ( intval( substr( $time, 9, 2 ) ) * 60 );
 378
 379                 if ( substr( $time, 6, 1 ) === '-' ) {
 380                         $tz = - $tz;
 381                 }
 382
 383                 $finalTimestamp = wfTimestamp( TS_EXIF, $unixTS + $tz );
 384                 if ( $finalTimestamp === false ) {
 385                         wfDebugLog( 'iptc', "IPTC: can't make final timestamp. Date: " . ( $unixTS + $tz ) );
 386                         return null;
 387                 }
 388                 if ( $dateOnly ) {
 389                         //return the date only
 390                         return substr( $finalTimestamp, 0, 10 );
 391                 } else {
 392                         return $finalTimestamp;
 393                 }
 394         }
 395
 396         /**
 397         * Helper function to convert charset for iptc values.
 398         * @param $data Mixed String or Array: The iptc string
 399         * @param $charset String: The charset
 400          *
 401          * @return string
 402         */
 403         private static function convIPTC ( $data, $charset ) {
 404                 if ( is_array( $data ) ) {
 405                         foreach ($data as &$val) {
 406                                 $val = self::convIPTCHelper( $val, $charset );
 407                         }
 408                 } else {
 409                         $data = self::convIPTCHelper( $data, $charset );
 410                 }
 411
 412                 return $data;
 413         }
 414         /**
 415         * Helper function of a helper function to convert charset for iptc values.
 416         * @param $data Mixed String or Array: The iptc string
 417         * @param $charset String: The charset
 418         *
 419         * @return string
 420         */
 421         private static function convIPTCHelper ( $data, $charset ) {
 422                 if ( $charset ) {
 423                         wfSuppressWarnings();
 424                         $data = iconv($charset, "UTF-8//IGNORE", $data);
 425                         wfRestoreWarnings();
 426                         if ($data === false) {
 427                                 $data = "";
 428                                 wfDebugLog('iptc', __METHOD__ . " Error converting iptc data charset $charset to utf-8");
 429                         }
 430                 } else {
 431                         //treat as utf-8 if is valid utf-8. otherwise pretend its windows-1252
 432                         // most of the time if there is no 1:90 tag, it is either ascii, latin1, or utf-8
 433                         $oldData = $data;
 434                         UtfNormal::quickIsNFCVerify( $data ); //make $data valid utf-8
 435                         if ($data === $oldData) {
 436                                 return $data; //if validation didn't change $data
 437                         } else {
 438                                 return self::convIPTCHelper( $oldData, 'Windows-1252' );
 439                         }
 440                 }
 441                 return trim( $data );
 442         }
 443
 444         /**
 445         * take the value of 1:90 tag and returns a charset
 446         * @param String $tag 1:90 tag.
 447         * @return string charset name or "?"
 448         * Warning, this function does not (and is not intended to) detect
 449         * all iso 2022 escape codes. In practise, the code for utf-8 is the
 450         * only code that seems to have wide use. It does detect that code.
 451         */
 452         static function getCharset($tag) {
 453
 454                 //According to iim standard, charset is defined by the tag 1:90.
 455                 //in which there are iso 2022 escape sequences to specify the character set.
 456                 //the iim standard seems to encourage that all necessary escape sequences are
 457                 //in the 1:90 tag, but says it doesn't have to be.
 458
 459                 //This is in need of more testing probably. This is definitely not complete.
 460                 //however reading the docs of some other iptc software, it appears that most iptc software
 461                 //only recognizes utf-8. If 1:90 tag is not present content is
 462                 // usually ascii or iso-8859-1 (and sometimes utf-8), but no guarantee.
 463
 464                 //This also won't work if there are more than one escape sequence in the 1:90 tag
 465                 //or if something is put in the G2, or G3 charsets, etc. It will only reliably recognize utf-8.
 466
 467                 // This is just going through the charsets mentioned in appendix C of the iim standard.
 468
 469                 //  \x1b = ESC.
 470                 switch ( $tag ) {
 471                         case "\x1b%G": //utf-8
 472                         //Also call things that are compatible with utf-8, utf-8 (e.g. ascii)
 473                         case "\x1b(B": // ascii
 474                         case "\x1b(@": // iso-646-IRV (ascii in latest version, $ different in older version)
 475                                 $c = 'UTF-8';
 476                                 break;
 477                         case "\x1b(A": //like ascii, but british.
