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  * @ingroup Maintenance
  21  * @file
  22  */
  23
  24 require_once( dirname( __FILE__ ) .'/../Maintenance.php' );
  25
  26 /**
  27  * Generate first letter data files for Collation.php
  28  */
  29 class GenerateCollationData extends Maintenance {
  30         /** The directory with source data files in it */
  31         var $dataDir;
  32
  33         /** The primary weights, indexed by codepoint */
  34         var $weights;
  35
  36         /**
  37          * A hashtable keyed by codepoint, where presence indicates that a character
  38          * has a decomposition mapping. This makes it non-preferred for group header
  39          * selection.
  40          */
  41         var $mappedChars;
  42
  43         var $debugOutFile;
  44
  45         /**
  46          * Important tertiary weights from UTS #10 section 7.2
  47          */
  48         const NORMAL_UPPERCASE = 0x08;
  49         const NORMAL_HIRAGANA = 0X0E;
  50
  51         public function __construct() {
  52                 parent::__construct();
  53                 $this->addOption( 'data-dir', 'A directory on the local filesystem ' .
  54                         'containing allkeys.txt and ucd.all.grouped.xml from unicode.org',
  55                         false, true );
  56                 $this->addOption( 'debug-output', 'Filename for sending debug output to',
  57                         false, true );
  58         }
  59
  60         public function execute() {
  61                 $this->dataDir = $this->getOption( 'data-dir', '.' );
  62                 if ( !file_exists( "{$this->dataDir}/allkeys.txt" ) ) {
  63                         $this->error( "Unable to find allkeys.txt. Please download it from " .
  64                                 "http://www.unicode.org/Public/UCA/latest/allkeys.txt and specify " .
  65                                 "its location with --data-dir=<DIR>" );
  66                         exit( 1 );
  67                 }
  68                 if ( !file_exists( "{$this->dataDir}/ucd.all.grouped.xml" ) ) {
  69                         $this->error( "Unable to find ucd.all.grouped.xml. Please download it " .
  70                                 "from http://www.unicode.org/Public/6.0.0/ucdxml/ucd.all.grouped.zip " .
  71                                 "and specify its location with --data-dir=<DIR>" );
  72                         exit( 1 );
  73                 }
  74                 $debugOutFileName = $this->getOption( 'debug-output' );
  75                 if ( $debugOutFileName ) {
  76                         $this->debugOutFile = fopen( $debugOutFileName, 'w' );
  77                         if ( !$this->debugOutFile ) {
  78                                 $this->error( "Unable to open debug output file for writing" );
  79                                 exit( 1 );
  80                         }
  81                 }
  82                 $this->loadUcd();
  83                 $this->generateFirstChars();
  84         }
  85
  86         function loadUcd() {
  87                 $uxr = new UcdXmlReader( "{$this->dataDir}/ucd.all.grouped.xml" );
  88                 $uxr->readChars( array( $this, 'charCallback' ) );
  89         }
  90
  91         function charCallback( $data ) {
  92                 // Skip non-printable characters,
  93                 // but do not skip a normal space (U+0020) since
  94                 // people like to use that as a fake no header symbol.
  95                 $category = substr( $data['gc'], 0, 1 );
  96                 if ( strpos( 'LNPS', $category ) === false
  97                         && $data['cp'] !== '0020' ) {
  98                         return;
  99                 }
 100                 $cp = hexdec( $data['cp'] );
 101
 102                 // Skip the CJK ideograph blocks, as an optimisation measure.
 103                 // UCA doesn't sort them properly anyway, without tailoring.
 104                 if ( IcuCollation::isCjk( $cp ) ) {
 105                         return;
 106                 }
 107
 108                 // Skip the composed Hangul syllables, we will use the bare Jamo
 109                 // as first letters
 110                 if ( $data['block'] == 'Hangul Syllables' ) {
 111                         return;
 112                 }
 113
 114                 // Calculate implicit weight per UTS #10 v6.0.0, sec 7.1.3
 115                 if ( $data['UIdeo'] === 'Y' ) {
 116                         if ( $data['block'] == 'CJK Unified Ideographs'
 117                                 || $data['block'] == 'CJK Compatibility Ideographs' )
 118                         {
 119                                 $base = 0xFB40;
 120                         } else {
 121                                 $base = 0xFB80;
 122                         }
 123                 } else {
 124                         $base = 0xFBC0;
 125                 }
 126                 $a = $base + ( $cp >> 15 );
 127                 $b = ( $cp & 0x7fff ) | 0x8000;
 128
 129                 $this->weights[$cp] = sprintf( ".%04X.%04X", $a, $b );
 130
 131                 if ( $data['dm'] !== '#' ) {
 132                         $this->mappedChars[$cp] = true;
 133                 }
 134
 135                 if ( $cp % 4096 == 0 ) {
 136                         print "{$data['cp']}\n";
 137                 }
 138         }
 139
 140         function generateFirstChars() {
 141                 $file = fopen( "{$this->dataDir}/allkeys.txt", 'r' );
 142                 if ( !$file ) {
 143                         $this->error( "Unable to open allkeys.txt" );
 144                         exit( 1 );
 145                 }
 146                 global $IP;
 147                 $outFile = fopen( "$IP/serialized/first-letters-root.ser", 'w' );
 148                 if ( !$outFile ) {
 149                         $this->error( "Unable to open output file first-letters-root.ser" );
 150                         exit( 1 );
 151                 }
 152
 153                 $goodTertiaryChars = array();
 154
 155                 // For each character with an entry in allkeys.txt, overwrite the implicit
 156                 // entry in $this->weights that came from the UCD.
