Add AMP_DOM_Document & meta tag sanitizer

includes/sanitizers/class-amp-script-sanitizer.php

@@ -50,7 +50,7 @@ public function sanitize() {
 			$fragment->appendChild( $this->dom->createComment( 'noscript' ) );
 			while ( $noscript->firstChild ) {
 				if ( $is_inside_head_el && ! $must_move_to_body ) {
-					$must_move_to_body = ! AMP_DOM_Utils::is_valid_head_node( $noscript->firstChild );
+					$must_move_to_body = ! $this->dom->is_valid_head_node( $noscript->firstChild );
 				}
 				$fragment->appendChild( $noscript->firstChild );
 			}

includes/sanitizers/class-amp-style-sanitizer.php

@@ -2795,7 +2795,7 @@ private function remove_admin_bar_if_css_excluded() {
 		if ( ! $included ) {
 			// Remove admin-bar class from body element.
 			// @todo It would be nice if any style rules which refer to .admin-bar could also be removed, but this would mean retroactively going back over the CSS again and re-shaking it.
-			if ( $this->dom->body instanceof DOMElement && $this->dom->body->hasAttribute( 'class' ) ) {
+			if ( $this->dom->body->hasAttribute( 'class' ) ) {
 				$this->dom->body->setAttribute(
 					'class',
 					preg_replace( '/(^|\s)admin-bar(\s|$)/', ' ', $this->dom->body->getAttribute( 'class' ) )

includes/utils/class-amp-dom-utils.php

@@ -47,39 +47,19 @@ class AMP_DOM_Utils {
 	 */
 	const HTML_BODY_CONTENTS_PATTERN = '#^.*?<body.*?>(.*)</body>.*?$#si';
 
-	/**
-	 * List of elements allowed in head.
-	 *
-	 * @link https://github.com/ampproject/amphtml/blob/445d6e3be8a5063e2738c6f90fdcd57f2b6208be/validator/engine/htmlparser.js#L83-L100
-	 * @link https://www.w3.org/TR/html5/document-metadata.html
-	 *
-	 * @todo Turn into const array once PHP minimum is bumped to 5.6+.
-	 *
-	 * @var array
-	 */
-	private static $elements_allowed_in_head = [
-		'title',
-		'base',
-		'link',
-		'meta',
-		'style',
-		'noscript',
-		'script',
-	];
-
 	/**
 	 * Return a valid Dom\Document representing HTML document passed as a parameter.
 	 *
 	 * @since 0.7
 	 * @see AMP_DOM_Utils::get_content_from_dom_node()
-	 * @deprecated Use Dom\Document::from_html( $html, $encoding ) instead.
+	 * @deprecated Use Amp\AmpWP\Dom\Document::from_html( $html, $encoding ) instead.
 	 *
 	 * @param string $document Valid HTML document to be represented by a Dom\Document.
 	 * @param string $encoding Optional. Encoding to use for the content.
 	 * @return Document|false Returns Dom\Document, or false if conversion failed.
 	 */
 	public static function get_dom( $document, $encoding = null ) {
-		_deprecated_function( __METHOD__, '1.5.0', 'Dom\Document::from_html()' );
+		_deprecated_function( __METHOD__, '1.5.0', 'Amp\AmpWP\Dom\Document::from_html()' );
 		return Document::from_html( $document, $encoding );
 	}
 
@@ -88,18 +68,14 @@ public static function get_dom( $document, $encoding = null ) {
 	 *
 	 * @link https://github.com/ampproject/amphtml/blob/445d6e3be8a5063e2738c6f90fdcd57f2b6208be/validator/engine/htmlparser.js#L83-L100
 	 * @link https://www.w3.org/TR/html5/document-metadata.html
+	 * @deprecated Use Amp\AmpWP\Dom\Document->is_valid_head_node() instead.
 	 *
 	 * @param DOMNode $node Node.
 	 * @return bool Whether valid head node.
 	 */
 	public static function is_valid_head_node( DOMNode $node ) {
-		return (
-			( $node instanceof DOMElement && in_array( $node->nodeName, self::$elements_allowed_in_head, true ) )
-			||
-			( $node instanceof DOMText && preg_match( '/^\s*$/', $node->nodeValue ) ) // Whitespace text nodes are OK.
-			||
-			$node instanceof DOMComment
-		);
+		_deprecated_function( __METHOD__, '1.5.0', 'Amp\AmpWP\Dom\Document->is_valid_head_node()' );
+		return Document::from_node( $node )->is_valid_head_node( $node );
 	}
 
 	/**
@@ -219,7 +195,7 @@ public static function get_content_from_dom( Document $dom ) {
 	 * @return string Returns the HTML content represented in the DOMNode
 	 */
 	public static function get_content_from_dom_node( Document $dom, $node ) {
-		_deprecated_function( __METHOD__, '1.5.0', 'Dom\Document->saveHtml( $node )' );
+		_deprecated_function( __METHOD__, '1.5.0', 'Amp\AmpWP\Dom\Document::saveHtml()' );
 		return $dom->saveHTML( $node );
 	}
 

src/Dom/Document.php

@@ -7,10 +7,11 @@
 
 namespace Amp\AmpWP\Dom;
 
