Optimize RTL check in ICU to avoid mmap access.

base/i18n/rtl.cc

 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/i18n/rtl.h"
 
+#include <algorithm>
+
 #include "base/files/file_path.h"
 #include "base/logging.h"
-#include "base/metrics/field_trial.h"
+#include "base/strings/string_split.h"
 #include "base/strings/string_util.h"
 #include "base/strings/sys_string_conversions.h"
 #include "base/strings/utf_string_conversions.h"
 #include "third_party/icu/source/common/unicode/locid.h"
 #include "third_party/icu/source/common/unicode/uchar.h"
 #include "third_party/icu/source/common/unicode/uscript.h"
 #include "third_party/icu/source/i18n/unicode/coll.h"
 
 #if defined(OS_IOS)
 #include "base/ios/ios_util.h"
@@ skipping 100 matching lines @@
   g_icu_text_direction = UNKNOWN_DIRECTION;
 }
 
 bool IsRTL() {
   return ICUIsRTL();
 }
 
 bool ICUIsRTL() {
   if (g_icu_text_direction == UNKNOWN_DIRECTION) {
     const icu::Locale& locale = icu::Locale::getDefault();
-    g_icu_text_direction = GetTextDirectionForLocale(locale.getName());
+    g_icu_text_direction = GetTextDirectionForLocaleInStartUp(locale.getName());
   }
   return g_icu_text_direction == RIGHT_TO_LEFT;
 }
 
+TextDirection GetTextDirectionForLocaleInStartUp(const char* locale_name) {
+// On iOS, check for RTL forcing.
+#if defined(OS_IOS)
+  if (ios::IsInForcedRTL())
+    return RIGHT_TO_LEFT;
+#endif
+
+  // This list needs to be updated in alphabetical order if we add more RTL
+  // locales.
+  static const char* kRTLLanguageCodes[] = {"ar", "fa", "he", "iw", "ur"};
+  std::vector<StringPiece> locale_split =
+      SplitStringPiece(locale_name, "-_", KEEP_WHITESPACE, SPLIT_WANT_ALL);
+  const StringPiece& language_code = locale_split[0];
+  if (std::binary_search(kRTLLanguageCodes,
+                         kRTLLanguageCodes + arraysize(kRTLLanguageCodes),
+                         language_code))
+    return RIGHT_TO_LEFT;
+  return LEFT_TO_RIGHT;
+}
+
 TextDirection GetTextDirectionForLocale(const char* locale_name) {

[Alexei Svitkine (slow), 2015/08/12 15:40:02]

Seems like the only place where GetTextDirectionForLocale() gets used after this
change (besides unit tests), is message_bundle.cc.

Does it still make sense to have this separate logic or can that call site also
be switched to the GetTextDirectionForLocaleInStartUp() codepath?

[danduong, 2015/08/12 17:22:42]

That's a good question. jshin@ I know your original suggestion was to have 2
different methods for this, but is it possible that we just make
GetTextDirectionForLocaleInStartUp the only method?

-  const std::string group_name =
-      FieldTrialList::FindFullName("LightSpeed");
-  // StartsWith allows flexibility for this experiment to apply to multiple
-  // group names. To start, this will apply to AvoidMMapOnStartup.
-  if (StartsWith(group_name, "AvoidMMap", CompareCase::SENSITIVE)) {
-    static const char kEnglishLocale[] = "en_";
-    if (StartsWith(locale_name, kEnglishLocale, CompareCase::SENSITIVE))
-      return LEFT_TO_RIGHT;
-  }
-
   // On iOS, check for RTL forcing.
 #if defined(OS_IOS)
   if (ios::IsInForcedRTL())
     return RIGHT_TO_LEFT;
 #endif
 
   UErrorCode status = U_ZERO_ERROR;
   ULayoutType layout_dir = uloc_getCharacterOrientation(locale_name, &status);
   DCHECK(U_SUCCESS(status));
   // Treat anything other than RTL as LTR.
@@ skipping 247 matching lines @@
       begin == kRightToLeftOverride)
     ++begin_index;
   size_t end_index = text.length() - 1;
   if (text[end_index] == kPopDirectionalFormatting)
     --end_index;
   return text.substr(begin_index, end_index - begin_index + 1);
 }
 
 }  // namespace i18n
 }  // namespace base