maintenance/language/generateCollationData.php

   1 <?php
   2 /**
   3  * @ingroup Maintenance
   4  * @file
   5  */
   6
   7 require_once( dirname( __FILE__ ) .'/../Maintenance.php' );
   8
   9 /**
  10  * Generate first letter data files for Collation.php
  11  */
  12 class GenerateCollationData extends Maintenance {
  13         /** The directory with source data files in it */
  14         var $dataDir;
  15
  16         /** The primary weights, indexed by codepoint */
  17         var $weights;
  18
  19         /**
  20          * A hashtable keyed by codepoint, where presence indicates that a character
  21          * has a decomposition mapping. This makes it non-preferred for group header
  22          * selection.
  23          */
  24         var $mappedChars;
  25
  26         var $debugOutFile;
  27
  28         /**
  29          * Important tertiary weights from UTS #10 section 7.2
  30          */
  31         const NORMAL_UPPERCASE = 0x08;
  32         const NORMAL_HIRAGANA = 0X0E;
  33
  34         public function __construct() {
  35                 parent::__construct();
  36                 $this->addOption( 'data-dir', 'A directory on the local filesystem ' .
  37                         'containing allkeys.txt and ucd.all.grouped.xml from unicode.org',
  38                         false, true );
  39                 $this->addOption( 'debug-output', 'Filename for sending debug output to',
  40                         false, true );
  41         }
  42
  43         public function execute() {
  44                 $this->dataDir = $this->getOption( 'data-dir', '.' );
  45                 if ( !file_exists( "{$this->dataDir}/allkeys.txt" ) ) {
  46                         $this->error( "Unable to find allkeys.txt. Please download it from " .
  47                                 "http://www.unicode.org/Public/UCA/latest/allkeys.txt and specify " .
  48                                 "its location with --data-dir=<DIR>" );
  49                         exit( 1 );
  50                 }
  51                 if ( !file_exists( "{$this->dataDir}/ucd.all.grouped.xml" ) ) {
  52                         $this->error( "Unable to find ucd.all.grouped.xml. Please download it " .
  53                                 "from http://www.unicode.org/Public/6.0.0/ucdxml/ucd.all.grouped.zip " .
  54                                 "and specify its location with --data-dir=<DIR>" );
  55                         exit( 1 );
  56                 }
  57                 $debugOutFileName = $this->getOption( 'debug-output' );
  58                 if ( $debugOutFileName ) {
  59                         $this->debugOutFile = fopen( $debugOutFileName, 'w' );
  60                         if ( !$this->debugOutFile ) {
  61                                 $this->error( "Unable to open debug output file for writing" );
  62                                 exit( 1 );
  63                         }
  64                 }
  65                 $this->loadUcd();
  66                 $this->generateFirstChars();
  67         }
  68
  69         function loadUcd() {
  70                 $uxr = new UcdXmlReader( "{$this->dataDir}/ucd.all.grouped.xml" );
  71                 $uxr->readChars( array( $this, 'charCallback' ) );
  72         }
  73
  74         function charCallback( $data ) {
  75                 // Skip non-printable characters,
  76                 // but do not skip a normal space (U+0020) since
  77                 // people like to use that as a fake no header symbol.
  78                 $category = substr( $data['gc'], 0, 1 );
  79                 if ( strpos( 'LNPS', $category ) === false
  80                         && $data['cp'] !== '0020' ) {
  81                         return;
  82                 }
  83                 $cp = hexdec( $data['cp'] );
  84
  85                 // Skip the CJK ideograph blocks, as an optimisation measure.
  86                 // UCA doesn't sort them properly anyway, without tailoring.
