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                 $category = substr( $data['gc'], 0, 1 );
  73                 if ( strpos( 'LNPS', $category ) === false ) {
  74                         return;
  75                 }
  76                 $cp = hexdec( $data['cp'] );
  77
  78                 // Skip the CJK ideograph blocks, as an optimisation measure.
  79                 // UCA doesn't sort them properly anyway, without tailoring.
  80                 if ( IcuCollation::isCjk( $cp ) ) {
  81                         return;
  82                 }
  83
  84                 // Skip the composed Hangul syllables, we will use the bare Jamo
  85                 // as first letters
  86                 if ( $data['block'] == 'Hangul Syllables' ) {
  87                         return;
  88                 }
  89
  90                 // Calculate implicit weight per UTS #10 v6.0.0, sec 7.1.3
  91                 if ( $data['UIdeo'] === 'Y' ) {
  92                         if ( $data['block'] == 'CJK Unified Ideographs'
  93                                 || $data['block'] == 'CJK Compatibility Ideographs' )
  94                         {
  95                                 $base = 0xFB40;
  96                         } else {
  97                                 $base = 0xFB80;
  98                         }
  99                 } else {
 100                         $base = 0xFBC0;
 101                 }
 102                 $a = $base + ( $cp >> 15 );
 103                 $b = ( $cp & 0x7fff ) | 0x8000;
 104
 105                 $this->weights[$cp] = sprintf( ".%04X.%04X", $a, $b );
 106
 107                 if ( $data['dm'] !== '#' ) {
 108                         $this->mappedChars[$cp] = true;
 109                 }
 110
 111                 if ( $cp % 4096 == 0 ) {
 112                         print "{$data['cp']}\n";
 113                 }
 114         }
 115
 116         function generateFirstChars() {
 117                 $file = fopen( "{$this->dataDir}/allkeys.txt", 'r' );
 118                 if ( !$file ) {
 119                         $this->error( "Unable to open allkeys.txt" );
 120                         exit( 1 );
 121                 }
 122                 global $IP;
 123                 $outFile = fopen( "$IP/serialized/first-letters-root.ser", 'w' );
 124                 if ( !$outFile ) {
 125                         $this->error( "Unable to open output file first-letters-root.ser" );
 126                         exit( 1 );
 127                 }
 128
 129                 $goodTertiaryChars = array();
 130
 131                 // For each character with an entry in allkeys.txt, overwrite the implicit
 132                 // entry in $this->weights that came from the UCD.
 133                 // Also gather a list of tertiary weights, for use in selecting the group header
 134                 while ( false !== ( $line = fgets( $file ) ) ) {
 135                         // We're only interested in single-character weights, pick them out with a regex
 136                         $line = trim( $line );
 137                         if ( !preg_match( '/^([0-9A-F]+)\s*;\s*([^#]*)/', $line, $m ) ) {
 138                                 continue;
 139                         }
 140
 141                         $cp = hexdec( $m[1] );
 142                         $allWeights = trim( $m[2] );
 143                         $primary = '';
 144                         $tertiary = '';
 145
 146                         if ( !isset( $this->weights[$cp] ) ) {
 147                                 // Non-printable, ignore
 148                                 continue;
 149                         }
 150                         foreach ( StringUtils::explode( '[', $allWeights ) as $weightStr ) {
 151                                 preg_match_all( '/[*.]([0-9A-F]+)/', $weightStr, $m );
 152                                 if ( !empty( $m[1] ) ) {
 153                                         if ( $m[1][0] !== '0000' ) {
 154                                                 $primary .= '.' . $m[1][0];
 155                                         }
 156                                         if ( $m[1][2] !== '0000' ) {
 157                                                 $tertiary .= '.' . $m[1][2];
 158                                         }
 159                                 }
 160                         }
 161                         $this->weights[$cp] = $primary;
 162                         if ( $tertiary === '.0008'
 163                                 || $tertiary === '.000E' )
 164                         {
 165                                 $goodTertiaryChars[$cp] = true;
 166                         }
 167                 }
 168                 fclose( $file );
 169
 170                 // Identify groups of characters with the same primary weight
 171                 $this->groups = array();
 172                 asort( $this->weights, SORT_STRING );
 173                 $prevWeight = reset( $this->weights );
 174                 $group = array();
 175                 foreach ( $this->weights as $cp => $weight ) {
 176                         if ( $weight !== $prevWeight ) {
 177                                 $this->groups[$prevWeight] = $group;
 178                                 $prevWeight = $weight;
 179                                 if ( isset( $this->groups[$weight] ) ) {
 180                                         $group = $this->groups[$weight];
 181                                 } else {
 182                                         $group = array();
 183                                 }
 184                         }
 185                         $group[] = $cp;
 186                 }
 187                 if ( $group ) {
 188                         $this->groups[$prevWeight] = $group;
 189                 }
 190
 191                 // If one character has a given primary weight sequence, and a second
 192                 // character has a longer primary weight sequence with an initial
 193                 // portion equal to the first character, then remove the second
 194                 // character. This avoids having characters like U+A732 (double A)
 195                 // polluting the basic latin sort area.
