From 1bb1ff8dd3e3777da88a20c598b331a16c21448f Mon Sep 17 00:00:00 2001
From: Aaron Schulz <aschulz@wikimedia.org>
Date: Wed, 31 Jul 2019 18:48:45 -0400
Subject: [PATCH] Short-circuit WebRequest::getGPCVal() for printable ASCII
 strings

This avoids accessing the content Language instance in many cases

Change-Id: I82bd66496180f947c1af16c2728a1548df03dcf9
---
 includes/WebRequest.php | 21 ++++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)

diff --git a/includes/WebRequest.php b/includes/WebRequest.php
index 6593e49d8a..defe07eaee 100644
--- a/includes/WebRequest.php
+++ b/includes/WebRequest.php
@@ -395,18 +395,25 @@ class WebRequest {
 		# https://www.php.net/variables.external#language.variables.external.dot-in-names
 		# Work around PHP *feature* to avoid *bugs* elsewhere.
 		$name = strtr( $name, '.', '_' );
-		if ( isset( $arr[$name] ) ) {
-			$data = $arr[$name];
+
+		if ( !isset( $arr[$name] ) ) {
+			return $default;
+		}
+
+		$data = $arr[$name];
+		# Optimisation: Skip UTF-8 normalization and legacy transcoding for simple ASCII strings.
+		$isAsciiStr = ( is_string( $data ) && preg_match( '/[^\x20-\x7E]/', $data ) === 0 );
+		if ( !$isAsciiStr ) {
 			if ( isset( $_GET[$name] ) && is_string( $data ) ) {
 				# Check for alternate/legacy character encoding.
-				$contLang = MediaWikiServices::getInstance()->getContentLanguage();
-				$data = $contLang->checkTitleEncoding( $data );
+				$data = MediaWikiServices::getInstance()
+					->getContentLanguage()
+					->checkTitleEncoding( $data );
 			}
 			$data = $this->normalizeUnicode( $data );
-			return $data;
-		} else {
-			return $default;
 		}
+
+		return $data;
 	}
 
 	/**
-- 
2.20.1