Use RemexHtml to properly remove <img> tags (without regexes)

edwardspec · Dec 30, 2024 · 2650b5a · 2650b5a
1 parent a648959
commit 2650b5a
Show file tree

Hide file tree

Showing 3 changed files with 51 additions and 23 deletions.
diff --git a/includes/EventCalendar.php b/includes/EventCalendar.php
@@ -230,15 +230,16 @@ public function findEvents( array $opt, Parser $recursiveParser ) {
 				if ( !$snippet ) {
 					// Full text of the page (no more than N first symbols) was requested.
 					// NOTE: we can't use getParserOutput() here, because we are already inside Parser::parse().
-					$parsedHtml = $recursiveParser->recursiveTagParseFully( $row->text );
+					$snippet = $recursiveParser->recursiveTagParseFully( $row->text );
 
-					// Remove the image tags: in 99,9% of cases they are too wide to be included into the calendar.
-					// TODO: properly remove <div class="thumb"> with all contents (currently hidden by CSS).
-					$parsedHtml = preg_replace( '/<img[^>]+>/', '', $parsedHtml );
-					$snippet = mb_substr( $parsedHtml, 0, $maxSymbols );
+					// Remove unwanted HTML tags (if any), such as images (too wide for the calendar).
+					$snippet = HtmlSanitizer::sanitizeSnippet( $snippet );
+
+					// Truncate to maximum allowed length.
+					$snippet = mb_substr( $snippet, 0, $maxSymbols );
 
 					// Remove truncated HTML tags (if any).
-					$snippet = HtmlSanitizer::sanitizeHTML( $snippet );
+					$snippet = HtmlSanitizer::sanitizeSnippet( $snippet );
 
 					// Store the snippet in cache.
 					// NOTE: the reason why we don't use $dbCache->set() here is that SqlBagOStuff::set() will do

diff --git a/includes/HtmlSanitizer.php b/includes/HtmlSanitizer.php
@@ -25,18 +25,58 @@
 
 use Wikimedia\RemexHtml\Serializer\HtmlFormatter;
 use Wikimedia\RemexHtml\Serializer\Serializer;
+use Wikimedia\RemexHtml\Serializer\SerializerNode;
 use Wikimedia\RemexHtml\Tokenizer\Tokenizer;
 use Wikimedia\RemexHtml\TreeBuilder\Dispatcher;
 use Wikimedia\RemexHtml\TreeBuilder\TreeBuilder;
 
 class HtmlSanitizer {
 	/**
-	 * Remove invalid/non-matching/truncated HTML tags and return correct (sanitized) HTML.
+	 * Remove unwanted/invalid/non-matching/truncated HTML tags and return correct (sanitized) HTML.
 	 * @param string $html
 	 * @return string
 	 */
-	public static function sanitizeHTML( $html ) {
-		$formatter = new HtmlFormatter;
+	public static function sanitizeSnippet( $html ) {
+		$formatter = new class () extends HtmlFormatter {
+			public function startDocument( $fragmentNamespace, $fragmentName ) {
+				// Remove DOCTYPE.
+				return '';
+			}
+
+			public function element( SerializerNode $parent, SerializerNode $node, $contents ) {
+				switch ( $node->name ) {
+					// Remove everything outside the <body> tag.
+					case 'head':
+						return '';
+
+					case 'html':
+					case 'body':
+						return $contents;
+
+					case 'img':
+						// Remove the image tags: in 99,9% of cases they are too wide
+						// to be included into the calendar.
+						// Not needed in MediaWiki 1.40+ (already removed with the <span> below).
+						return '';
+
+					case 'span':
+						if ( $node->attrs['typeof'] ?? '' === 'mw:File' ) {
+							// Wrapper around the image.
+							return '';
+						}
+						break;
+
+					// TODO: properly remove <div class="thumb"> with all contents (currently hidden by CSS).
+
+					case 'p':
+						// Remove trailing newline inside <p> tags.
+						$contents = trim( $contents );
+				}
+
+				return parent::element( $parent, $node, $contents );
+			}
+		};
+
 		$serializer = new Serializer( $formatter );
 		$treeBuilder = new TreeBuilder( $serializer );
 		$dispatcher = new Dispatcher( $treeBuilder );
@@ -45,13 +85,6 @@ public static function sanitizeHTML( $html ) {
 		$tokenizer->execute();
 		$html = $serializer->getResult();
 
-		// Remove doctype, <head>, etc.: everything outside the <body> tag.
-		// TODO: this can probably be implemented by subclassing HtmlFormatter class.
-		$html = preg_replace( '@^.*<body>(.*)</body>.*$@s', '$1', $html );
-
-		// Remove trailing newline inside <p></p> tags.
-		$html = preg_replace( "@\n</p>@", '</p>', $html );
-
 		return $html;
 	}
 }
diff --git a/tests/phpunit/EventCalendarTest.php b/tests/phpunit/EventCalendarTest.php
@@ -778,13 +778,7 @@ public function testSnippetForCompressedRevision() {
 	public function testSnippetSanitizer() {
 		$filename = 'Testimage.png';
 		$pageText = "Expected snippet [[File:$filename]]";
-
-		$isMW39 = version_compare( MW_VERSION, '1.40.0', '<' );
-		$expectedSnippet = '<p>Expected snippet ' .
-			( $isMW39 ? '' : '<span class="mw-default-size" typeof="mw:File">' ) .
-			"<a href=\"/wiki/File:$filename\" class=\"" .
-			( $isMW39 ? 'image' : 'mw-file-description' ) .
-			'"></a>' . ( $isMW39 ? '' : '</span>' ) . '</p>';
+		$expectedSnippet = '<p>Expected snippet</p>';
 
 		// Upload a test file, so that [[File:]] syntax would create an actual thumbnail, not a redlink.
 		$user = $this->getTestUser()->getUser();