Parser: Move Sanitizer::normalizeCharReferences into RemexCompatFormatter

Choosing a particular encoding of HTML entities is logically a task
of the Remex formatter (which serializes HTML).  Move it out of the
Parser so that it is part of the serialization specification.

This is a follow up to Ic8965e81882d7cf024bdced437f684064a30ac86.

Change-Id: If45907baf24d60987b39cd1f7709c5f7caf19f37

includes/parser/Parser.php

@@ -1682,8 +1682,6 @@ class Parser {
 
 		$text = $this->mStripState->unstripGeneral( $text );
 
-		$text = Sanitizer::normalizeCharReferences( $text );
-
 		$text = $this->remexDriver->tidy( $text, [ Sanitizer::class, 'armorFrenchSpaces' ] );
 
 		if ( $isMain ) {

includes/tidy/RemexCompatFormatter.php

@@ -5,6 +5,7 @@ namespace MediaWiki\Tidy;
 use RemexHtml\HTMLData;
 use RemexHtml\Serializer\HtmlFormatter;
 use RemexHtml\Serializer\SerializerNode;
+use Sanitizer;
 
 /**
  * @internal
@@ -34,6 +35,7 @@ class RemexCompatFormatter extends HtmlFormatter {
 
 	public function characters( SerializerNode $parent, $text, $start, $length ) {
 		$text = parent::characters( $parent, $text, $start, $length );
+
 		if ( $parent->namespace !== HTMLData::NS_HTML
 			|| !isset( $this->rawTextElements[$parent->name] )
 		) {
@@ -41,6 +43,9 @@ class RemexCompatFormatter extends HtmlFormatter {
 				$text = call_user_func( $this->textProcessor, $text );
 			}
 		}
+
+		// Ensure a consistent representation for all entities
+		$text = Sanitizer::normalizeCharReferences( $text );
 		return $text;
 	}
 
@@ -65,6 +70,7 @@ class RemexCompatFormatter extends HtmlFormatter {
 		$s = "<$name";
 		foreach ( $attrs->getValues() as $attrName => $attrValue ) {
 			$encValue = strtr( $attrValue, $this->attributeEscapes );
+			$encValue = Sanitizer::normalizeCharReferences( $encValue );
 			$s .= " $attrName=\"$encValue\"";
 		}
 		if ( $node->namespace === HTMLData::NS_HTML && isset( $this->voidElements[$name] ) ) {

tests/phpunit/unit/includes/parser/TidyTest.php

@@ -37,6 +37,7 @@ class TidyTest extends \MediaWikiUnitTestCase {
     </mrow>
   </math>
 MathML;
+		$testMathML = Sanitizer::normalizeCharReferences( $testMathML );
 		return [
 			[
 				'<mw:editsection page="foo" section="bar">foo</mw:editsection>',