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