GettextFormat.php

<?php
declare( strict_types = 1 );
 
namespace MediaWiki\Extension\Translate\FileFormatSupport;
 
use InvalidArgumentException;
use LanguageCode;
use MediaWiki\Extension\Translate\MessageGroupConfiguration\MetaYamlSchemaExtender;
use MediaWiki\Extension\Translate\MessageLoading\Message;
use MediaWiki\Extension\Translate\MessageLoading\MessageCollection;
use MediaWiki\Extension\Translate\Utilities\GettextPlural;
use MediaWiki\Extension\Translate\Utilities\Utilities;
use MediaWiki\Logger\LoggerFactory;
use MediaWiki\MediaWikiServices;
use MediaWiki\Title\Title;
use RuntimeException;
use SpecialVersion;
 
class GettextFormat extends SimpleFormat implements MetaYamlSchemaExtender {
    private bool $allowPotMode = false;
    private bool $offlineMode = false;
 
    public function supportsFuzzy(): string {
        return 'yes';
    }
 
    public function getFileExtensions(): array {
        return [ '.pot', '.po' ];
    }
 
    public function setOfflineMode( bool $value ): void {
        $this->offlineMode = $value;
    }
 
    public function read( $languageCode ) {
        // This is somewhat hacky, but pot mode should only ever be used for the source language.
        // See https://phabricator.wikimedia.org/T230361
        $this->allowPotMode = $this->getGroup()->getSourceLanguage() === $languageCode;
 
        try {
            return parent::read( $languageCode );
        } finally {
            $this->allowPotMode = false;
        }
    }
 
    public function readFromVariable( string $data ): array {
        # Authors first
        $matches = [];
        preg_match_all( '/^#\s*Author:\s*(.*)$/m', $data, $matches );
        $authors = $matches[1];
 
        # Then messages and everything else
        $parsedData = $this->parseGettext( $data );
        $parsedData['AUTHORS'] = $authors;
 
        foreach ( $parsedData['MESSAGES'] as $key => $value ) {
            if ( $value === '' ) {
                unset( $parsedData['MESSAGES'][$key] );
            }
        }
 
        return $parsedData;
    }
 
    private function parseGettext( string $data ): array {
        $mangler = $this->group->getMangler();
        $useCtxtAsKey = $this->extra['CtxtAsKey'] ?? false;
        $keyAlgorithm = 'simple';
        if ( isset( $this->extra['keyAlgorithm'] ) ) {
            $keyAlgorithm = $this->extra['keyAlgorithm'];
        }
 
        $potmode = false;
 
        // Normalise newlines, to make processing easier
        $data = str_replace( "\r\n", "\n", $data );
 
        /* Delimit the file into sections, which are separated by two newlines.
         * We are permissive and accept more than two. This parsing method isn't
         * efficient wrt memory, but was easy to implement */
        $sections = preg_split( '/\n{2,}/', $data );
 
        /* First one isn't an actual message. We'll handle it specially below */
        $headerSection = array_shift( $sections );
        /* Since this is the header section, we are only interested in the tags
         * and msgid is empty. Somewhere we should extract the header comments
         * too */
        $match = $this->expectKeyword( 'msgstr', $headerSection );
        if ( $match !== null ) {
            $headerBlock = $this->formatForWiki( $match, 'trim' );
            $headers = $this->parseHeaderTags( $headerBlock );
 
            // Check for pot-mode by checking if the header is fuzzy
            $flags = $this->parseFlags( $headerSection );
            if ( in_array( 'fuzzy', $flags, true ) ) {
                $potmode = $this->allowPotMode;
            }
        } else {
            $message = "Gettext file header was not found:\n\n$data";
            throw new GettextParseException( $message );
        }
 
        $template = [];
        $messages = [];
 
        // Extract some metadata from headers for easier use
        $metadata = [];
        if ( isset( $headers['X-Language-Code'] ) ) {
            $metadata['code'] = $headers['X-Language-Code'];
        }
 
        if ( isset( $headers['X-Message-Group'] ) ) {
            $metadata['group'] = $headers['X-Message-Group'];
        }
 
        /* At this stage we are only interested how many plurals forms we should
         * be expecting when parsing the rest of this file. */
        $pluralCount = null;
        if ( $potmode ) {
            $pluralCount = 2;
        } elseif ( isset( $headers['Plural-Forms'] ) ) {
            $pluralCount = $metadata['plural'] = GettextPlural::getPluralCount( $headers['Plural-Forms'] );
        }
 
