HTML API: Defer applying attribute updates until necessary.

When making repeated updates to a document, the Tag Processor will end
up copying the entire document once for every update. This can lead to
catastrophic behavior in the worse case.

However, when batch-applying updates it's able to copy chunks of the
document in one thread and only end up copying the entire document once
for the entire batch.

Previously the Tag Processor has been eagerly applying udpates, but in
this patch it defers applying those updates as long as is possible.

Developed in #6120
Discussed in https://core.trac.wordpress.org/ticket/60697

Props: dmsnell, bernhard-reiter, jonsurrell, westonruter.
Fixes #60697.
Follow-up to [55706], [56941], [57348].



git-svn-id: https://develop.svn.wordpress.org/trunk@57805 602fd350-edb4-49c9-b593-d223f7449a82

Author: dmsnell

src/wp-includes/html-api/class-wp-html-tag-processor.php

@@ -837,8 +837,27 @@ public function next_tag( $query = null ) {
 	 * @return bool Whether a token was parsed.
 	 */
 	public function next_token() {
+		return $this->base_class_next_token();
+	}
+
+	/**
+	 * Internal method which finds the next token in the HTML document.
+	 *
+	 * This method is a protected internal function which implements the logic for
+	 * finding the next token in a document. It exists so that the parser can update
+	 * its state without affecting the location of the cursor in the document and
+	 * without triggering subclass methods for things like `next_token()`, e.g. when
+	 * applying patches before searching for the next token.
+	 *
+	 * @since 6.5.0
+	 *
+	 * @access private
+	 *
+	 * @return bool Whether a token was parsed.
+	 */
+	private function base_class_next_token() {
 		$was_at = $this->bytes_already_parsed;
-		$this->get_updated_html();
+		$this->after_tag();
 
 		// Don't proceed if there's nothing more to scan.
 		if (
@@ -2041,6 +2060,45 @@ private function skip_whitespace() {
 	 * @since 6.2.0
 	 */
 	private function after_tag() {
+		/*
+		 * There could be lexical updates enqueued for an attribute that
+		 * also exists on the next tag. In order to avoid conflating the
+		 * attributes across the two tags, lexical updates with names
+		 * need to be flushed to raw lexical updates.
+		 */
+		$this->class_name_updates_to_attributes_updates();
+
+		/*
+		 * Purge updates if there are too many. The actual count isn't
+		 * scientific, but a few values from 100 to a few thousand were
+		 * tests to find a practially-useful limit.
+		 *
+		 * If the update queue grows too big, then the Tag Processor
+		 * will spend more time iterating through them and lose the
+		 * efficiency gains of deferring applying them.
+		 */
+		if ( 1000 < count( $this->lexical_updates ) ) {
+			$this->get_updated_html();
+		}
+
+		foreach ( $this->lexical_updates as $name => $update ) {
+			/*
+			 * Any updates appearing after the cursor should be applied
+			 * before proceeding, otherwise they may be overlooked.
+			 */
+			if ( $update->start >= $this->bytes_already_parsed ) {
+				$this->get_updated_html();
+				break;
+			}
+
+			if ( is_int( $name ) ) {
+				continue;
+			}
+
+			$this->lexical_updates[] = $update;
+			unset( $this->lexical_updates[ $name ] );
+		}
+
 		$this->token_starts_at      = null;
 		$this->token_length         = null;
 		$this->tag_name_starts_at   = null;
@@ -2230,7 +2288,7 @@ private function apply_attributes_updates( $shift_this_point = 0 ) {
 			$shift = strlen( $diff->text ) - $diff->length;
 
 			// Adjust the cursor position by however much an update affects it.
-			if ( $diff->start <= $this->bytes_already_parsed ) {
+			if ( $diff->start < $this->bytes_already_parsed ) {
 				$this->bytes_already_parsed += $shift;
 			}
 
@@ -3164,15 +3222,7 @@ public function get_updated_html() {
 		 *                 └←─┘ back up by strlen("em") + 1 ==> 3
 		 */
 		$this->bytes_already_parsed = $before_current_tag;
-		$this->parse_next_tag();
-		// Reparse the attributes.
-		while ( $this->parse_next_attribute() ) {
-			continue;
-		}
-
-		$tag_ends_at                = strpos( $this->html, '>', $this->bytes_already_parsed );
-		$this->token_length         = $tag_ends_at - $this->token_starts_at;
-		$this->bytes_already_parsed = $tag_ends_at;
+		$this->base_class_next_token();
 
 		return $this->html;
 	}

tests/phpunit/tests/html-api/wpHtmlTagProcessor-bookmark.php

@@ -293,21 +293,30 @@ public function test_bookmarks_complex_use_case() {
 
 	/**
 	 * @ticket 56299
+	 * @ticket 60697
 	 *
 	 * @covers WP_HTML_Tag_Processor::seek
 	 */
 	public function test_updates_bookmark_for_additions_after_both_sides() {
 		$processor = new WP_HTML_Tag_Processor( '<div>First</div><div>Second</div>' );
 		$processor->next_tag();
+		$processor->set_attribute( 'id', 'one' );
 		$processor->set_bookmark( 'first' );
 		$processor->next_tag();
+		$processor->set_attribute( 'id', 'two' );
 		$processor->add_class( 'second' );
 
 		$processor->seek( 'first' );
 		$processor->add_class( 'first' );
 
 		$this->assertSame(
-			'<div class="first">First</div><div class="second">Second</div>',
+			'one',
+			$processor->get_attribute( 'id' ),
+			'Should have remembered attribute change from before the seek.'
+		);
+
+		$this->assertSame(
+			'<div class="first" id="one">First</div><div class="second" id="two">Second</div>',
 			$processor->get_updated_html(),
 			'The bookmark was updated incorrectly in response to HTML markup updates'
 		);

tests/phpunit/tests/html-api/wpHtmlTagProcessor.php

@@ -2727,4 +2727,49 @@ public function test_single_text_node_with_taglike_text() {
 		$this->assertSame( '#text', $processor->get_token_type(), 'Did not find text node.' );
 		$this->assertSame( 'test< /A>', $processor->get_modifiable_text(), 'Did not find complete text node.' );
 	}
+
+	/**
+	 * Ensures that updates which are enqueued in front of the cursor
+	 * are applied before moving forward in the document.
+	 *
+	 * @ticket 60697
+	 */
+	public function test_applies_updates_before_proceeding() {
+		$html = '<div><img></div><div><img></div>';
+
+		$subclass = new class( $html ) extends WP_HTML_Tag_Processor {
+			/**
+			 * Inserts raw text after the current token.
+			 *
+			 * @param string $new_html Raw text to insert.
+			 */
+			public function insert_after( $new_html ) {
+				$this->set_bookmark( 'here' );
+				$this->lexical_updates[] = new WP_HTML_Text_Replacement(
+					$this->bookmarks['here']->start + $this->bookmarks['here']->length + 1,
+					0,
+					$new_html
+				);
+			}
+		};
+
+		$subclass->next_tag( 'img' );
+		$subclass->insert_after( '<p>snow-capped</p>' );
+
+		$subclass->next_tag();
+		$this->assertSame(
+			'P',
+			$subclass->get_tag(),
+			'Should have matched inserted HTML as next tag.'
+		);
+
+		$subclass->next_tag( 'img' );
+		$subclass->set_attribute( 'alt', 'mountain' );
+
+		$this->assertSame(
+			'<div><img><p>snow-capped</p></div><div><img alt="mountain"></div>',
+			$subclass->get_updated_html(),
+			'Should have properly applied the update from in front of the cursor.'
+		);
+	}
 }