Summary

Locale.Builder should throw IllformedLocaleException for duplicate extension singletons, U extension attributes, and U extension keys in BCP47 language tags.

Problem

Locale.Builder is a strict API when constructing Locale instances. Locale.Builder.setLanguageTag(String) is specified as,

the language tag must be well-formed (see Locale) or an exception is thrown (unlike Locale.forLanguageTag, which just discards ill-formed and following portions of the tag).

However, this API currently accepts and silently discards duplicate extension singletons, U extension attributes, and U extension keys.

Regarding BCP 47 language tags, RFC 5646 states,

Each singleton subtag MUST appear at most one time in each tag (other than as a private use subtag). That is, singleton subtags MUST NOT be repeated.

Regarding the BCP 47 Extension U, RFC 6067 states,

Only the first occurrence of an attribute or key conveys meaning in a language tag.

Based off of the specification, it would be permissible to throw an exception for these cases since such tags are considered ill-formed.

Solution

These duplicate occurrences should be defined in the specification as ill-formed. The APIs already describe their respective behavior when a tag is ill-formed or not.

As such, Locale.Builder.setLanguageTag(String) should throw IllformedLocaleException for duplicate extension singletons, U extension attributes, and U extension keys. Locale.Builder.setExtension(char, String) should also throw IllformedLocaleException for duplicate U extension attributes and U extension keys.

Specification

For Locale.java

Under the BCP47 extension definition,

  *   <dt><a id="def_extensions"><b>extensions</b></a></dt>
  *
  *   <dd> A map from single character keys to string values, indicating
- *   extensions apart from language identification.</dd>
+ *   extensions apart from language identification. Keys must not be repeated, unless
+ *   used as values for the key 'x'.</dd>
  *   <dd> <em> BCP 47 deviation:</em> The {@code
  *   extensions} in {@code Locale} implement the semantics and syntax of BCP 47
  *   extension subtags <em>and</em> private use subtags. The {@code extensions}

Under the BCP47 U-extension definition,

  * can be empty, or a series of subtags 3-8 alphanums in length).  A
  * well-formed locale attribute has the form
  * {@code [0-9a-zA-Z]{3,8}} (it is a single subtag with the same
- * form as a locale type subtag).
+ * form as a locale type subtag). When duplicate keys or attributes occur, they
+ * are considered ill-formed.
  *
  * <p>The Unicode locale extension specifies optional behavior in
  * locale-sensitive services.  Although the LDML specification defines

Add RFC 6067 as an external specification,

  *      RFC 4647: Matching of Language Tags
  * @spec https://www.rfc-editor.org/info/rfc5646
  *      RFC 5646: Tags for Identifying Languages
+ * @spec https://www.rfc-editor.org/info/rfc6067
+ *      RFC 6067: BCP 47 Extension U