        $metadata['plural'] = $pluralCount;
 
        // Then parse the messages
        foreach ( $sections as $section ) {
            $item = $this->parseGettextSection( $section, $pluralCount );
            if ( $item === null ) {
                continue;
            }
 
            if ( $useCtxtAsKey ) {
                if ( !isset( $item['ctxt'] ) ) {
                    error_log( "ctxt missing for: $section" );
                    continue;
                }
                $key = $item['ctxt'];
            } else {
                $key = $this->generateKeyFromItem( $item, $keyAlgorithm );
            }
 
            $key = $mangler->mangle( $key );
            $messages[$key] = $potmode ? $item['id'] : $item['str'];
            $template[$key] = $item;
        }
 
        return [
            'MESSAGES' => $messages,
            'EXTRA' => [
                'TEMPLATE' => $template,
                'METADATA' => $metadata,
                'HEADERS' => $headers,
            ],
        ];
    }
 
    private function parseGettextSection( string $section, ?int $pluralCount ): ?array {
        if ( trim( $section ) === '' ) {
            return null;
        }
 
        /* These inactive sections are of no interest to us. Multiline mode
         * is needed because there may be flags or other annoying stuff
         * before the commented out sections.
         */
        if ( preg_match( '/^#~/m', $section ) ) {
            return null;
        }
 
        $item = [
            'ctxt' => false,
            'id' => '',
            'str' => '',
            'flags' => [],
            'comments' => [],
        ];
 
        $match = $this->expectKeyword( 'msgid', $section );
        if ( $match !== null ) {
            $item['id'] = $this->formatForWiki( $match );
        } else {
            throw new RuntimeException( "Unable to parse msgid:\n\n$section" );
        }
 
        $match = $this->expectKeyword( 'msgctxt', $section );
        if ( $match !== null ) {
            $item['ctxt'] = $this->formatForWiki( $match );
        }
 
        $pluralMessage = false;
        $match = $this->expectKeyword( 'msgid_plural', $section );
        if ( $match !== null ) {
            $pluralMessage = true;
            $plural = $this->formatForWiki( $match );
            $item['id'] = GettextPlural::flatten( [ $item['id'], $plural ] );
        }
 
        if ( $pluralMessage ) {
            $pluralMessageText = $this->processGettextPluralMessage( $pluralCount, $section );
 
            // Keep the translation empty if no form has translation
            if ( $pluralMessageText !== '' ) {
                $item['str'] = $pluralMessageText;
            }
        } else {
            $match = $this->expectKeyword( 'msgstr', $section );
            if ( $match !== null ) {
                $item['str'] = $this->formatForWiki( $match );
            } else {
                throw new RuntimeException( "Unable to parse msgstr:\n\n$section" );
            }
        }
 
        // Parse flags
        $flags = $this->parseFlags( $section );
        foreach ( $flags as $key => $flag ) {
            if ( $flag === 'fuzzy' ) {
                $item['str'] = TRANSLATE_FUZZY . $item['str'];
                unset( $flags[$key] );
            }
        }
        $item['flags'] = $flags;
 
        // Rest of the comments
        $matches = [];
        if ( preg_match_all( '/^#(.?) (.*)$/m', $section, $matches, PREG_SET_ORDER ) ) {
            foreach ( $matches as $match ) {
                if ( $match[1] !== ',' && !str_starts_with( $match[1], '[Wiki]' ) ) {
                    $item['comments'][$match[1]][] = $match[2];
                }
            }
        }
 
        return $item;
    }
 
    private function processGettextPluralMessage( ?int $pluralCount, string $section ): string {
        $actualForms = [];
 
        for ( $i = 0; $i < $pluralCount; $i++ ) {
            $match = $this->expectKeyword( "msgstr\\[$i\\]", $section );
 
            if ( $match !== null ) {
                $actualForms[] = $this->formatForWiki( $match );
            } else {
                $actualForms[] = '';
                error_log( "Plural $i not found, expecting total of $pluralCount for $section" );
            }
        }
 
        if ( array_sum( array_map( 'strlen', $actualForms ) ) > 0 ) {
            return GettextPlural::flatten( $actualForms );
        } else {
            return '';
        }
    }
 
    private function parseFlags( string $section ): array {
        $matches = [];
        if ( preg_match( '/^#,(.*)$/mu', $section, $matches ) ) {
            return array_map( 'trim', explode( ',', $matches[1] ) );
        } else {
            return [];
        }
    }
 
    private function expectKeyword( string $name, string $section ): ?string {
        /* Catches the multiline textblock that comes after keywords msgid,
         * msgstr, msgid_plural, msgctxt.
         */
        $poformat = '".*"\n?(^".*"$\n?)*';
 