  87                 if ( IcuCollation::isCjk( $cp ) ) {
  88                         return;
  89                 }
  90
  91                 // Skip the composed Hangul syllables, we will use the bare Jamo
  92                 // as first letters
  93                 if ( $data['block'] == 'Hangul Syllables' ) {
  94                         return;
  95                 }
  96
  97                 // Calculate implicit weight per UTS #10 v6.0.0, sec 7.1.3
  98                 if ( $data['UIdeo'] === 'Y' ) {
  99                         if ( $data['block'] == 'CJK Unified Ideographs'
 100                                 || $data['block'] == 'CJK Compatibility Ideographs' )
 101                         {
 102                                 $base = 0xFB40;
 103                         } else {
 104                                 $base = 0xFB80;
 105                         }
 106                 } else {
 107                         $base = 0xFBC0;
 108                 }
 109                 $a = $base + ( $cp >> 15 );
 110                 $b = ( $cp & 0x7fff ) | 0x8000;
 111
 112                 $this->weights[$cp] = sprintf( ".%04X.%04X", $a, $b );
 113
 114                 if ( $data['dm'] !== '#' ) {
 115                         $this->mappedChars[$cp] = true;
 116                 }
 117
 118                 if ( $cp % 4096 == 0 ) {
 119                         print "{$data['cp']}\n";
 120                 }
 121         }
 122
 123         function generateFirstChars() {
 124                 $file = fopen( "{$this->dataDir}/allkeys.txt", 'r' );
 125                 if ( !$file ) {
 126                         $this->error( "Unable to open allkeys.txt" );
 127                         exit( 1 );
 128                 }
 129                 global $IP;
 130                 $outFile = fopen( "$IP/serialized/first-letters-root.ser", 'w' );
 131                 if ( !$outFile ) {
 132                         $this->error( "Unable to open output file first-letters-root.ser" );
 133                         exit( 1 );
 134                 }
 135
 136                 $goodTertiaryChars = array();
 137
 138                 // For each character with an entry in allkeys.txt, overwrite the implicit
 139                 // entry in $this->weights that came from the UCD.
 140                 // Also gather a list of tertiary weights, for use in selecting the group header
 141                 while ( false !== ( $line = fgets( $file ) ) ) {
 142                         // We're only interested in single-character weights, pick them out with a regex
 143                         $line = trim( $line );
 144                         if ( !preg_match( '/^([0-9A-F]+)\s*;\s*([^#]*)/', $line, $m ) ) {
 145                                 continue;
 146                         }
 147
 148                         $cp = hexdec( $m[1] );
 149                         $allWeights = trim( $m[2] );
 150                         $primary = '';
 151                         $tertiary = '';
 152
 153                         if ( !isset( $this->weights[$cp] ) ) {
 154                                 // Non-printable, ignore
 155                                 continue;
 156                         }
 157                         foreach ( StringUtils::explode( '[', $allWeights ) as $weightStr ) {
 158                                 preg_match_all( '/[*.]([0-9A-F]+)/', $weightStr, $m );
 159                                 if ( !empty( $m[1] ) ) {
 160                                         if ( $m[1][0] !== '0000' ) {
 161                                                 $primary .= '.' . $m[1][0];
 162                                         }
 163                                         if ( $m[1][2] !== '0000' ) {
 164                                                 $tertiary .= '.' . $m[1][2];
 165                                         }
 166                                 }
 167                         }
 168                         $this->weights[$cp] = $primary;
 169                         if ( $tertiary === '.0008'
 170                                 || $tertiary === '.000E' )
 171                         {
 172                                 $goodTertiaryChars[$cp] = true;
 173                         }
 174                 }
 175                 fclose( $file );
 176
 177                 // Identify groups of characters with the same primary weight
 178                 $this->groups = array();
 179                 asort( $this->weights, SORT_STRING );
 180                 $prevWeight = reset( $this->weights );
 181                 $group = array();
 182                 foreach ( $this->weights as $cp => $weight ) {
 183                         if ( $weight !== $prevWeight ) {
 184                                 $this->groups[$prevWeight] = $group;
 185                                 $prevWeight = $weight;
 186                                 if ( isset( $this->groups[$weight] ) ) {
 187                                         $group = $this->groups[$weight];
 188                                 } else {
 189                                         $group = array();
 190                                 }
 191                         }
 192                         $group[] = $cp;
 193                 }
 194                 if ( $group ) {
 195                         $this->groups[$prevWeight] = $group;
 196                 }
 197
 198                 // If one character has a given primary weight sequence, and a second
 199                 // character has a longer primary weight sequence with an initial
 200                 // portion equal to the first character, then remove the second
 201                 // character. This avoids having characters like U+A732 (double A)
 202                 // polluting the basic latin sort area.
