wikimedia/mediawiki-extensions-Translate

<?php
declare( strict_types = 1 );

namespace MediaWiki\Extension\Translate\WebService;

use FormatJson;

/**
 * Implements support Caighdean translator api.
 * @author Niklas Laxström
 * @license GPL-2.0-or-later
 * @ingroup TranslationWebService
 * @since 2017.04
 * @see https://github.com/kscanne/caighdean/blob/master/API.md
 */
class CaighdeanWebService extends TranslationWebService {
    /** @inheritDoc */
    public function getType(): string {
        return 'mt';
    }

    /** @inheritDoc */
    protected function mapCode( string $code ): string {
        return $code;
    }

    /** @inheritDoc */
    protected function doPairs(): array {
        $pairs = [
            'gd' => [ 'ga' => true ],
            'gv' => [ 'ga' => true ],
        ];

        return $pairs;
    }

    /** @inheritDoc */
    protected function getQuery( string $text, string $sourceLanguage, string $targetLanguage ): TranslationQuery {
        if ( !isset( $this->config['url'] ) ) {
            throw new TranslationWebServiceConfigurationException( '`url` not set in configuration' );
        }

        $text = trim( $text );
        if ( $text === '' ) {
            throw new TranslationWebServiceInvalidInputException( 'Input is empty' );
        }

        $data = wfArrayToCgi( [
            'foinse' => $sourceLanguage,
            'teacs' => $text,
        ] );

        // Maximum payload is 16 KiB. Based ont testing 16000 bytes is safe by leaving 224
        // bytes for other things.
        if ( strlen( $data ) > 16000 ) {
            throw new TranslationWebServiceInvalidInputException( 'Input is over 16000 bytes long' );
        }

        return TranslationQuery::factory( $this->config['url'] )
            ->timeout( intval( $this->config['timeout'] ) )
            ->postWithData( $data )
            ->attachProcessingInstructions( $text );
    }

    /** @inheritDoc */
    protected function parseResponse( TranslationQueryResponse $reply ): string {
        $body = $reply->getBody();
        $response = FormatJson::decode( $body );
        if ( !is_array( $response ) ) {
            throw new TranslationWebServiceException( 'Invalid json: ' . serialize( $body ) );
        }

        $text = '';
        $originalText = $reply->getQuery()->getProcessingInstructions();
        foreach ( $response as [ $sourceToken, $targetToken ] ) {
            $separator = ' ';
            $pos = strpos( $originalText, $sourceToken );
            // Try to keep the effects local. If we fail to match at token, we could accidentally
            // scan very far ahead in the text, find a false match and not find matches for all
            // of the tokens in the between.
            if ( $pos !== false && $pos < 50 ) {
                // Remove the portion of text we have processed. $pos should be zero, unless
                // we failed to match something earlier.
                $originalText = substr( $originalText, $pos + strlen( $sourceToken ) );
                if ( preg_match( '/^\s+/', $originalText, $match ) ) {
                    $separator = $match[ 0 ];
                    $originalText = substr( $originalText, strlen( $separator ) );
                } else {
                    $separator = '';
                }
            }

            $text .= $targetToken . $separator;
        }

        return $text;
    }
}