maintenance/language/generateCollationData.php

   1 <?php
   2 /**
   3  * Maintenance script to generate first letter data files for Collation.php.
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup MaintenanceLanguage
  22  */
  23
  24 require_once __DIR__ . '/../Maintenance.php';
  25
  26 /**
  27  * Generate first letter data files for Collation.php
  28  *
  29  * @ingroup MaintenanceLanguage
  30  */
  31 class GenerateCollationData extends Maintenance {
  32         /** The directory with source data files in it */
  33         public $dataDir;
  34
  35         /** The primary weights, indexed by codepoint */
  36         public $weights;
  37
  38         /**
  39          * A hashtable keyed by codepoint, where presence indicates that a character
  40          * has a decomposition mapping. This makes it non-preferred for group header
  41          * selection.
  42          */
  43         public $mappedChars;
  44
  45         public $debugOutFile;
  46
  47         /**
  48          * Important tertiary weights from UTS #10 section 7.2
  49          */
  50         const NORMAL_UPPERCASE = 0x08;
  51         const NORMAL_HIRAGANA = 0x0E;
  52
  53         public function __construct() {
  54                 parent::__construct();
  55                 $this->addOption( 'data-dir', 'A directory on the local filesystem ' .
  56                         'containing allkeys.txt and ucd.all.grouped.xml from unicode.org',
  57                         false, true );
  58                 $this->addOption( 'debug-output', 'Filename for sending debug output to',
  59                         false, true );
  60         }
  61
  62         public function execute() {
  63                 $this->dataDir = $this->getOption( 'data-dir', '.' );
  64
  65                 $allkeysPresent = file_exists( "{$this->dataDir}/allkeys.txt" );
  66                 $ucdallPresent = file_exists( "{$this->dataDir}/ucd.all.grouped.xml" );
  67
  68                 // As of January 2013, these links work for all versions of Unicode
  69                 // between 5.1 and 6.2, inclusive.
  70                 $allkeysURL = "https://www.unicode.org/Public/UCA/<Unicode version>/allkeys.txt";
  71                 $ucdallURL = "https://www.unicode.org/Public/<Unicode version>/ucdxml/ucd.all.grouped.zip";
  72
  73                 if ( !$allkeysPresent || !$ucdallPresent ) {
  74                         $icuVersion = INTL_ICU_VERSION;
  75                         $unicodeVersion = IcuCollation::getUnicodeVersionForICU();
  76
  77                         $error = "";
  78
  79                         if ( !$allkeysPresent ) {
  80                                 $error .= "Unable to find allkeys.txt. "
  81                                         . "Download it and specify its location with --data-dir=<DIR>. "
  82                                         . "\n\n";
  83                         }
  84                         if ( !$ucdallPresent ) {
  85                                 $error .= "Unable to find ucd.all.grouped.xml. "
  86                                         . "Download it, unzip, and specify its location with --data-dir=<DIR>. "
  87                                         . "\n\n";
  88                         }
  89
  90                         $versionKnown = false;
  91                         if ( version_compare( $icuVersion, "4.0", "<" ) ) {
  92                                 // Extra old version
  93                                 $error .= "You are using outdated version of ICU ($icuVersion), intended for "
  94                                         . ( $unicodeVersion ? "Unicode $unicodeVersion" : "an unknown version of Unicode" )
  95                                         . "; this file might not be avalaible for it, and it's not supported by MediaWiki. "
  96                                         . " You are on your own; consider upgrading PHP's intl extension or try "
  97                                         . "one of the files available at:";
  98                         } elseif ( version_compare( $icuVersion, "51.0", ">=" ) ) {
  99                                 // Extra recent version
 100                                 $error .= "You are using ICU $icuVersion, released after this script was last updated. "
 101                                         . "Check what is the Unicode version it is using at http://site.icu-project.org/download . "
 102                                         . "It can't be guaranteed everything will work, but appropriate file(s) should "
 103                                         . "be available at:";
 104                         } else {
 105                                 // ICU 4.0 to 50.x
 106                                 $versionKnown = true;
