maintenance/language/generateNormalizerData.php

   1 <?php
   2
   3 require_once( dirname( __FILE__ ) . '/../Maintenance.php' );
   4
   5 require_once( dirname( __FILE__ ) . '/../../includes/normal/UtfNormalUtil.php' );
   6
   7 /**
   8  * Generates normalizer data files for Arabic and Malayalam.
   9  * For NFC see includes/normal.
  10  */
  11 class GenerateNormalizerData extends Maintenance {
  12         var $dataFile;
  13
  14         public function __construct() {
  15                 parent::__construct();
  16                 $this->addOption( 'unicode-data-file', 'The local location of the data file ' .
  17                         'from http://unicode.org/Public/UNIDATA/UnicodeData.txt', false, true );
  18         }
  19
  20         public function execute() {
  21                 if ( !$this->hasOption( 'unicode-data-file' ) ) {
  22                         $this->dataFile = 'UnicodeData.txt';
  23                         if ( !file_exists( $this->dataFile ) ) {
  24                                 $this->error( "Unable to find UnicodeData.txt. Please specify its location with --unicode-data-file=<FILE>" );
  25                                 exit( 1 );
  26                         }
  27                 } else {
  28                         $this->dataFile = $this->getOption( 'unicode-data-file' );
  29                         if ( !file_exists( $this->dataFile ) ) {
  30                                 $this->error( 'Unable to find the specified data file.' );
  31                                 exit( 1 );
  32                         }
  33                 }
  34
  35                 $this->generateArabic();
  36                 $this->generateMalayalam();
  37         }
  38
  39         function generateArabic() {
  40                 $file = fopen( $this->dataFile, 'r' );
  41                 if ( !$file ) {
  42                         $this->error( 'Unable to open the data file.' );
  43                         exit( 1 );
  44                 }
  45
  46                 // For the file format, see http://www.unicode.org/reports/tr44/
  47                 $fieldNames = array(
  48                         'Code',
  49                         'Name',
  50                         'General_Category',
  51                         'Canonical_Combining_Class',
  52                         'Bidi_Class',
  53                         'Decomposition_Type_Mapping',
  54                         'Numeric_Type_Value',
  55                         'Bidi_Mirrored',
  56                         'Unicode_1_Name',
  57                         'ISO_Comment',
  58                         'Simple_Uppercase_Mapping',
  59                         'Simple_Lowercase_Mapping',
  60                         'Simple_Titlecase_Mapping'
  61                 );
  62
  63                 $pairs = array();
  64
  65                 $lineNum = 0;
  66                 while ( false !== ( $line = fgets( $file ) ) ) {
  67                         ++$lineNum;
  68
  69                         # Strip comments
  70                         $line = trim( substr( $line, 0, strcspn( $line, '#' ) ) );
  71                         if ( $line === '' ) {
  72                                 continue;
  73                         }
  74
  75                         # Split fields
  76                         $numberedData = explode( ';', $line );
  77                         $data = array();
  78                         foreach ( $fieldNames as $number => $name ) {
  79                                 $data[$name] = $numberedData[$number];
  80                         }
  81
  82                         $code = base_convert( $data['Code'], 16, 10 );
  83                         if ( ( $code >= 0xFB50 && $code <= 0xFDFF ) # Arabic presentation forms A
  84                                 || ( $code >= 0xFE70 && $code <= 0xFEFF ) ) # Arabic presentation forms B
  85                         {
  86                                 if ( $data['Decomposition_Type_Mapping'] === '' ) {
  87                                         // No decomposition
  88                                         continue;
  89                                 }
  90                                 if ( !preg_match( '/^ *(<\w*>) +([0-9A-F ]*)$/',
  91                                         $data['Decomposition_Type_Mapping'], $m ) )
  92                                 {
  93                                         $this->error( "Can't parse Decomposition_Type/Mapping on line $lineNum" );
  94                                         $this->error( $line );
  95                                         continue;
  96                                 }
  97
  98                                 $source = hexSequenceToUtf8( $data['Code'] );
  99                                 $dest = hexSequenceToUtf8( $m[2] );
 100                                 $pairs[$source] = $dest;
 101                         }
 102                 }
 103
 104                 global $IP;
 105                 file_put_contents( "$IP/serialized/normalize-ar.ser", serialize( $pairs ) );
 106                 echo "ar: " . count( $pairs ) . " pairs written.\n";
 107         }
 108
 109         function generateMalayalam() {
 110                 $hexPairs = array(
 111                         # From http://unicode.org/versions/Unicode5.1.0/#Malayalam_Chillu_Characters
 112                         '0D23 0D4D 200D' => '0D7A',
 113                         '0D28 0D4D 200D' => '0D7B',
 114                         '0D30 0D4D 200D' => '0D7C',
 115                         '0D32 0D4D 200D' => '0D7D',
 116                         '0D33 0D4D 200D' => '0D7E',
 117
 118                         # From http://permalink.gmane.org/gmane.science.linguistics.wikipedia.technical/46413
 119                         '0D15 0D4D 200D' => '0D7F',
 120                 );
 121
 122                 $pairs = array();
 123                 foreach ( $hexPairs as $hexSource => $hexDest ) {
 124                         $source = hexSequenceToUtf8( $hexSource );
 125                         $dest = hexSequenceToUtf8( $hexDest );
 126                         $pairs[$source] = $dest;
 127                 }
 128
 129                 global $IP;
 130                 file_put_contents( "$IP/serialized/normalize-ml.ser", serialize( $pairs ) );
 131                 echo "ml: " . count( $pairs ) . " pairs written.\n";
 132         }
 133 }
 134
 135 $maintClass = 'GenerateNormalizerData';
 136 require_once( DO_MAINTENANCE );
 137