maintenance/language/generateCollationData.php

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