 107                                 $error .= "You are using ICU $icuVersion, intended for "
 108                                         . ( $unicodeVersion ? "Unicode $unicodeVersion" : "an unknown version of Unicode" )
 109                                         . ". Appropriate file(s) should be available at:";
 110                         }
 111                         $error .= "\n";
 112
 113                         if ( $versionKnown && $unicodeVersion ) {
 114                                 $allkeysURL = str_replace( "<Unicode version>", "$unicodeVersion.0", $allkeysURL );
 115                                 $ucdallURL = str_replace( "<Unicode version>", "$unicodeVersion.0", $ucdallURL );
 116                         }
 117
 118                         if ( !$allkeysPresent ) {
 119                                 $error .= "* $allkeysURL\n";
 120                         }
 121                         if ( !$ucdallPresent ) {
 122                                 $error .= "* $ucdallURL\n";
 123                         }
 124
 125                         $this->fatalError( $error );
 126                 }
 127
 128                 $debugOutFileName = $this->getOption( 'debug-output' );
 129                 if ( $debugOutFileName ) {
 130                         $this->debugOutFile = fopen( $debugOutFileName, 'w' );
 131                         if ( !$this->debugOutFile ) {
 132                                 $this->fatalError( "Unable to open debug output file for writing" );
 133                         }
 134                 }
 135                 $this->loadUcd();
 136                 $this->generateFirstChars();
 137         }
 138
 139         function loadUcd() {
 140                 $uxr = new UcdXmlReader( "{$this->dataDir}/ucd.all.grouped.xml" );
 141                 $uxr->readChars( [ $this, 'charCallback' ] );
 142         }
 143
 144         function charCallback( $data ) {
 145                 // Skip non-printable characters,
 146                 // but do not skip a normal space (U+0020) since
 147                 // people like to use that as a fake no header symbol.
 148                 $category = substr( $data['gc'], 0, 1 );
 149                 if ( strpos( 'LNPS', $category ) === false
 150                         && $data['cp'] !== '0020'
 151                 ) {
 152                         return;
 153                 }
 154                 $cp = hexdec( $data['cp'] );
 155
 156                 // Skip the CJK ideograph blocks, as an optimisation measure.
 157                 // UCA doesn't sort them properly anyway, without tailoring.
 158                 if ( IcuCollation::isCjk( $cp ) ) {
 159                         return;
 160                 }
 161
 162                 // Skip the composed Hangul syllables, we will use the bare Jamo
 163                 // as first letters
 164                 if ( $data['block'] == 'Hangul Syllables' ) {
 165                         return;
 166                 }
 167
 168                 // Calculate implicit weight per UTS #10 v6.0.0, sec 7.1.3
 169                 if ( $data['UIdeo'] === 'Y' ) {
 170                         if ( $data['block'] == 'CJK Unified Ideographs'
 171                                 || $data['block'] == 'CJK Compatibility Ideographs'
 172                         ) {
 173                                 $base = 0xFB40;
 174                         } else {
 175                                 $base = 0xFB80;
 176                         }
 177                 } else {
 178                         $base = 0xFBC0;
 179                 }
 180                 $a = $base + ( $cp >> 15 );
 181                 $b = ( $cp & 0x7fff ) | 0x8000;
 182
 183                 $this->weights[$cp] = sprintf( ".%04X.%04X", $a, $b );
 184
 185                 if ( $data['dm'] !== '#' ) {
 186                         $this->mappedChars[$cp] = true;
 187                 }
 188
 189                 if ( $cp % 4096 == 0 ) {
 190                         print "{$data['cp']}\n";
 191                 }
 192         }
 193
 194         function generateFirstChars() {
 195                 $file = fopen( "{$this->dataDir}/allkeys.txt", 'r' );
 196                 if ( !$file ) {
 197                         $this->fatalError( "Unable to open allkeys.txt" );
 198                 }
 199                 global $IP;
 200                 $outFile = fopen( "$IP/serialized/first-letters-root.ser", 'w' );
 201                 if ( !$outFile ) {
 202                         $this->fatalError( "Unable to open output file first-letters-root.ser" );
 203                 }
 204
 205                 $goodTertiaryChars = [];
 206
 207                 // For each character with an entry in allkeys.txt, overwrite the implicit
 208                 // entry in $this->weights that came from the UCD.