        $matches = [];
        if ( preg_match( "/^$name\s($poformat)/mx", $section, $matches ) ) {
            return $matches[1];
        } else {
            return null;
        }
    }
 
    public function generateKeyFromItem( array $item, string $algorithm = 'simple' ): string {
        $lang = MediaWikiServices::getInstance()->getLanguageFactory()->getLanguage( 'en' );
 
        if ( $item['ctxt'] === '' ) {
            /* Messages with msgctxt as empty string should be different
             * from messages without any msgctxt. To avoid BC break make
             * the empty ctxt a special case */
            $hash = sha1( $item['id'] . 'MSGEMPTYCTXT' );
        } else {
            $hash = sha1( $item['ctxt'] . $item['id'] );
        }
 
        if ( $algorithm === 'simple' ) {
            $hash = substr( $hash, 0, 6 );
            $snippet = $lang->truncateForDatabase( $item['id'], 30, '' );
            $snippet = str_replace( ' ', '_', trim( $snippet ) );
        } else { // legacy
            $legalChars = Title::legalChars();
            $snippet = $item['id'];
            $snippet = preg_replace( "/[^$legalChars]/", ' ', $snippet );
            $snippet = preg_replace( "/[:&%\/_]/", ' ', $snippet );
            $snippet = preg_replace( '/ {2,}/', ' ', $snippet );
            $snippet = $lang->truncateForDatabase( $snippet, 30, '' );
            $snippet = str_replace( ' ', '_', trim( $snippet ) );
        }
 
        return "$hash-$snippet";
    }
 
    private function processData( string $data ): string {
        $quotePattern = '/(^"|"$\n?)/m';
        $data = preg_replace( $quotePattern, '', $data );
        return stripcslashes( $data );
    }
 
    private function handleWhitespace( string $data, string $whitespace ): string {
        if ( preg_match( '/\s$/', $data ) ) {
            if ( $whitespace === 'mark' ) {
                $data .= '\\';
            } elseif ( $whitespace === 'trim' ) {
                $data = rtrim( $data );
            } else {
                // This condition will never happen as long as $whitespace is 'mark' or 'trim'
                throw new InvalidArgumentException( "Unknown action for whitespace: $whitespace" );
            }
        }
 
        return $data;
    }
 
    private function formatForWiki( string $data, string $whitespace = 'mark' ): string {
        $data = $this->processData( $data );
        return $this->handleWhitespace( $data, $whitespace );
    }
 
    private function parseHeaderTags( string $headers ): array {
        $tags = [];
        foreach ( explode( "\n", $headers ) as $line ) {
            if ( !str_contains( $line, ':' ) ) {
                error_log( __METHOD__ . ": $line" );
            }
            [ $key, $value ] = explode( ':', $line, 2 );
            $tags[trim( $key )] = trim( $value );
        }
 
        return $tags;
    }
 
    protected function writeReal( MessageCollection $collection ): string {
        // FIXME: this should be the source language
        $pot = $this->read( 'en' ) ?? [];
        $code = $collection->code;
        $template = $this->read( $code ) ?? [];
        $output = $this->doGettextHeader( $collection, $template['EXTRA'] ?? [] );
 
        $pluralRule = GettextPlural::getPluralRule( $code );
        if ( !$pluralRule ) {
            $pluralRule = GettextPlural::getPluralRule( 'en' );
            LoggerFactory::getInstance( 'Translate' )->warning(
                "T235180: Missing Gettext plural rule for '{languagecode}'",
                [ 'languagecode' => $code ]
            );
        }
        $pluralCount = GettextPlural::getPluralCount( $pluralRule );
 
        $documentationLanguageCode = MediaWikiServices::getInstance()
            ->getMainConfig()
            ->get( 'TranslateDocumentationLanguageCode' );
        $documentationCollection = null;
        if ( is_string( $documentationLanguageCode ) ) {
            $documentationCollection = clone $collection;
            $documentationCollection->resetForNewLanguage( $documentationLanguageCode );
            $documentationCollection->loadTranslations();
        }
 
        foreach ( $collection as $key => $m ) {
            $transTemplate = $template['EXTRA']['TEMPLATE'][$key] ?? [];
            $potTemplate = $pot['EXTRA']['TEMPLATE'][$key] ?? [];
            $documentation = isset( $documentationCollection[$key] ) ?
                $documentationCollection[$key]->translation() : null;
 