 203
 204                 foreach ( $this->groups as $weight => $group ) {
 205                         if ( preg_match( '/(\.[0-9A-F]*)\./', $weight, $m ) ) {
 206                                 if ( isset( $this->groups[$m[1]] ) ) {
 207                                         unset( $this->groups[$weight] );
 208                                 }
 209                         }
 210                 }
 211
 212                 ksort( $this->groups, SORT_STRING );
 213
 214                 // Identify the header character in each group
 215                 $headerChars = array();
 216                 $prevChar = "\000";
 217                 $tertiaryCollator = new Collator( 'root' );
 218                 $primaryCollator = new Collator( 'root' );
 219                 $primaryCollator->setStrength( Collator::PRIMARY );
 220                 $numOutOfOrder = 0;
 221                 foreach ( $this->groups as $weight => $group ) {
 222                         $uncomposedChars = array();
 223                         $goodChars = array();
 224                         foreach ( $group as $cp ) {
 225                                 if ( isset( $goodTertiaryChars[$cp] ) ) {
 226                                         $goodChars[] = $cp;
 227                                 }
 228                                 if ( !isset( $this->mappedChars[$cp] ) ) {
 229                                         $uncomposedChars[] = $cp;
 230                                 }
 231                         }
 232                         $x = array_intersect( $goodChars, $uncomposedChars );
 233                         if ( !$x ) {
 234                                 $x = $uncomposedChars;
 235                                 if ( !$x ) {
 236                                         $x = $group;
 237                                 }
 238                         }
 239
 240                         // Use ICU to pick the lowest sorting character in the selection
 241                         $tertiaryCollator->sort( $x );
 242                         $cp = $x[0];
 243
 244                         $char = codepointToUtf8( $cp );
 245                         $headerChars[] = $char;
 246                         if ( $primaryCollator->compare( $char, $prevChar ) <= 0 ) {
 247                                 $numOutOfOrder ++;
 248                                 /*
 249                                 printf( "Out of order: U+%05X > U+%05X\n",
 250                                         utf8ToCodepoint( $prevChar ),
 251                                         utf8ToCodepoint( $char ) );
 252                                  */
 253                         }
 254                         $prevChar = $char;
 255
 256                         if ( $this->debugOutFile ) {
 257                                 fwrite( $this->debugOutFile, sprintf( "%05X %s %s (%s)\n", $cp, $weight, $char,
 258                                         implode( ' ', array_map( 'codepointToUtf8', $group ) ) ) );
 259                         }
 260                 }
 261
 262                 print "Out of order: $numOutOfOrder / " . count( $headerChars ) . "\n";
 263
 264                 fwrite( $outFile, serialize( $headerChars ) );
 265         }
 266 }
 267
 268 class UcdXmlReader {
 269         var $fileName;
 270         var $callback;
 271         var $groupAttrs;
 272         var $xml;
 273         var $blocks = array();
 274         var $currentBlock;
 275
 276         function __construct( $fileName ) {
 277                 $this->fileName = $fileName;
 278         }
 279
 280         public function readChars( $callback ) {
 281                 $this->getBlocks();
 282                 $this->currentBlock = reset( $this->blocks );
 283                 $xml = $this->open();
 284                 $this->callback = $callback;
 285
 286                 while ( $xml->name !== 'repertoire' && $xml->next() );
 287
 288                 while ( $xml->read() ) {
 289                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 290                                 if ( $xml->name === 'group' ) {
 291                                         $this->groupAttrs = $this->readAttributes();
 292                                 } elseif ( $xml->name === 'char' ) {
 293                                         $this->handleChar();
