Expose getCanonicalLocales() for Intl object.

Also add a test for the return object of getCanonicalLocaleList().

See https://github.com/tc39/test262/issues/745 for more details.

BUG=v8:5012
TEST=test262/intl402/Intl/getCanonicalLocales/*
TEST=intl/general/getCanonicalLocales

Review-Url: https://codereview.chromium.org/2239523002
Cr-Commit-Position: refs/heads/master@{#38733}

src/js/i18n.js

@@ -587,6 +587,7 @@ function setOptions(inOptions, extensionMap, keyValues, getOption, outOptions) {
  * Given an array-like, outputs an Array with the numbered
  * properties copied over and defined
  * configurable: false, writable: false, enumerable: true.
+ * When |expandable| is true, the result array can be expanded.
  */
 function freezeArray(input) {
   var array = [];
@@ -604,6 +605,12 @@ function freezeArray(input) {
   return array;
 }
 
+/* Make JS array[] out of InternalArray */
+function makeArray(input) {
+  var array = [];
+  %MoveArrayContents(input, array);
+  return array;
+}
 
 /**
  * It's sometimes desireable to leave user requested locale instead of ICU
@@ -738,6 +745,7 @@ function toTitleCaseTimezoneLocation(location) {
 
 /**
  * Canonicalizes the language tag, or throws in case the tag is invalid.
+ * ECMA 402 9.2.1 steps 7.c ii ~ v.
  */
 function canonicalizeLanguageTag(localeID) {
   // null is typeof 'object' so we have to do extra check.
@@ -755,11 +763,14 @@ function canonicalizeLanguageTag(localeID) {
 
   var localeString = TO_STRING(localeID);
 
-  if (isValidLanguageTag(localeString) === false) {
+  if (isStructuallyValidLanguageTag(localeString) === false) {
     throw %make_range_error(kInvalidLanguageTag, localeString);
   }
 
+  // ECMA 402 6.2.3
   var tag = %CanonicalizeLanguageTag(localeString);
+  // TODO(jshin): This should not happen because the structual validity
+  // is already checked. If that's the case, remove this.
   if (tag === 'invalid-tag') {
     throw %make_range_error(kInvalidLanguageTag, localeString);
   }
@@ -769,20 +780,22 @@ function canonicalizeLanguageTag(localeID) {
 
 
 /**
- * Returns an array where all locales are canonicalized and duplicates removed.
+ * Returns an InternalArray where all locales are canonicalized and duplicates
+ * removed.
  * Throws on locales that are not well formed BCP47 tags.
+ * ECMA 402 8.2.1 steps 1 (ECMA 402 9.2.1) and 2.
  */
-function initializeLocaleList(locales) {
+function canonicalizeLocaleList(locales) {
   var seen = new InternalArray();
   if (!IS_UNDEFINED(locales)) {
     // We allow single string localeID.
     if (typeof locales === 'string') {
       %_Call(ArrayPush, seen, canonicalizeLanguageTag(locales));
-      return freezeArray(seen);
+      return seen;
     }
 
     var o = TO_OBJECT(locales);
-    var len = TO_UINT32(o.length);
+    var len = TO_LENGTH(o.length);
 
     for (var k = 0; k < len; k++) {
       if (k in o) {
@@ -797,20 +810,30 @@ function initializeLocaleList(locales) {
     }
   }
 
-  return freezeArray(seen);
+  return seen;
 }
 
+function initializeLocaleList(locales) {
+  return freezeArray(canonicalizeLocaleList(locales));
+}
 
 /**
- * Validates the language tag. Section 2.2.9 of the bcp47 spec
- * defines a valid tag.
+ * Check the structual Validity of the language tag per ECMA 402 6.2.2:
+ *   - Well-formed per RFC 5646 2.1
+ *   - There are no duplicate variant subtags
+ *   - There are no duplicate singletion (extension) subtags
+ *
+ * One extra-check is done (from RFC 5646 2.2.9): the tag is compared
+ * against the list of grandfathered tags. However, subtags for
+ * primary/extended language, script, region, variant are not checked
+ * against the IANA language subtag registry.
  *
  * ICU is too permissible and lets invalid tags, like
  * hant-cmn-cn, through.
  *
  * Returns false if the language tag is invalid.
  */
-function isValidLanguageTag(locale) {
+function isStructuallyValidLanguageTag(locale) {
   // Check if it's well-formed, including grandfadered tags.
   if (IS_NULL(InternalRegExpMatch(GetLanguageTagRE(), locale))) {
     return false;
@@ -905,6 +928,16 @@ var resolvedAccessor = {
   }
 };
 
+// ECMA 402 section 8.2.1
+InstallFunction(Intl, 'getCanonicalLocales', function(locales) {
+    if (!IS_UNDEFINED(new.target)) {
+      throw %make_type_error(kOrdinaryFunctionCalledAsConstructor);
+    }
+
+    return makeArray(canonicalizeLocaleList(locales));
+  }
+);
+
 /**
  * Initializes the given object so it's a valid Collator instance.
  * Useful for subclassing.

src/runtime/runtime-i18n.cc

@@ -63,6 +63,7 @@ const UChar* GetUCharBufferFromFlat(const String::FlatContent& flat,
 
 }  // namespace
 
+// ECMA 402 6.2.3
 RUNTIME_FUNCTION(Runtime_CanonicalizeLanguageTag) {
   HandleScope scope(isolate);
   Factory* factory = isolate->factory();
@@ -73,6 +74,8 @@ RUNTIME_FUNCTION(Runtime_CanonicalizeLanguageTag) {
   v8::String::Utf8Value locale_id(v8::Utils::ToLocal(locale_id_str));
 
   // Return value which denotes invalid language tag.
+  // TODO(jshin): Can uloc_{for,to}TanguageTag fail even for structually valid
+  // language tags? If not, just add CHECK instead of returning 'invalid-tag'.
   const char* const kInvalidTag = "invalid-tag";
 
   UErrorCode error = U_ZERO_ERROR;

test/intl/general/getCanonicalLocales.js

+// Copyright 2016 the V8 project authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+var locales = ['en-US', 'fr'];
+var result = Intl.getCanonicalLocales(locales);
+var len = result.length
+
+// TODO(jshin): Remove the following when
+// https://github.com/tc39/test262/issues/745 is resolved and
+// test262 in v8 is updated.
+
+assertEquals(Object.getPrototypeOf(result), Array.prototype);
+assertEquals(result.constructor, Array);
+
+for (var key in result) {
+  var desc = Object.getOwnPropertyDescriptor(result, key);
+  assertTrue(desc.writable);
+  assertTrue(desc.configurable);
+  assertTrue(desc.enumerable);
+}
+
+var desc = Object.getOwnPropertyDescriptor(result, 'length');
+assertTrue(desc.writable);
+assertEquals(result.push('de'), desc.value + 1);

test/test262/test262.status

@@ -330,7 +330,10 @@
   'language/module-code/*': [SKIP],
 
   # https://bugs.chromium.org/p/v8/issues/detail?id=5012
-  'intl402/Intl/getCanonicalLocales/*': [FAIL],
+  # http://bugs.icu-project.org/trac/ticket/12671
+  'intl402/Intl/getCanonicalLocales/weird-cases': [FAIL],
+  # https://github.com/tc39/test262/issues/743
+  'intl402/Intl/getCanonicalLocales/main': [FAIL],
 
   # https://bugs.chromium.org/p/v8/issues/detail?id=5115
   'language/statements/class/subclass/class-definition-null-proto-missing-return-override': [FAIL],