 196                 $prevWeights = array();
 197                 foreach ( $this->groups as $weight => $group ) {
 198                         if ( preg_match( '/(\.[0-9A-F]*)\./', $weight, $m ) ) {
 199                                 if ( isset( $this->groups[$m[1]] ) ) {
 200                                         unset( $this->groups[$weight] );
 201                                 }
 202                         }
 203                 }
 204
 205                 ksort( $this->groups, SORT_STRING );
 206
 207                 // Identify the header character in each group
 208                 $headerChars = array();
 209                 $prevChar = "\000";
 210                 $tertiaryCollator = new Collator( 'root' );
 211                 $primaryCollator = new Collator( 'root' );
 212                 $primaryCollator->setStrength( Collator::PRIMARY );
 213                 $numOutOfOrder = 0;
 214                 foreach ( $this->groups as $weight => $group ) {
 215                         $uncomposedChars = array();
 216                         $goodChars = array();
 217                         foreach ( $group as $cp ) {
 218                                 if ( isset( $goodTertiaryChars[$cp] ) ) {
 219                                         $goodChars[] = $cp;
 220                                 }
 221                                 if ( !isset( $this->mappedChars[$cp] ) ) {
 222                                         $uncomposedChars[] = $cp;
 223                                 }
 224                         }
 225                         $x = array_intersect( $goodChars, $uncomposedChars );
 226                         if ( !$x ) {
 227                                 $x = $uncomposedChars;
 228                                 if ( !$x ) {
 229                                         $x = $group;
 230                                 }
 231                         }
 232
 233                         // Use ICU to pick the lowest sorting character in the selection
 234                         $tertiaryCollator->sort( $x );
 235                         $cp = $x[0];
 236
 237                         $char = codepointToUtf8( $cp );
 238                         $headerChars[] = $char;
 239                         if ( $primaryCollator->compare( $char, $prevChar ) <= 0 ) {
 240                                 $numOutOfOrder ++;
 241                                 /*
 242                                 printf( "Out of order: U+%05X > U+%05X\n",
 243                                         utf8ToCodepoint( $prevChar ),
 244                                         utf8ToCodepoint( $char ) );
 245                                  */
 246                         }
 247                         $prevChar = $char;
 248
 249                         if ( $this->debugOutFile ) {
 250                                 fwrite( $this->debugOutFile, sprintf( "%05X %s %s (%s)\n", $cp, $weight, $char,
 251                                         implode( ' ', array_map( 'codepointToUtf8', $group ) ) ) );
 252                         }
 253                 }
 254
 255                 print "Out of order: $numOutOfOrder / " . count( $headerChars ) . "\n";
 256
 257                 fwrite( $outFile, serialize( $headerChars ) );
 258         }
 259 }
 260
 261 class UcdXmlReader {
 262         var $fileName;
 263         var $callback;
 264         var $groupAttrs;
 265         var $xml;
 266         var $blocks = array();
 267         var $currentBlock;
 268
 269         function __construct( $fileName ) {
 270                 $this->fileName = $fileName;
 271         }
 272
 273         public function readChars( $callback ) {
 274                 $this->getBlocks();
 275                 $this->currentBlock = reset( $this->blocks );
 276                 $xml = $this->open();
 277                 $this->callback = $callback;
 278
 279                 while ( $xml->name !== 'repertoire' && $xml->next() );
 280
 281                 while ( $xml->read() ) {
 282                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 283                                 if ( $xml->name === 'group' ) {
 284                                         $this->groupAttrs = $this->readAttributes();
 285                                 } elseif ( $xml->name === 'char' ) {
