Use BCP47 compliant format for locale representation

base/android/java/src/org/chromium/base/LocaleUtils.java

Index: base/android/java/src/org/chromium/base/LocaleUtils.java
diff --git a/base/android/java/src/org/chromium/base/LocaleUtils.java b/base/android/java/src/org/chromium/base/LocaleUtils.java
index 5c26e7a50c28618ad009b7499bd1954b44b4541c..ba4fb6dcfbfafba91dc496b04f75beb736b57bfa 100644
--- a/base/android/java/src/org/chromium/base/LocaleUtils.java
+++ b/base/android/java/src/org/chromium/base/LocaleUtils.java
@@ -4,6 +4,8 @@
 
 package org.chromium.base;
 
+import android.os.Build;
+
 import org.chromium.base.annotations.CalledByNative;
 
 import java.util.Locale;
@@ -19,26 +21,73 @@ public class LocaleUtils {
     }
 
     /**
-     * @return the string for the given locale, translating
-     * Android deprecated language codes into the modern ones
-     * used by Chromium.
+     * @return the string for the given locale with language and/or country code, translating
+     * Android deprecated language codes into the modern ones used by Chromium.
      */
     public static String getLocale(Locale locale) {
-        String language = locale.getLanguage();
+        String language = languageAdjust(locale.getLanguage());
         String country = locale.getCountry();
 
-        // Android uses deprecated lanuages codes for Hebrew and Indonesian but Chromium uses the
-        // updated codes. Also, Android uses "tl" while Chromium uses "fil" for Tagalog/Filipino.
-        // So apply a mapping.
+        return country.isEmpty() ? language : language + "-" + country;
+    }
+
+    /**
+     * Handle the special cases in converting a language code/region code pair into an ISO-639-1
+     * language tag.
+     * @return updated a language code if necessary.
+     **/
+    public static String languageAdjust(String language) {

[ksk1, 2016/10/12 05:52:50]

Can this be private?

[Yirui Huang, 2016/10/12 09:02:04]

Done.

+        // Android uses deprecated lanuages codes for Hebrew, Yiddish and Indonesian but
+        // Chromium uses the updated codes. Also, Android uses "tl" while Chromium uses "fil"
+        // for Tagalog/Filipino. So apply a mapping.
         // See http://developer.android.com/reference/java/util/Locale.html
         if ("iw".equals(language)) {
-            language = "he";
+            return "he";
+        } else if ("ji".equals(language)) {
+            return "yi";
         } else if ("in".equals(language)) {
-            language = "id";
+            return "id";
         } else if ("tl".equals(language)) {
-            language = "fil";
+            return "fil";
+        } else {
+            return language;
+        }
+    }
+
+    /**
+     * @return the locale that best represents the language tag that has language and country code.
+     */

[ksk1, 2016/10/12 05:52:50]

Can we have unit tests for forLanguageTag and toLanguageTag?

[Yirui Huang, 2016/10/12 09:02:04]

Done.

+    public static Locale forLanguageTag(String languageTag) {
+        if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.LOLLIPOP) {
+            Locale locale = Locale.forLanguageTag(languageTag);

[ksk1, 2016/10/12 05:52:50]

return Locale.forLanguageTag(languageTag);

[Yirui Huang, 2016/10/12 09:02:04]

Done.

+            return locale;
+        } else {

[ksk1, 2016/10/12 05:52:50]

How about removing else?

[Yirui Huang, 2016/10/12 09:02:04]

Done.

+            String language;
+            String country;
+            if (languageTag.length() == 2 || languageTag.length() == 3) {
+                language = languageAdjust(languageTag);
+                return new Locale(language);
+            } else if (languageTag.charAt(2) == '-') {

[ksk1, 2016/10/12 05:52:50]

This throws IndexOutOfBoundsException when languageTag length is smaller than 3.
Please have length check before calling charAt and substring.
How about:
String[] tags = languageTag.split("-");
if (tags.length == 0) {
return new Locale("");
}
String language = tags[0];
if (language.length() != 2 && language.length() != 3) {
return new Locale("");
}
if (tag.length == 1) {
return new Locale(language);
}
String country = tag[1];
if (country.length() != 2 && country.length() != 3) {
return new Locale(language);
}
return new Locale(language, country);

[Yirui Huang, 2016/10/12 09:02:04]

For original input languageTag, I assumed that all language tags are in well
formed, since LocalesUtils.getDefaultString() is called to form all tags. But I
agreed that this way, it is more safer to deal to exceptions. Thanks!

+                language = languageAdjust(languageTag.substring(0, 2));
+                country = languageTag.substring(3);
+            } else { // length of the language code is 3.
+                language = languageAdjust(languageTag.substring(0, 3));
+                country = languageTag.substring(4);
+            }
+            return new Locale(language, country);
+        }
+    }
+
+    /**
+     * @return a well-formed IETF BCP 47 language tag with language and country code that
+     * represents this locale.
+     */
+    public static String toLanguageTag(Locale locale) {
+        if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.LOLLIPOP) {
+            return locale.toLanguageTag();
+        } else {
+            return getLocale(locale);
         }
-        return country.isEmpty() ? language : language + "-" + country;
     }
 
     /**