 478                                 $c = 'ISO646-GB';
 479                                 break;
 480                         case "\x1b(C": //some obscure sweedish/finland encoding
 481                                 $c = 'ISO-IR-8-1';
 482                                 break;
 483                         case "\x1b(D":
 484                                 $c = 'ISO-IR-8-2';
 485                                 break;
 486                         case "\x1b(E": //some obscure danish/norway encoding
 487                                 $c = 'ISO-IR-9-1';
 488                                 break;
 489                         case "\x1b(F":
 490                                 $c = 'ISO-IR-9-2';
 491                                 break;
 492                         case "\x1b(G":
 493                                 $c = 'SEN_850200_B'; // aka iso 646-SE; ascii-like
 494                                 break;
 495                         case "\x1b(I":
 496                                 $c = "ISO646-IT";
 497                                 break;
 498                         case "\x1b(L":
 499                                 $c = "ISO646-PT";
 500                                 break;
 501                         case "\x1b(Z":
 502                                 $c = "ISO646-ES";
 503                                 break;
 504                         case "\x1b([":
 505                                 $c = "GREEK7-OLD";
 506                                 break;
 507                         case "\x1b(K":
 508                                 $c = "ISO646-DE";
 509                                 break;
 510                         case "\x1b(N":  //crylic
 511                                 $c = "ISO_5427";
 512                                 break;
 513                         case "\x1b(`": //iso646-NO
 514                                 $c = "NS_4551-1";
 515                                 break;
 516                         case "\x1b(f": //iso646-FR
 517                                 $c = "NF_Z_62-010";
 518                                 break;
 519                         case "\x1b(g":
 520                                 $c = "PT2"; //iso646-PT2
 521                                 break;
 522                         case "\x1b(h":
 523                                 $c = "ES2";
 524                                 break;
 525                         case "\x1b(i": //iso646-HU
 526                                 $c = "MSZ_7795.3";
 527                                 break;
 528                         case "\x1b(w":
 529                                 $c = "CSA_Z243.4-1985-1";
 530                                 break;
 531                         case "\x1b(x":
 532                                 $c = "CSA_Z243.4-1985-2";
 533                                 break;
 534                         case "\x1b\$(B":
 535                         case "\x1b\$B":
 536                         case "\x1b&@\x1b\$B":
 537                         case "\x1b&@\x1b\$(B":
 538                                 $c = "JIS_C6226-1983";
 539                                 break;
 540                         case "\x1b-A": // iso-8859-1. at least for the high code characters.
 541                         case "\x1b(@\x1b-A":
 542                         case "\x1b(B\x1b-A":
 543                                 $c = 'ISO-8859-1';
 544                                 break;
 545                         case "\x1b-B": // iso-8859-2. at least for the high code characters.
 546                                 $c = 'ISO-8859-2';
 547                                 break;
 548                         case "\x1b-C": // iso-8859-3. at least for the high code characters.
 549                                 $c = 'ISO-8859-3';
 550                                 break;
 551                         case "\x1b-D": // iso-8859-4. at least for the high code characters.
 552                                 $c = 'ISO-8859-4';
 553                                 break;
 554                         case "\x1b-E": // iso-8859-5. at least for the high code characters.
 555                                 $c = 'ISO-8859-5';
 556                                 break;
 557                         case "\x1b-F": // iso-8859-6. at least for the high code characters.
 558                                 $c = 'ISO-8859-6';
 559                                 break;
 560                         case "\x1b-G": // iso-8859-7. at least for the high code characters.
 561                                 $c = 'ISO-8859-7';
 562                                 break;
 563                         case "\x1b-H": // iso-8859-8. at least for the high code characters.
 564                                 $c = 'ISO-8859-8';
 565                                 break;
 566                         case "\x1b-I": // CSN_369103. at least for the high code characters.
 567                                 $c = 'CSN_369103';
 568                                 break;
 569                         default:
 570                                 wfDebugLog('iptc', __METHOD__ . 'Unknown charset in iptc 1:90: ' . bin2hex( $tag ) );
 571                                 //at this point just give up and refuse to parse iptc?
 572                                 $c = false;
 573                 }
 574                 return $c;
 575         }
 576 }