 294                                 }
 295                         } elseif ( $xml->nodeType === XMLReader::END_ELEMENT ) {
 296                                 if ( $xml->name === 'group' ) {
 297                                         $this->groupAttrs = array();
 298                                 }
 299                         }
 300                 }
 301                 $xml->close();
 302         }
 303
 304         protected function open() {
 305                 $this->xml = new XMLReader;
 306                 $this->xml->open( $this->fileName );
 307                 if ( !$this->xml ) {
 308                         throw new MWException( __METHOD__.": unable to open {$this->fileName}" );
 309                 }
 310                 while ( $this->xml->name !== 'ucd' && $this->xml->read() );
 311                 $this->xml->read();
 312                 return $this->xml;
 313         }
 314
 315         /**
 316          * Read the attributes of the current element node and return them
 317          * as an array
 318          * @return array
 319          */
 320         protected function readAttributes() {
 321                 $attrs = array();
 322                 while ( $this->xml->moveToNextAttribute() ) {
 323                         $attrs[$this->xml->name] = $this->xml->value;
 324                 }
 325                 return $attrs;
 326         }
 327
 328         protected function handleChar() {
 329                 $attrs = $this->readAttributes() + $this->groupAttrs;
 330                 if ( isset( $attrs['cp'] ) ) {
 331                         $first = $last = hexdec( $attrs['cp'] );
 332                 } else {
 333                         $first = hexdec( $attrs['first-cp'] );
 334                         $last = hexdec( $attrs['last-cp'] );
 335                         unset( $attrs['first-cp'] );
 336                         unset( $attrs['last-cp'] );
 337                 }
 338
 339                 for ( $cp = $first; $cp <= $last; $cp++ ) {
 340                         $hexCp = sprintf( "%04X", $cp );
 341                         foreach ( array( 'na', 'na1' ) as $nameProp ) {
 342                                 if ( isset( $attrs[$nameProp] ) ) {
 343                                         $attrs[$nameProp] = str_replace( '#', $hexCp, $attrs[$nameProp] );
 344                                 }
 345                         }
 346
 347                         while ( $this->currentBlock ) {
 348                                 if ( $cp < $this->currentBlock[0] ) {
 349                                         break;
 350                                 } elseif ( $cp <= $this->currentBlock[1] ) {
 351                                         $attrs['block'] = key( $this->blocks );
 352                                         break;
 353                                 } else {
 354                                         $this->currentBlock = next( $this->blocks );
 355                                 }
 356                         }
 357
 358                         $attrs['cp'] = $hexCp;
 359                         call_user_func( $this->callback, $attrs );
 360                 }
 361         }
 362
 363         public function getBlocks() {
 364                 if ( $this->blocks ) {
 365                         return $this->blocks;
 366                 }
 367
 368                 $xml = $this->open();
 369                 while ( $xml->name !== 'blocks' && $xml->read() );
 370
 371                 while ( $xml->read() ) {
 372                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 373                                 if ( $xml->name === 'block' ) {
 374                                         $attrs = $this->readAttributes();
 375                                         $first = hexdec( $attrs['first-cp'] );
 376                                         $last = hexdec( $attrs['last-cp'] );
 377                                         $this->blocks[$attrs['name']] = array( $first, $last );
 378                                 }
 379                         }
 380                 }
 381                 $xml->close();
 382                 return $this->blocks;
 383         }
 384
 385 }
 386
 387 $maintClass = 'GenerateCollationData';
 388 require_once( RUN_MAINTENANCE_IF_MAIN );
 389