            $output .= $this->formatMessageBlock(
                $key,
                $m,
                $transTemplate,
                $potTemplate,
                $pluralCount,
                $documentation
            );
        }
 
        return $output;
    }
 
    private function doGettextHeader( MessageCollection $collection, array $template ): string {
        global $wgSitename;
 
        $code = $collection->code;
        $name = Utilities::getLanguageName( $code );
        $native = Utilities::getLanguageName( $code, $code );
        $authors = $this->doAuthors( $collection );
        if ( isset( $this->extra['header'] ) ) {
            $extra = "# --\n" . $this->extra['header'];
        } else {
            $extra = '';
        }
 
        $group = $this->getGroup();
        $output =
            <<<EOT
            # Translation of {$group->getLabel()} to $name ($native)
            # Exported from $wgSitename
            #
            $authors$extra
            EOT;
 
        // Make sure there is no empty line before msgid
        $output = trim( $output ) . "\n";
 
        $specs = $template['HEADERS'] ?? [];
 
        $timestamp = wfTimestampNow();
        $specs['PO-Revision-Date'] = $this->formatTime( $timestamp );
        if ( $this->offlineMode ) {
            $specs['POT-Creation-Date'] = $this->formatTime( $timestamp );
        } else {
            $specs['X-POT-Import-Date'] = $this->formatTime( wfTimestamp( TS_MW, $this->getPotTime() ) );
        }
        $specs['Content-Type'] = 'text/plain; charset=UTF-8';
        $specs['Content-Transfer-Encoding'] = '8bit';
 
        $specs['Language'] = LanguageCode::bcp47( $this->group->mapCode( $code ) );
        MediaWikiServices::getInstance()->getHookContainer()
            ->run( 'Translate:GettextFFS:headerFields', [ &$specs, $this->group, $code ] );
        $specs['X-Generator'] = 'MediaWiki '
            . SpecialVersion::getVersion()
            . '; Translate '
            . Utilities::getVersion();
 
        if ( $this->offlineMode ) {
            $specs['X-Language-Code'] = $code;
            $specs['X-Message-Group'] = $group->getId();
        }
 
        $specs['Plural-Forms'] = GettextPlural::getPluralRule( $code )
            ?: GettextPlural::getPluralRule( 'en' );
 
        $output .= 'msgid ""' . "\n";
        $output .= 'msgstr ""' . "\n";
        $output .= '""' . "\n";
 
        foreach ( $specs as $k => $v ) {
            $output .= $this->escape( "$k: $v\n" ) . "\n";
        }
 
        $output .= "\n";
 
        return $output;
    }
 
    private function doAuthors( MessageCollection $collection ): string {
        $output = '';
        $authors = $collection->getAuthors();
        $authors = $this->filterAuthors( $authors, $collection->code );
 
        foreach ( $authors as $author ) {
            $output .= "# Author: $author\n";
        }
 
        return $output;
    }
 
    private function formatMessageBlock(
        string $key,
        Message $message,
        array $trans,
        array $pot,
        int $pluralCount,
        ?string $documentation
    ): string {
        $header = $this->formatDocumentation( $documentation );
        $content = '';
 
        $comments = $pot['comments'] ?? $trans['comments'] ?? [];
        foreach ( $comments as $type => $typecomments ) {
            foreach ( $typecomments as $comment ) {
                $header .= "#$type $comment\n";
            }
        }
 
        $flags = $pot['flags'] ?? $trans['flags'] ?? [];
        $flags = array_merge( $message->getTags(), $flags );
 
        if ( $this->offlineMode ) {
            $content .= 'msgctxt ' . $this->escape( $key ) . "\n";
        } else {
            $ctxt = $pot['ctxt'] ?? $trans['ctxt'] ?? false;
            if ( $ctxt !== false ) {
                $content .= 'msgctxt ' . $this->escape( $ctxt ) . "\n";
            }
        }
 