 209                 // Also gather a list of tertiary weights, for use in selecting the group header
 210                 while ( false !== ( $line = fgets( $file ) ) ) {
 211                         // We're only interested in single-character weights, pick them out with a regex
 212                         $line = trim( $line );
 213                         if ( !preg_match( '/^([0-9A-F]+)\s*;\s*([^#]*)/', $line, $m ) ) {
 214                                 continue;
 215                         }
 216
 217                         $cp = hexdec( $m[1] );
 218                         $allWeights = trim( $m[2] );
 219                         $primary = '';
 220                         $tertiary = '';
 221
 222                         if ( !isset( $this->weights[$cp] ) ) {
 223                                 // Non-printable, ignore
 224                                 continue;
 225                         }
 226                         foreach ( StringUtils::explode( '[', $allWeights ) as $weightStr ) {
 227                                 preg_match_all( '/[*.]([0-9A-F]+)/', $weightStr, $m );
 228                                 if ( !empty( $m[1] ) ) {
 229                                         if ( $m[1][0] !== '0000' ) {
 230                                                 $primary .= '.' . $m[1][0];
 231                                         }
 232                                         if ( $m[1][2] !== '0000' ) {
 233                                                 $tertiary .= '.' . $m[1][2];
 234                                         }
 235                                 }
 236                         }
 237                         $this->weights[$cp] = $primary;
 238                         if ( $tertiary === '.0008'
 239                                 || $tertiary === '.000E'
 240                         ) {
 241                                 $goodTertiaryChars[$cp] = true;
 242                         }
 243                 }
 244                 fclose( $file );
 245
 246                 // Identify groups of characters with the same primary weight
 247                 $this->groups = [];
 248                 asort( $this->weights, SORT_STRING );
 249                 $prevWeight = reset( $this->weights );
 250                 $group = [];
 251                 foreach ( $this->weights as $cp => $weight ) {
 252                         if ( $weight !== $prevWeight ) {
 253                                 $this->groups[$prevWeight] = $group;
 254                                 $prevWeight = $weight;
 255                                 if ( isset( $this->groups[$weight] ) ) {
 256                                         $group = $this->groups[$weight];
 257                                 } else {
 258                                         $group = [];
 259                                 }
 260                         }
 261                         $group[] = $cp;
 262                 }
 263                 if ( $group ) {
 264                         $this->groups[$prevWeight] = $group;
 265                 }
 266
 267                 // If one character has a given primary weight sequence, and a second
 268                 // character has a longer primary weight sequence with an initial
 269                 // portion equal to the first character, then remove the second
 270                 // character. This avoids having characters like U+A732 (double A)
 271                 // polluting the basic latin sort area.
 272
 273                 foreach ( $this->groups as $weight => $group ) {
 274                         if ( preg_match( '/(\.[0-9A-F]*)\./', $weight, $m ) ) {
 275                                 if ( isset( $this->groups[$m[1]] ) ) {
 276                                         unset( $this->groups[$weight] );
 277                                 }
 278                         }
 279                 }
 280
 281                 ksort( $this->groups, SORT_STRING );
 282
 283                 // Identify the header character in each group
 284                 $headerChars = [];
 285                 $prevChar = "\000";
 286                 $tertiaryCollator = new Collator( 'root' );
 287                 $primaryCollator = new Collator( 'root' );
 288                 $primaryCollator->setStrength( Collator::PRIMARY );
 289                 $numOutOfOrder = 0;
 290                 foreach ( $this->groups as $weight => $group ) {
 291                         $uncomposedChars = [];
 292                         $goodChars = [];
 293                         foreach ( $group as $cp ) {
 294                                 if ( isset( $goodTertiaryChars[$cp] ) ) {
 295                                         $goodChars[] = $cp;
 296                                 }
 297                                 if ( !isset( $this->mappedChars[$cp] ) ) {
 298                                         $uncomposedChars[] = $cp;
 299                                 }
 300                         }
 301                         $x = array_intersect( $goodChars, $uncomposedChars );
 302                         if ( !$x ) {
 303                                 $x = $uncomposedChars;
 304                                 if ( !$x ) {
 305                                         $x = $group;
 306                                 }
 307                         }
 308
 309                         // Use ICU to pick the lowest sorting character in the selection
 310                         $tertiaryCollator->sort( $x );
 311                         $cp = $x[0];
 312
 313                         $char = UtfNormal\Utils::codepointToUtf8( $cp );
 314                         $headerChars[] = $char;
 315                         if ( $primaryCollator->compare( $char, $prevChar ) <= 0 ) {
 316                                 $numOutOfOrder++;
 317                         }
 318                         $prevChar = $char;
 319
