ParsoidMediaWikiParser.php

<?php

namespace PortableInfobox\Parsoid;

use MediaWiki\Title\Title;
use PortableInfobox\Services\Parser\ExternalParser;
use Wikimedia\Parsoid\DOM\Element;
use Wikimedia\Parsoid\Ext\ParsoidExtensionAPI;
use Wikimedia\Parsoid\Utils\DOMCompat;

class ParsoidMediaWikiParser implements ExternalParser {

	public function __construct(
		private readonly ParsoidExtensionAPI $api,
	) {
	}

	public function parseRecursive( $wikitext ) {
		if ( $wikitext === null ) {
			return null;
		}

		$paramParsed = $this->api->wikitextToDOM( $wikitext, [
			// this differs from earlier as we need the frame to be able to grab the
			// params the user passed - parsoid handles this internally it appears
			'processInNewFrame' => false,
			'parseOpts' => [ 'context' => 'inline' ],
		], true );
		'@phan-var Element $paramParsed';

		// we don't want Parsoid to wrap in a span or add a typeof here,
		// just interested in the content
		return DOMCompat::getOuterHTML( $paramParsed );
	}

	/** @param $wikitext @phan-unused-param */
	public function replaceVariables( $wikitext ) {
		// no-op - I think handled by ->wikiTextToDOM?
	}

	/**
	 * @param $title @phan-unused-param
	 * @param array $sizeParams @phan-unused-param
	 */
	public function addImage( $title, array $sizeParams ): ?string {
		// no-op at present. Used to extend the image tag with specific information for the PageImages extension.
		// that extension relies on Parser hooks and therefore is not safe to assume that will work indefinitely.
		// could potentially just do it for now whilst the hooks still exist, and maybe remove at a later date if
		// PageImages is not made Parsoid-compat.
		return '';
	}

	public function getParsoidExtensionApi(): ParsoidExtensionAPI {
		return $this->api;
	}

	/**
	 * Extract the gallery and return the filename -> captions. PortableInfobox currently does this
	 * a lot cleaner as it piggybacks on onAfterParserFetchFileAndTitle hook to set the images into the data bag.
	 * This hook is NOT available on Parsoid, and we have no other way to get the resultant class which PortableInfobox
	 * currently relies on. So we need to fake it as best we can and hope WMF comes up with something later down
	 * the line.
	 * @param string $wikitext
	 * @return array an array of the images
	 */
	public function extractGallery( string $wikitext ): array {
		// the legacy implementation reuturns an array where each element is an array of the caption
		// and the title object for that specific image. We don't have access to this by default,
		// since there is no concept of half parsing in Parsoid - we either ask Parsoid for the Parsed wt->html
		// or we work with the WT and grab what we need.
		$result = [];

		// this is quicker than passing the wikitext to Parsoid and extracting
		// the images etc from it.
		if ( preg_match( '/<gallery[^>]*>(.*?)<\/gallery>/s', $wikitext, $matches ) ) {
			$galleryContent = trim( $matches[1] );
			if ( !$galleryContent ) {
				return [];
			}

			$lines = explode( "\n", $galleryContent );
			foreach ( $lines as $line ) {
				$line = trim( $line );
				if ( !$line ) {
					continue;
				}

				$parts = explode( '|', $line, 2 );
				$filename = trim( $parts[0] );
				$caption = trim( $parts[1] ?? '' );
				if ( !$filename ) {
					continue;
				}

				$title = Title::newFromText( $filename, NS_FILE );
				if ( $title !== null ) {
					$result[] = [
						'label' => $caption,
						'title' => $title,
					];
				}
			}
		}

		return $result;
	}
}