 157                 // Also gather a list of tertiary weights, for use in selecting the group header
 158                 while ( false !== ( $line = fgets( $file ) ) ) {
 159                         // We're only interested in single-character weights, pick them out with a regex
 160                         $line = trim( $line );
 161                         if ( !preg_match( '/^([0-9A-F]+)\s*;\s*([^#]*)/', $line, $m ) ) {
 162                                 continue;
 163                         }
 164
 165                         $cp = hexdec( $m[1] );
 166                         $allWeights = trim( $m[2] );
 167                         $primary = '';
 168                         $tertiary = '';
 169
 170                         if ( !isset( $this->weights[$cp] ) ) {
 171                                 // Non-printable, ignore
 172                                 continue;
 173                         }
 174                         foreach ( StringUtils::explode( '[', $allWeights ) as $weightStr ) {
 175                                 preg_match_all( '/[*.]([0-9A-F]+)/', $weightStr, $m );
 176                                 if ( !empty( $m[1] ) ) {
 177                                         if ( $m[1][0] !== '0000' ) {
 178                                                 $primary .= '.' . $m[1][0];
 179                                         }
 180                                         if ( $m[1][2] !== '0000' ) {
 181                                                 $tertiary .= '.' . $m[1][2];
 182                                         }
 183                                 }
 184                         }
 185                         $this->weights[$cp] = $primary;
 186                         if ( $tertiary === '.0008'
 187                                 || $tertiary === '.000E' )
 188                         {
 189                                 $goodTertiaryChars[$cp] = true;
 190                         }
 191                 }
 192                 fclose( $file );
 193
 194                 // Identify groups of characters with the same primary weight
 195                 $this->groups = array();
 196                 asort( $this->weights, SORT_STRING );
 197                 $prevWeight = reset( $this->weights );
 198                 $group = array();
 199                 foreach ( $this->weights as $cp => $weight ) {
 200                         if ( $weight !== $prevWeight ) {
 201                                 $this->groups[$prevWeight] = $group;
 202                                 $prevWeight = $weight;
 203                                 if ( isset( $this->groups[$weight] ) ) {
 204                                         $group = $this->groups[$weight];
 205                                 } else {
 206                                         $group = array();
 207                                 }
 208                         }
 209                         $group[] = $cp;
 210                 }
 211                 if ( $group ) {
 212                         $this->groups[$prevWeight] = $group;
 213                 }
 214
 215                 // If one character has a given primary weight sequence, and a second
 216                 // character has a longer primary weight sequence with an initial
 217                 // portion equal to the first character, then remove the second
 218                 // character. This avoids having characters like U+A732 (double A)
 219                 // polluting the basic latin sort area.
 220
 221                 foreach ( $this->groups as $weight => $group ) {
 222                         if ( preg_match( '/(\.[0-9A-F]*)\./', $weight, $m ) ) {
 223                                 if ( isset( $this->groups[$m[1]] ) ) {
 224                                         unset( $this->groups[$weight] );
 225                                 }
 226                         }
 227                 }
 228
 229                 ksort( $this->groups, SORT_STRING );
 230
 231                 // Identify the header character in each group
 232                 $headerChars = array();
 233                 $prevChar = "\000";
 234                 $tertiaryCollator = new Collator( 'root' );
 235                 $primaryCollator = new Collator( 'root' );
 236                 $primaryCollator->setStrength( Collator::PRIMARY );
 237                 $numOutOfOrder = 0;
 238                 foreach ( $this->groups as $weight => $group ) {
 239                         $uncomposedChars = array();
 240                         $goodChars = array();
 241                         foreach ( $group as $cp ) {
 242                                 if ( isset( $goodTertiaryChars[$cp] ) ) {
 243                                         $goodChars[] = $cp;
 244                                 }
 245                                 if ( !isset( $this->mappedChars[$cp] ) ) {
 246                                         $uncomposedChars[] = $cp;
 247                                 }
 248                         }
 249                         $x = array_intersect( $goodChars, $uncomposedChars );
 250                         if ( !$x ) {
 251                                 $x = $uncomposedChars;
 252                                 if ( !$x ) {
 253                                         $x = $group;
 254                                 }
 255                         }
 256
 257                         // Use ICU to pick the lowest sorting character in the selection
 258                         $tertiaryCollator->sort( $x );
 259                         $cp = $x[0];
 260
 261                         $char = codepointToUtf8( $cp );
 262                         $headerChars[] = $char;
 263                         if ( $primaryCollator->compare( $char, $prevChar ) <= 0 ) {
 264                                 $numOutOfOrder ++;
 265                                 /*
 266                                 printf( "Out of order: U+%05X > U+%05X\n",
 267                                         utf8ToCodepoint( $prevChar ),
