generateNormalizerDataAr.php

Go to the documentation of this file.

<?php
require_once __DIR__ . '/../Maintenance.php';

use Wikimedia\StaticArrayWriter;

class GenerateNormalizerDataAr extends Maintenance {
    public function __construct() {
        parent::__construct();
        $this->addDescription( 'Generate the normalizer data file for Arabic' );
        $this->addOption( 'unicode-data-file', 'The local location of the data file ' .
            'from https://unicode.org/Public/6.0.0/ucd/UnicodeData.txt', false, true );
    }

    public function getDbType() {
        return Maintenance::DB_NONE;
    }

    public function execute() {
        if ( !$this->hasOption( 'unicode-data-file' ) ) {
            $dataFile = 'UnicodeData.txt';
            if ( !file_exists( $dataFile ) ) {
                $this->fatalError( "Unable to find UnicodeData.txt. Please specify " .
                    "its location with --unicode-data-file=<FILE>" );
            }
        } else {
            $dataFile = $this->getOption( 'unicode-data-file' );
            if ( !file_exists( $dataFile ) ) {
                $this->fatalError( 'Unable to find the specified data file.' );
            }
        }

        $file = fopen( $dataFile, 'r' );
        if ( !$file ) {
            $this->fatalError( 'Unable to open the data file.' );
        }

        // For the file format, see https://www.unicode.org/reports/tr44/
        $fieldNames = [
            'Code',
            'Name',
            'General_Category',
            'Canonical_Combining_Class',
            'Bidi_Class',
            'Decomposition_Type_Mapping',
            'Numeric_Type_Value_6',
            'Numeric_Type_Value_7',
            'Numeric_Type_Value_8',
            'Bidi_Mirrored',
            'Unicode_1_Name',
            'ISO_Comment',
            'Simple_Uppercase_Mapping',
            'Simple_Lowercase_Mapping',
            'Simple_Titlecase_Mapping'
        ];

        $pairs = [];

        $lineNum = 0;
        while ( ( $line = fgets( $file ) ) !== false ) {
            ++$lineNum;

            # Strip comments
            $line = trim( substr( $line, 0, strcspn( $line, '#' ) ) );
            if ( $line === '' ) {
                continue;
            }

            # Split fields
            $numberedData = explode( ';', $line );
            $data = [];
            foreach ( $fieldNames as $number => $name ) {
                $data[$name] = $numberedData[$number];
            }

            $code = base_convert( $data['Code'], 16, 10 );
            if ( ( $code >= 0xFB50 && $code <= 0xFDFF ) # Arabic presentation forms A
                || ( $code >= 0xFE70 && $code <= 0xFEFF ) # Arabic presentation forms B
            ) {
                if ( $data['Decomposition_Type_Mapping'] === '' ) {
                    // No decomposition
                    continue;
                }
                if ( !preg_match( '/^ *(<\w*>) +([0-9A-F ]*)$/',
                    $data['Decomposition_Type_Mapping'], $m )
                ) {
                    $this->error( "Can't parse Decomposition_Type/Mapping on line $lineNum" );
                    $this->error( $line );
                    continue;
                }

                $source = UtfNormal\Utils::hexSequenceToUtf8( $data['Code'] );
                $dest = UtfNormal\Utils::hexSequenceToUtf8( $m[2] );
                $pairs[$source] = $dest;
            }
        }

        global $IP;
        $writer = new StaticArrayWriter();
        file_put_contents( "$IP/languages/data/normalize-ar.php", $writer->create(
            $pairs,
            'File created by generateNormalizerDataAr.php'
        ) );

        echo "ar: " . count( $pairs ) . " pairs written.\n";
    }
}

$maintClass = GenerateNormalizerDataAr::class;
require_once RUN_MAINTENANCE_IF_MAIN;