 320                         if ( $this->debugOutFile ) {
 321                                 fwrite( $this->debugOutFile, sprintf( "%05X %s %s (%s)\n", $cp, $weight, $char,
 322                                         implode( ' ', array_map( 'UtfNormal\Utils::codepointToUtf8', $group ) ) ) );
 323                         }
 324                 }
 325
 326                 print "Out of order: $numOutOfOrder / " . count( $headerChars ) . "\n";
 327
 328                 fwrite( $outFile, serialize( $headerChars ) );
 329         }
 330 }
 331
 332 class UcdXmlReader {
 333         public $fileName;
 334         public $callback;
 335         public $groupAttrs;
 336         public $xml;
 337         public $blocks = [];
 338         public $currentBlock;
 339
 340         function __construct( $fileName ) {
 341                 $this->fileName = $fileName;
 342         }
 343
 344         public function readChars( $callback ) {
 345                 $this->getBlocks();
 346                 $this->currentBlock = reset( $this->blocks );
 347                 $xml = $this->open();
 348                 $this->callback = $callback;
 349
 350                 while ( $xml->name !== 'repertoire' && $xml->next() );
 351
 352                 while ( $xml->read() ) {
 353                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 354                                 if ( $xml->name === 'group' ) {
 355                                         $this->groupAttrs = $this->readAttributes();
 356                                 } elseif ( $xml->name === 'char' ) {
 357                                         $this->handleChar();
 358                                 }
 359                         } elseif ( $xml->nodeType === XMLReader::END_ELEMENT ) {
 360                                 if ( $xml->name === 'group' ) {
 361                                         $this->groupAttrs = [];
 362                                 }
 363                         }
 364                 }
 365                 $xml->close();
 366         }
 367
 368         protected function open() {
 369                 $this->xml = new XMLReader;
 370                 $this->xml->open( $this->fileName );
 371                 if ( !$this->xml ) {
 372                         throw new MWException( __METHOD__ . ": unable to open {$this->fileName}" );
 373                 }
 374                 while ( $this->xml->name !== 'ucd' && $this->xml->read() );
 375                 $this->xml->read();
 376
 377                 return $this->xml;
 378         }
 379
 380         /**
 381          * Read the attributes of the current element node and return them
 382          * as an array
 383          * @return array
 384          */
 385         protected function readAttributes() {
 386                 $attrs = [];
 387                 while ( $this->xml->moveToNextAttribute() ) {
 388                         $attrs[$this->xml->name] = $this->xml->value;
 389                 }
 390
 391                 return $attrs;
 392         }
 393
 394         protected function handleChar() {
 395                 $attrs = $this->readAttributes() + $this->groupAttrs;
 396                 if ( isset( $attrs['cp'] ) ) {
 397                         $first = $last = hexdec( $attrs['cp'] );
 398                 } else {
 399                         $first = hexdec( $attrs['first-cp'] );
 400                         $last = hexdec( $attrs['last-cp'] );
 401                         unset( $attrs['first-cp'] );
 402                         unset( $attrs['last-cp'] );
 403                 }
 404
 405                 for ( $cp = $first; $cp <= $last; $cp++ ) {
 406                         $hexCp = sprintf( "%04X", $cp );
 407                         foreach ( [ 'na', 'na1' ] as $nameProp ) {
 408                                 if ( isset( $attrs[$nameProp] ) ) {
 409                                         $attrs[$nameProp] = str_replace( '#', $hexCp, $attrs[$nameProp] );
 410                                 }
 411                         }
 412
 413                         while ( $this->currentBlock ) {
 414                                 if ( $cp < $this->currentBlock[0] ) {
 415                                         break;
 416                                 } elseif ( $cp <= $this->currentBlock[1] ) {
 417                                         $attrs['block'] = key( $this->blocks );
 418                                         break;
 419                                 } else {
 420                                         $this->currentBlock = next( $this->blocks );
 421                                 }
 422                         }
 423
 424                         $attrs['cp'] = $hexCp;
 425                         call_user_func( $this->callback, $attrs );
 426                 }
 427         }
 428
 429         public function getBlocks() {
 430                 if ( $this->blocks ) {
 431                         return $this->blocks;
 432                 }
 433
 434                 $xml = $this->open();
 435                 while ( $xml->name !== 'blocks' && $xml->read() );
 436
 437                 while ( $xml->read() ) {
 438                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 439                                 if ( $xml->name === 'block' ) {
 440                                         $attrs = $this->readAttributes();
 441                                         $first = hexdec( $attrs['first-cp'] );
 442                                         $last = hexdec( $attrs['last-cp'] );
 443                                         $this->blocks[$attrs['name']] = [ $first, $last ];
 444                                 }
 445                         }
 446                 }
 447                 $xml->close();
 448
 449                 return $this->blocks;
 450         }
 451 }
 452
 453 $maintClass = GenerateCollationData::class;
 454 require_once RUN_MAINTENANCE_IF_MAIN;