-use AMP_DOM_Utils;
+use DOMComment;
 use DOMDocument;
 use DOMElement;
 use DOMNode;
+use DOMText;
 use DOMXPath;
 
 /**
@@ -119,7 +120,6 @@ final class Document extends DOMDocument {
 	const HTML_FIND_TAG_WITH_ATTRIBUTE_PATTERN    = '/<%1$s [^>]*?\s*%2$s=[^>]*?>[^<]*(?:<\/%1$s>)?/i';
 	const HTML_EXTRACT_ATTRIBUTE_VALUE_PATTERN    = '/%s=(?:([\'"])(?<full>.*)?\1|(?<partial>[^ \'";]+))/';
 
-
 	// Tags constants used throughout.
 	const TAG_HEAD     = 'head';
 	const TAG_BODY     = 'body';
@@ -181,6 +181,26 @@ final class Document extends DOMDocument {
 		'wbr',
 	];
 
+	/**
+	 * List of elements allowed in head.
+	 *
+	 * @link https://github.com/ampproject/amphtml/blob/445d6e3be8a5063e2738c6f90fdcd57f2b6208be/validator/engine/htmlparser.js#L83-L100
+	 * @link https://www.w3.org/TR/html5/document-metadata.html
+	 *
+	 * @todo Turn into const array once PHP minimum is bumped to 5.6+.
+	 *
+	 * @var string[]
+	 */
+	private static $elements_allowed_in_head = [
+		'title',
+		'base',
+		'link',
+		'meta',
+		'style',
+		'noscript',
+		'script',
+	];
+
 	/**
 	 * Store the placeholder comments that were generated to replace <noscript> elements.
 	 *
@@ -258,6 +278,22 @@ public static function from_node( DOMNode &$node ) {
 		return $dom;
 	}
 
+	/**
+	 * Reset the internal optimizations of the Document object.
+	 *
+	 * This might be needed if you are doing an operation that causes the cached
+	 * nodes and XPath objects to point to the wrong document.
+	 *
+	 * @return self Reset version of the Document object.
+	 */
+	public function reset() {
+		// Drop references to old DOM document.
+		unset( $this->xpath, $this->head, $this->body );
+
+		// Reference of the document itself doesn't change here, but might need to change in the future.
+		return $this;
+	}
+
 	/**
 	 * Load HTML from a string.
 	 *
@@ -268,15 +304,14 @@ public static function from_node( DOMNode &$node ) {
 	 * @return bool true on success or false on failure.
 	 */
 	public function loadHTML( $source, $options = 0 ) {
-		// Drop references to old DOM document.
-		unset( $this->xpath, $this->head, $this->body );
+		$this->reset();
 
 		$source = $this->convert_amp_bind_attributes( $source );
 		$source = $this->replace_self_closing_tags( $source );
 		$source = $this->normalize_document_structure( $source );
 		$source = $this->maybe_replace_noscript_elements( $source );
 
-		$this->original_encoding = $this->detect_and_strip_encoding( $source );
+		list( $source, $this->original_encoding ) = $this->detect_and_strip_encoding( $source );
 
 		if ( self::AMP_ENCODING !== strtolower( $this->original_encoding ) ) {
 			$source = $this->adapt_encoding( $source );
@@ -306,7 +341,8 @@ public function loadHTML( $source, $options = 0 ) {
 		}
 
 		// Add the required utf-8 meta charset tag.
-		$charset = AMP_DOM_Utils::create_node( $this, 'meta', [ 'charset' => self::AMP_ENCODING ] );
+		$charset = $this->createElement( 'meta' );
+		$charset->setAttribute( 'charset', self::AMP_ENCODING );
 		$this->head->insertBefore( $charset, $this->head->firstChild );
 
 		// Do some further clean-up.
@@ -328,15 +364,9 @@ public function saveHTML( DOMNode $node = null ) {
 