 286                                         $this->handleChar();
 287                                 }
 288                         } elseif ( $xml->nodeType === XMLReader::END_ELEMENT ) {
 289                                 if ( $xml->name === 'group' ) {
 290                                         $this->groupAttrs = array();
 291                                 }
 292                         }
 293                 }
 294                 $xml->close();
 295         }
 296
 297         protected function open() {
 298                 $this->xml = new XMLReader;
 299                 $this->xml->open( $this->fileName );
 300                 if ( !$this->xml ) {
 301                         throw new MWException( __METHOD__.": unable to open {$this->fileName}" );
 302                 }
 303                 while ( $this->xml->name !== 'ucd' && $this->xml->read() );
 304                 $this->xml->read();
 305                 return $this->xml;
 306         }
 307
 308         /**
 309          * Read the attributes of the current element node and return them
 310          * as an array
 311          */
 312         protected function readAttributes() {
 313                 $attrs = array();
 314                 while ( $this->xml->moveToNextAttribute() ) {
 315                         $attrs[$this->xml->name] = $this->xml->value;
 316                 }
 317                 return $attrs;
 318         }
 319
 320         protected function handleChar() {
 321                 $attrs = $this->readAttributes() + $this->groupAttrs;
 322                 if ( isset( $attrs['cp'] ) ) {
 323                         $first = $last = hexdec( $attrs['cp'] );
 324                 } else {
 325                         $first = hexdec( $attrs['first-cp'] );
 326                         $last = hexdec( $attrs['last-cp'] );
 327                         unset( $attrs['first-cp'] );
 328                         unset( $attrs['last-cp'] );
 329                 }
 330
 331                 for ( $cp = $first; $cp <= $last; $cp++ ) {
 332                         $hexCp = sprintf( "%04X", $cp );
 333                         foreach ( array( 'na', 'na1' ) as $nameProp ) {
 334                                 if ( isset( $attrs[$nameProp] ) ) {
 335                                         $attrs[$nameProp] = str_replace( '#', $hexCp, $attrs[$nameProp] );
 336                                 }
 337                         }
 338
 339                         while ( $this->currentBlock ) {
 340                                 if ( $cp < $this->currentBlock[0] ) {
 341                                         break;
 342                                 } elseif ( $cp <= $this->currentBlock[1] ) {
 343                                         $attrs['block'] = key( $this->blocks );
 344                                         break;
 345                                 } else {
 346                                         $this->currentBlock = next( $this->blocks );
 347                                 }
 348                         }
 349
 350                         $attrs['cp'] = $hexCp;
 351                         call_user_func( $this->callback, $attrs );
 352                 }
 353         }
 354
 355         public function getBlocks() {
 356                 if ( $this->blocks ) {
 357                         return $this->blocks;
 358                 }
 359
 360                 $xml = $this->open();
 361                 while ( $xml->name !== 'blocks' && $xml->read() );
 362
 363                 while ( $xml->read() ) {
 364                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 365                                 if ( $xml->name === 'block' ) {
 366                                         $attrs = $this->readAttributes();
 367                                         $first = hexdec( $attrs['first-cp'] );
 368                                         $last = hexdec( $attrs['last-cp'] );
 369                                         $this->blocks[$attrs['name']] = array( $first, $last );
 370                                 }
 371                         }
 372                 }
 373                 $xml->close();
 374                 return $this->blocks;
 375         }
 376
 377 }
 378
 379 $maintClass = 'GenerateCollationData';
 380 require_once( DO_MAINTENANCE );
 381