 268                                         utf8ToCodepoint( $char ) );
 269                                  */
 270                         }
 271                         $prevChar = $char;
 272
 273                         if ( $this->debugOutFile ) {
 274                                 fwrite( $this->debugOutFile, sprintf( "%05X %s %s (%s)\n", $cp, $weight, $char,
 275                                         implode( ' ', array_map( 'codepointToUtf8', $group ) ) ) );
 276                         }
 277                 }
 278
 279                 print "Out of order: $numOutOfOrder / " . count( $headerChars ) . "\n";
 280
 281                 fwrite( $outFile, serialize( $headerChars ) );
 282         }
 283 }
 284
 285 class UcdXmlReader {
 286         var $fileName;
 287         var $callback;
 288         var $groupAttrs;
 289         var $xml;
 290         var $blocks = array();
 291         var $currentBlock;
 292
 293         function __construct( $fileName ) {
 294                 $this->fileName = $fileName;
 295         }
 296
 297         public function readChars( $callback ) {
 298                 $this->getBlocks();
 299                 $this->currentBlock = reset( $this->blocks );
 300                 $xml = $this->open();
 301                 $this->callback = $callback;
 302
 303                 while ( $xml->name !== 'repertoire' && $xml->next() );
 304
 305                 while ( $xml->read() ) {
 306                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 307                                 if ( $xml->name === 'group' ) {
 308                                         $this->groupAttrs = $this->readAttributes();
 309                                 } elseif ( $xml->name === 'char' ) {
 310                                         $this->handleChar();
 311                                 }
 312                         } elseif ( $xml->nodeType === XMLReader::END_ELEMENT ) {
 313                                 if ( $xml->name === 'group' ) {
 314                                         $this->groupAttrs = array();
 315                                 }
 316                         }
 317                 }
 318                 $xml->close();
 319         }
 320
 321         protected function open() {
 322                 $this->xml = new XMLReader;
 323                 $this->xml->open( $this->fileName );
 324                 if ( !$this->xml ) {
 325                         throw new MWException( __METHOD__.": unable to open {$this->fileName}" );
 326                 }
 327                 while ( $this->xml->name !== 'ucd' && $this->xml->read() );
 328                 $this->xml->read();
 329                 return $this->xml;
 330         }
 331
 332         /**
 333          * Read the attributes of the current element node and return them
 334          * as an array
 335          * @return array
 336          */
 337         protected function readAttributes() {
 338                 $attrs = array();
 339                 while ( $this->xml->moveToNextAttribute() ) {
 340                         $attrs[$this->xml->name] = $this->xml->value;
 341                 }
 342                 return $attrs;
 343         }
 344
 345         protected function handleChar() {
 346                 $attrs = $this->readAttributes() + $this->groupAttrs;
 347                 if ( isset( $attrs['cp'] ) ) {
 348                         $first = $last = hexdec( $attrs['cp'] );
 349                 } else {
 350                         $first = hexdec( $attrs['first-cp'] );
 351                         $last = hexdec( $attrs['last-cp'] );
 352                         unset( $attrs['first-cp'] );
 353                         unset( $attrs['last-cp'] );
 354                 }
 355
 356                 for ( $cp = $first; $cp <= $last; $cp++ ) {
 357                         $hexCp = sprintf( "%04X", $cp );
 358                         foreach ( array( 'na', 'na1' ) as $nameProp ) {
 359                                 if ( isset( $attrs[$nameProp] ) ) {
 360                                         $attrs[$nameProp] = str_replace( '#', $hexCp, $attrs[$nameProp] );
 361                                 }
 362                         }
 363
 364                         while ( $this->currentBlock ) {
 365                                 if ( $cp < $this->currentBlock[0] ) {
 366                                         break;
 367                                 } elseif ( $cp <= $this->currentBlock[1] ) {
 368                                         $attrs['block'] = key( $this->blocks );
 369                                         break;
 370                                 } else {
 371                                         $this->currentBlock = next( $this->blocks );
 372                                 }
 373                         }
 374
 375                         $attrs['cp'] = $hexCp;
 376                         call_user_func( $this->callback, $attrs );
 377                 }
 378         }
 379
 380         public function getBlocks() {
 381                 if ( $this->blocks ) {
 382                         return $this->blocks;
 383                 }
 384
 385                 $xml = $this->open();
 386                 while ( $xml->name !== 'blocks' && $xml->read() );
 387
 388                 while ( $xml->read() ) {
 389                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 390                                 if ( $xml->name === 'block' ) {
 391                                         $attrs = $this->readAttributes();
 392                                         $first = hexdec( $attrs['first-cp'] );
 393                                         $last = hexdec( $attrs['last-cp'] );
 394                                         $this->blocks[$attrs['name']] = array( $first, $last );
 395                                 }
 396                         }
 397                 }
 398                 $xml->close();
 399                 return $this->blocks;
 400         }
 401
 402 }
 403
 404 $maintClass = 'GenerateCollationData';
 405 require_once( RUN_MAINTENANCE_IF_MAIN );