        $msgid = $message->definition();
        $msgstr = $message->translation() ?? '';
        if ( strpos( $msgstr, TRANSLATE_FUZZY ) !== false ) {
            $msgstr = str_replace( TRANSLATE_FUZZY, '', $msgstr );
            // Might be fuzzy infile
            $flags[] = 'fuzzy';
        }
 
        if ( GettextPlural::hasPlural( $msgid ) ) {
            $forms = GettextPlural::unflatten( $msgid, 2 );
            $content .= 'msgid ' . $this->escape( $forms[0] ) . "\n";
            $content .= 'msgid_plural ' . $this->escape( $forms[1] ) . "\n";
 
            try {
                $forms = GettextPlural::unflatten( $msgstr, $pluralCount );
                foreach ( $forms as $index => $form ) {
                    $content .= "msgstr[$index] " . $this->escape( $form ) . "\n";
                }
            } catch ( GettextPluralException $e ) {
                $flags[] = 'invalid-plural';
                for ( $i = 0; $i < $pluralCount; $i++ ) {
                    $content .= "msgstr[$i] \"\"\n";
                }
            }
        } else {
            $content .= 'msgid ' . $this->escape( $msgid ) . "\n";
            $content .= 'msgstr ' . $this->escape( $msgstr ) . "\n";
        }
 
        if ( $flags ) {
            sort( $flags );
            $header .= '#, ' . implode( ', ', array_unique( $flags ) ) . "\n";
        }
 
        $output = $header ?: "#\n";
        $output .= $content . "\n";
 
        return $output;
    }
 
    private function formatTime( string $time ): string {
        $lang = MediaWikiServices::getInstance()->getLanguageFactory()->getLanguage( 'en' );
 
        return $lang->sprintfDate( 'xnY-xnm-xnd xnH:xni:xns+0000', $time );
    }
 
    private function getPotTime(): string {
        $cache = $this->group->getMessageGroupCache( $this->group->getSourceLanguage() );
 
        return $cache->exists() ? $cache->getTimestamp() : wfTimestampNow();
    }
 
    private function formatDocumentation( ?string $documentation ): string {
        if ( !is_string( $documentation ) ) {
            return '';
        }
 
        if ( !$this->offlineMode ) {
            return '';
        }
 
        $lines = explode( "\n", $documentation );
        $out = '';
        foreach ( $lines as $line ) {
            $out .= "#. [Wiki] $line\n";
        }
 
        return $out;
    }
 
    private function escape( string $line ): string {
        // There may be \ as a last character, for keeping trailing whitespace
        $line = preg_replace( '/(\s)\\\\$/', '\1', $line );
        $line = addcslashes( $line, '\\"' );
        $line = str_replace( "\n", '\n', $line );
        return '"' . $line . '"';
    }
 
    public function shouldOverwrite( string $a, string $b ): bool {
        $regex = '/^"(.+)-Date: \d\d\d\d-\d\d-\d\d \d\d:\d\d:\d\d\+\d\d\d\d\\\\n"$/m';
 
        $a = preg_replace( $regex, '', $a );
        $b = preg_replace( $regex, '', $b );
 
        return $a !== $b;
    }
 
    public static function getExtraSchema(): array {
        return [
            'root' => [
                '_type' => 'array',
                '_children' => [
                    'FILES' => [
                        '_type' => 'array',
                        '_children' => [
                            'header' => [
                                '_type' => 'text',
                            ],
                            'keyAlgorithm' => [
                                '_type' => 'enum',
                                '_values' => [ 'simple', 'legacy' ],
                            ],
                            'CtxtAsKey' => [
                                '_type' => 'boolean',
                            ],
                        ]
                    ]
                ]
            ]
        ];
    }
 
    public function isContentEqual( ?string $a, ?string $b ): bool {
        if ( $a === $b ) {
            return true;
        }
 
        if ( $a === null || $b === null ) {
            return false;
        }
 
        try {
            $parsedA = GettextPlural::parsePluralForms( $a );
            $parsedB = GettextPlural::parsePluralForms( $b );
 
            // if they have the different number of plural forms, just fail
            if ( count( $parsedA[1] ) !== count( $parsedB[1] ) ) {
                return false;
            }
 
        } catch ( GettextPluralException $e ) {
            // Something failed, invalid syntax?
            return false;
        }
 
        $expectedPluralCount = count( $parsedA[1] );
 
        // GettextPlural::unflatten() will return an empty array when $expectedPluralCount is 0
        // So if they do not have translations and are different strings, they are not equal
        if ( $expectedPluralCount === 0 ) {
            return false;
        }
 
        return GettextPlural::unflatten( $a, $expectedPluralCount )
            === GettextPlural::unflatten( $b, $expectedPluralCount );
    }
}
 
class_alias( GettextFormat::class, 'GettextFFS' );