 		// Force-add http-equiv charset to make DOMDocument behave as it should.
 		// See: http://php.net/manual/en/domdocument.loadhtml.php#78243.
-		$charset = AMP_DOM_Utils::create_node(
-			$this,
-			'meta',
-			[
-				'http-equiv' => self::HTML_HTTP_EQUIV_VALUE,
-				'content'    => self::HTML_HTTP_EQUIV_CONTENT_VALUE,
-			]
-		);
-
+		$charset = $this->createElement( 'meta' );
+		$charset->setAttribute( 'http-equiv', self::HTML_HTTP_EQUIV_VALUE );
+		$charset->setAttribute( 'content', self::HTML_HTTP_EQUIV_CONTENT_VALUE );
 		$this->head->insertBefore( $charset, $this->head->firstChild );
 
 		if ( null === $node || PHP_VERSION_ID >= 70300 ) {
@@ -479,7 +509,7 @@ private function move_invalid_head_nodes_to_body() {
 		$node = $this->head->lastChild;
 		while ( $node ) {
 			$next_sibling = $node->previousSibling;
-			if ( ! AMP_DOM_Utils::is_valid_head_node( $node ) ) {
+			if ( ! $this->is_valid_head_node( $node ) ) {
 				$this->body->insertBefore( $this->head->removeChild( $node ), $this->body->firstChild );
 			}
 			$node = $next_sibling;
@@ -661,7 +691,7 @@ private function convert_amp_bind_attributes( $html ) {
 			$html
 		);
 
-		/**
+		/*
 		 * If the regex engine incurred an error during processing, for example exceeding the backtrack
 		 * limit, $converted will be null. In this case we return the originally passed document to allow
 		 * DOMDocument to attempt to load it.  If the AMP HTML doesn't make use of amp-bind or similar
@@ -706,9 +736,14 @@ private function adapt_encoding( $source ) {
 	 * Detect the encoding of the document.
 	 *
 	 * @param string $content Content of which to detect the encoding.
-	 * @return string|false Detected encoding of the document, or false if none.
+	 * @return array {
+	 *     Detected encoding of the document, or false if none.
+	 *
+	 *     @type string $content Potentially modified content.
+	 *     @type string|false $encoding Encoding of the content, or false if not detected.
+	 * }
 	 */
-	private function detect_and_strip_encoding( &$content ) {
+	private function detect_and_strip_encoding( $content ) {
 		$encoding = self::UNKNOWN_ENCODING;
 
 		// Check for HTML 4 http-equiv meta tags.
@@ -734,7 +769,7 @@ private function detect_and_strip_encoding( &$content ) {
 			$content = str_replace( $charset_tag, '', $content );
 		}
 
-		return $encoding;
+		return [ $content, $encoding ];
 	}
 
 	/**
@@ -828,11 +863,13 @@ private function sanitize_encoding( $encoding ) {
 			$known_encodings = array_map( 'strtolower', mb_list_encodings() );
 		}
 
-		if ( array_key_exists( strtolower( $encoding ), self::$encoding_map ) ) {
-			$encoding = self::$encoding_map[ strtolower( $encoding ) ];
+		$lc_encoding = strtolower( $encoding );
+
+		if ( isset( self::$encoding_map[ $lc_encoding ] ) ) {
+			$encoding = self::$encoding_map[ $lc_encoding ];
 		}
 
-		if ( ! in_array( strtolower( $encoding ), $known_encodings, true ) ) {
+		if ( ! in_array( $lc_encoding, $known_encodings, true ) ) {
 			return self::UNKNOWN_ENCODING;
 		}
 
@@ -932,6 +969,25 @@ private function get_mustache_tag_placeholders() {
 		return $placeholders;
 	}
 
+	/**
+	 * Determine whether a node can be in the head.
+	 *
+	 * @link https://github.com/ampproject/amphtml/blob/445d6e3be8a5063e2738c6f90fdcd57f2b6208be/validator/engine/htmlparser.js#L83-L100
+	 * @link https://www.w3.org/TR/html5/document-metadata.html
+	 *
+	 * @param DOMNode $node Node.
+	 * @return bool Whether valid head node.
+	 */
+	public function is_valid_head_node( DOMNode $node ) {
+		return (
+			( $node instanceof DOMElement && in_array( $node->nodeName, self::$elements_allowed_in_head, true ) )
+			||
+			( $node instanceof DOMText && preg_match( '/^\s*$/', $node->nodeValue ) ) // Whitespace text nodes are OK.
+			||
+			$node instanceof DOMComment
+		);
+	}
+
 	/**
 	 * Magic getter to implement lazily-created, cached properties for the document.
 	 *
@@ -965,4 +1021,13 @@ public function __get( $name ) {
 		trigger_error( self::PROPERTY_GETTER_ERROR_MESSAGE . $name, E_NOTICE ); // phpcs:ignore WordPress.PHP.DevelopmentFunctions,WordPress.Security.EscapeOutput
 		return null;
 	}
+
+	/**
+	 * Make sure we properly reinitialize on clone.
+	 *
+	 * @return void
+	 */
+	public function __clone() {
+		$this->reset();
+	}
 }

tests/php/test-amp-style-sanitizer.php

@@ -2406,7 +2406,7 @@ public function test_prioritized_stylesheets( $html_generator, $assert ) {
 		$html = $html_generator();
 
 		$original_dom = Document::from_html( $html );
-		$amphtml_dom  = Document::from_html( $html );
+		$amphtml_dom  = clone( $original_dom );
 
 		$error_codes = [];
 		$args        = [