BCP47: add function for finding the first variant not matching its prefixes

Part of the implementation of #3307.

src/common/bcp47.cpp

@@ -379,16 +379,9 @@ language_c::matches_prefix(language_c const &prefix,
 }
 
 bool
-language_c::validate_extlang() {
-  if (m_extended_language_subtag.empty())
-    return true;
-
-  auto extlang = mtx::iana::language_subtag_registry::look_up_extlang(m_extended_language_subtag);
-
-  if (!extlang)                 // Should not happen as the parsing checks this already.
-    return false;
-
-  if (extlang->prefixes.empty())
+language_c::validate_prefixes(std::vector<std::string> const &prefixes)
+  const noexcept {
+  if (prefixes.empty())
     return true;
 
   prefix_restrictions_t restrictions;
@@ -399,7 +392,7 @@ language_c::validate_extlang() {
       value = true;
   };
 
-  for (auto const &prefix : extlang->prefixes) {
+  for (auto const &prefix : prefixes) {
     parsed_prefixes.emplace_back(parse(prefix));
     auto const &tag = parsed_prefixes.back();
 
@@ -414,6 +407,44 @@ language_c::validate_extlang() {
     if (matches_prefix(parsed_prefix, restrictions))
       return true;
 
+  return false;
+}
+
+std::string
+language_c::get_first_variant_not_matching_prefixes()
+  const noexcept {
+  if (m_variants.empty())
+    return {};
+
+  for (auto const &variant_str : m_variants) {
+    auto variant = mtx::iana::language_subtag_registry::look_up_variant(variant_str);
+
+    if (!variant)               // Should not happen as the parsing checks this already.
+      continue;
+
+    if (variant->prefixes.empty())
+      continue;
+
+    if (!validate_prefixes(variant->prefixes))
+      return variant_str;
+  }
+
+  return {};
+}
+
+bool
+language_c::validate_extlang() {
+  if (m_extended_language_subtag.empty())
+    return true;
+
+  auto extlang = mtx::iana::language_subtag_registry::look_up_extlang(m_extended_language_subtag);
+
+  if (!extlang)                 // Should not happen as the parsing checks this already.
+    return false;
+
+  if (validate_prefixes(extlang->prefixes))
+    return true;
+
   auto message   = Y("The extended language subtag '{}' must only be used with one of the following prefixes: {}.");
   m_parser_error = fmt::format(message, m_extended_language_subtag, fmt::join(extlang->prefixes, ", "));
 

src/common/bcp47.h

@@ -111,6 +111,8 @@ public:
   std::vector<std::string> const &get_private_use() const noexcept;
   std::string const &get_grandfathered() const noexcept;
 
+  std::string get_first_variant_not_matching_prefixes() const noexcept;
+
 protected:
   std::string format_internal(bool force) const noexcept;
 
@@ -124,6 +126,7 @@ protected:
   bool validate_extensions();
   bool validate_extlang();
   bool validate_variants();
+  bool validate_prefixes(std::vector<std::string> const &prefixes) const noexcept;
   bool matches_prefix(language_c const &prefix, prefix_restrictions_t const &restrictions) const noexcept;
 
   language_c &canonicalize_preferred_values();

tests/unit/common/bcp47.cpp

@@ -567,4 +567,76 @@ TEST(BCP47LanguageTags, NormalizationForDCNCTags) {
   EXPECT_EQ("pt-BR"s,       language_c::parse("QBP"s).format());
 }
 
+TEST(BCP47LanguageTags, VariantPrefixValidation) {
+  language_c::set_normalization_mode(norm_e::none);
+
+  auto l = language_c::parse("pt-BR"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // no variant
+
+  l = language_c::parse("da-DK-fonipa"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // variant without prefixes
+
+  l = language_c::parse("de-1996"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid
+
+  l = language_c::parse("de-DE-1996"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid
+
+  l = language_c::parse("fr-1996"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("1996"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: language code doesn't match
+
+  l = language_c::parse("pt-BR-abl1943"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid
+
+  l = language_c::parse("pt-abl1943"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("abl1943"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: pt-BR is valid, pt isn't
+
+  l = language_c::parse("zh-cmn-Latn-tongyong"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid
+
+  l = language_c::parse("yue-jyutping"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid
+
+  l = language_c::parse("zh-yue-jyutping"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("jyutping"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: yue would be a valid prefix but zh-yue isn't
+
+  l.to_canonical_form();
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid: zh-yue-jyutping canonicals to yue-jyutping & yue is a valid prefix
+
+  l = language_c::parse("cmn-pinyin"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("pinyin"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: missing Latn
+
+  l = language_c::parse("zh-cmn-pinyin"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("pinyin"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: missing Latn
+
+  l = language_c::parse("cmn-Latn-pinyin"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("pinyin"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: only valid for zh-Latn, not cmn-Latn
+
+  l.to_extlang_form();
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid now as extlang form is zh-cmn-Latn-pinyin
+
+  l = language_c::parse("zh-cmn-Latn-pinyin"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ(""s, l.get_first_variant_not_matching_prefixes()); // prefixes valid (directly)
+
+  l = language_c::parse("zh-cmn-Hans-pinyin"s);
+  EXPECT_TRUE(l.is_valid());
+  EXPECT_EQ("pinyin"s, l.get_first_variant_not_matching_prefixes()); // prefixes invalid: script not Latn
+}
+
 }