TranslatablePageParser.php

<?php
declare( strict_types = 1 );
 
namespace MediaWiki\Extension\Translate\PageTranslation;
 
use MediaWiki\Extension\Translate\Utilities\ParsingPlaceholderFactory;
use Wikimedia\Message\MessageValue;
 
class TranslatablePageParser {
    private ParsingPlaceholderFactory $placeholderFactory;
 
    public function __construct( ParsingPlaceholderFactory $placeholderFactory ) {
        $this->placeholderFactory = $placeholderFactory;
    }
 
    public function containsMarkup( string $text ): bool {
        $nowiki = [];
        $text = $this->armourNowiki( $nowiki, $text );
        return preg_match( '~</?translate[ >]~', $text ) !== 0;
    }
 
    public function cleanupTags( string $text ): string {
        $nowiki = [];
        $text = $this->armourNowiki( $nowiki, $text );
        $text = preg_replace( '~<translate( nowrap)?>\n?~s', '', $text );
        $text = preg_replace( '~\n?</translate>~s', '', $text );
        // Markers: headers and the rest
        $ic = preg_quote( TranslationUnit::UNIT_MARKER_INVALID_CHARS, '~' );
        $text = preg_replace( "~(^=.*=) <!--T:[^$ic]+-->$~um", '\1', $text );
        $text = preg_replace( "~<!--T:[^$ic]+-->[\n ]?~um", '', $text );
        // Remove variables
        $unit = new TranslationUnit( $text );
        $text = $unit->getTextForTrans();
 
        return $this->unarmourNowiki( $nowiki, $text );
    }
 
    public function parse( string $text ): ParserOutput {
        $nowiki = [];
        $text = $this->armourNowiki( $nowiki, $text );
 
        $sections = [];
        $tagPlaceHolders = [];
 
        while ( true ) {
            $re = '~(<translate(?: nowrap)?>)(.*?)</translate>~s';
            $matches = [];
            $ok = preg_match( $re, $text, $matches, PREG_OFFSET_CAPTURE );
 
            if ( $ok === 0 || $ok === false ) {
                break; // No match or failure
            }
 
            $contentWithTags = $matches[0][0];
            $contentWithoutTags = $matches[2][0];
            // These are offsets to the content inside the tags in $text
            $offsetStart = $matches[0][1];
            $offsetEnd = $offsetStart + strlen( $contentWithTags );
 
            // Replace the whole match with a placeholder
            $ph = $this->placeholderFactory->make();
            $text = substr( $text, 0, $offsetStart ) . $ph . substr( $text, $offsetEnd );
 
            if ( preg_match( '~<translate( nowrap)?>~', $contentWithoutTags ) !== 0 ) {
                throw new ParsingFailure(
                    'Nested tags',
                    new MessageValue( 'pt-parse-nested', [ wfEscapeWikiText( $contentWithoutTags ) ] )
                );
            }
 
            $openTag = $matches[1][0];
            $canWrap = $openTag !== '<translate nowrap>';
 
            // Parse the content inside the tags
            $contentWithoutTags = $this->unarmourNowiki( $nowiki, $contentWithoutTags );
            $parse = $this->parseSection( $contentWithoutTags, $canWrap );
 
            // Update list of sections and the template with the results
            $sections += $parse['sections'];
            $tagPlaceHolders[$ph] = new Section( $openTag, $parse['template'], '</translate>' );
        }
 
        $prettyTemplate = $text;
        foreach ( $tagPlaceHolders as $ph => $value ) {
            $prettyTemplate = str_replace( $ph, '[...]', $prettyTemplate );
        }
 
        if ( preg_match( '~<translate( nowrap)?>~', $text ) !== 0 ) {
            throw new ParsingFailure(
                'Unmatched opening tag',
                // TODO this and the below one should use ->plaintextParams() instead of wfEscapeWikiText,
                // but MediaWiki\EditPage\Constraint\EditFilterMergedContentHookConstraint::formatStatusErrors()
                // calls ->plain() instead of ->parse(), at which point the meaning of plain text params gets lost
                new MessageValue( 'pt-parse-open', [ wfEscapeWikiText( $prettyTemplate ) ] )
            );
        } elseif ( str_contains( $text, '</translate>' ) ) {
            throw new ParsingFailure(
                "Unmatched closing tag",
                new MessageValue( 'pt-parse-close', [ wfEscapeWikiText( $prettyTemplate ) ] )
            );
        }
 
        $text = $this->unarmourNowiki( $nowiki, $text );
 
        return new ParserOutput( $text, $tagPlaceHolders, $sections );
    }
 
    public function parseSection( string $text, bool $canWrap ): array {
        $flags = PREG_SPLIT_NO_EMPTY | PREG_SPLIT_DELIM_CAPTURE;
        $parts = preg_split( '~(^\s*|\s*\n\n\s*|\s*$)~', $text, -1, $flags );
 
        $inline = preg_match( '~\n~', $text ) === 0;
 
        $template = '';
        $sections = [];
 
        foreach ( $parts as $_ ) {
            if ( trim( $_ ) === '' ) {
                $template .= $_;
            } else {
                $ph = $this->placeholderFactory->make();
                $tpSection = $this->parseUnit( $_ );
                $tpSection->setIsInline( $inline );
                $tpSection->setCanWrap( $canWrap );
                $sections[$ph] = $tpSection;
                $template .= $ph;
            }
        }
 
        return [
            'template' => $template,
            'sections' => $sections,
        ];
    }
 
    public function parseUnit( string $content ): TranslationUnit {
        $re = '~<!--T:(.*?)-->~';
        $matches = [];
        $count = preg_match_all( $re, $content, $matches, PREG_SET_ORDER );
 
        if ( $count > 1 ) {
            throw new ParsingFailure(
                'Multiple translation unit markers',
                new MessageValue( 'pt-shake-multiple', [ wfEscapeWikiText( $content ) ] )
            );
        }
 
        // If no id given in the source, default to a new section id
        $id = TranslationUnit::NEW_UNIT_ID;
        if ( $count === 1 ) {
            foreach ( $matches as $match ) {
                [ /*full*/, $id ] = $match;
 
                // Currently handle only these two standard places.
                // Is this too strict?
                $rer1 = '~^<!--T:(.*?)-->( |\n)~'; // Normal sections
                $rer2 = '~\s*<!--T:(.*?)-->$~m'; // Sections with title
                $content = preg_replace( $rer1, '', $content );
                $content = preg_replace( $rer2, '', $content );
 
                if ( preg_match( $re, $content ) === 1 ) {
                    throw new ParsingFailure(
                        'Translation unit marker is in unsupported position',
                        new MessageValue( 'pt-shake-position', [ wfEscapeWikiText( $content ) ] )
                    );
                } elseif ( trim( $content ) === '' ) {
                    throw new ParsingFailure(
                        'Translation unit has no content besides marker',
                        new MessageValue( 'pt-shake-empty', [ wfEscapeWikiText( $id ) ] )
                    );
                }
            }
        }
 
        return new TranslationUnit( $content, $id );
    }
 
    public function armourNowiki( array &$holders, string $text ): string {
        $re = '~(<nowiki>)(.*?)(</nowiki>)~s';
 
        while ( preg_match( $re, $text, $matches ) ) {
            $ph = $this->placeholderFactory->make();
            $text = str_replace( $matches[0], $ph, $text );
            $holders[$ph] = $matches[0];
        }
 
        return $text;
    }
 
    public function unarmourNowiki( array $holders, string $text ): string {
        return strtr( $text, $holders );
    }
}