Implement idna_is_valid_u_label

src/core/idna/idna.cc

@@ -147,6 +147,52 @@ auto idna_passes_contextj(const std::u32string_view label,
   return true;
 }
 
+// TODO: Reject labels that are not NFC-normalized
+auto idna_is_valid_u_label(const std::u32string_view label) noexcept -> bool {
+  if (label.empty()) {
+    return false;
+  }
+
+  // RFC 5891 §4.2.3.1: must not start or end with a hyphen, and must not
+  // have a hyphen in both positions 3 and 4 (the IDNA A-label prefix
+  // shape "xn--" must not appear in U-labels).
+  if (label.front() == U'-' || label.back() == U'-') {
+    return false;
+  }
+  if (label.size() >= 4 && label[2] == U'-' && label[3] == U'-') {
+    return false;
+  }
+
+  // RFC 5891 §4.2.3.2: must not start with a combining mark.
+  if (is_combining_mark(label.front())) {
+    return false;
+  }
+
+  // RFC 5891 §4.2.3.3: every codepoint must be PVALID or satisfy its
+  // CONTEXTJ / CONTEXTO contextual rule. DISALLOWED and UNASSIGNED reject.
+  for (std::size_t position = 0; position < label.size(); ++position) {
+    switch (idna_property(label[position])) {
+      case IDNAProperty::PValid:
+        break;
+      case IDNAProperty::ContextJ:
+        if (!idna_passes_contextj(label, position)) {
+          return false;
+        }
+        break;
+      case IDNAProperty::ContextO:
+        if (!idna_passes_contexto(label, position)) {
+          return false;
+        }
+        break;
+      case IDNAProperty::Disallowed:
+      case IDNAProperty::Unassigned:
+        return false;
+    }
+  }
+
+  return true;
+}
+
 auto idna_passes_bidi_rule(const std::u32string_view label) noexcept -> bool {
   if (label.empty()) {
     return false;

src/core/idna/include/sourcemeta/core/idna.h

@@ -98,6 +98,23 @@ auto idna_passes_contextj(const std::u32string_view label,
 SOURCEMETA_CORE_IDNA_EXPORT
 auto idna_passes_bidi_rule(const std::u32string_view label) noexcept -> bool;
 
+/// @ingroup idna
+/// Return whether the given label is a valid U-label per RFC 5891 §4. See
+/// https://www.rfc-editor.org/rfc/rfc5891#section-4 for the criteria.
+/// The Bidi rule is not checked here because Bidi domain detection is a
+/// property of the whole domain, not of a single label. For example:
+///
+/// ```cpp
+/// #include <sourcemeta/core/idna.h>
+/// #include <cassert>
+///
+/// assert(sourcemeta::core::idna_is_valid_u_label(U"d\u00E9j\u00E0"));
+/// assert(!sourcemeta::core::idna_is_valid_u_label(U"ab--cd"));
+/// assert(!sourcemeta::core::idna_is_valid_u_label(U"-abc"));
+/// ```
+SOURCEMETA_CORE_IDNA_EXPORT
+auto idna_is_valid_u_label(const std::u32string_view label) noexcept -> bool;
+
 } // namespace sourcemeta::core
 
 #endif

test/idna/CMakeLists.txt

@@ -3,7 +3,8 @@ sourcemeta_googletest(NAMESPACE sourcemeta PROJECT core NAME idna
     idna_property_test.cc
     idna_passes_contexto_test.cc
     idna_passes_contextj_test.cc
-    idna_passes_bidi_rule_test.cc)
+    idna_passes_bidi_rule_test.cc
+    idna_is_valid_u_label_test.cc)
 
 target_link_libraries(sourcemeta_core_idna_unit
   PRIVATE sourcemeta::core::idna)

test/idna/idna_is_valid_u_label_test.cc

@@ -0,0 +1,145 @@
+#include <gtest/gtest.h>
+
+#include <sourcemeta/core/idna.h>
+
+TEST(IDNA_is_valid_u_label, ascii_letters) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"abc"));
+}
+
+TEST(IDNA_is_valid_u_label, ascii_letters_and_digits) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"abc123"));
+}
+
+TEST(IDNA_is_valid_u_label, ascii_with_internal_hyphen) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"ab-cd"));
+}
+
+TEST(IDNA_is_valid_u_label, latin_with_diacritic) {
+  // d, e+acute, j, a+grave
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"d\u00E9j\u00E0"));
+}
+
+TEST(IDNA_is_valid_u_label, hebrew_label) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"\u05D0\u05D1\u05D2"));
+}
+
+TEST(IDNA_is_valid_u_label, arabic_label) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"\u0627\u0628\u062A"));
+}
+
+TEST(IDNA_is_valid_u_label, devanagari_label) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"\u0905\u0915\u092E"));
+}
+
+TEST(IDNA_is_valid_u_label, devanagari_with_virama_zwj) {
+  // KA + VIRAMA + ZWJ + KA forms a valid ContextJ sequence
+  EXPECT_TRUE(
+      sourcemeta::core::idna_is_valid_u_label(U"\u0915\u094D\u200D\u0915"));
+}
+
+TEST(IDNA_is_valid_u_label, arabic_with_zwnj_context) {
+  // Arabic BEH (D) + ZWNJ + Arabic ALEF (R) satisfies ContextJ A.1
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"\u0628\u200C\u0627"));
+}
+
+TEST(IDNA_is_valid_u_label, catalan_middle_dot) {
+  // l + U+00B7 + l satisfies ContextO A.3
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"l\u00B7l"));
+}
+
+TEST(IDNA_is_valid_u_label, single_letter) {
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"a"));
+}
+
+TEST(IDNA_is_valid_u_label, empty_label) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U""));
+}
+
+TEST(IDNA_is_valid_u_label, leading_hyphen) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"-abc"));
+}
+
+TEST(IDNA_is_valid_u_label, trailing_hyphen) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"abc-"));
+}
+
+TEST(IDNA_is_valid_u_label, double_hyphen_at_positions_3_and_4) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"ab--cd"));
+}
+
+TEST(IDNA_is_valid_u_label, xn_prefix_pattern) {
+  // The "xn--" pattern is the A-label prefix and must not appear in U-labels
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"xn--abc"));
+}
+
+TEST(IDNA_is_valid_u_label, exactly_four_chars_with_double_hyphen) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"ab--"));
+}
+
+TEST(IDNA_is_valid_u_label, leading_nonspacing_combining_mark) {
+  // U+0301 combining acute is general category Mn
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"\u0301abc"));
+}
+
+TEST(IDNA_is_valid_u_label, leading_spacing_combining_mark) {
+  // U+093E Devanagari vowel sign Aa is general category Mc
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"\u093E\u0915"));
+}
+
+TEST(IDNA_is_valid_u_label, contains_uppercase_letter) {
+  // Uppercase ASCII is DISALLOWED in IDNA2008
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"abcD"));
+}
+
+TEST(IDNA_is_valid_u_label, contains_full_stop) {
+  // ASCII full stop is DISALLOWED in a label
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"abc.def"));
+}
+
+TEST(IDNA_is_valid_u_label, contains_disallowed_control) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u0007b"));
+}
+
+TEST(IDNA_is_valid_u_label, contains_disallowed_bidi_mark) {
+  // U+200E LEFT-TO-RIGHT MARK is DISALLOWED
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u200Eb"));
+}
+
+TEST(IDNA_is_valid_u_label, contains_unassigned_codepoint) {
+  // U+0E80 is unassigned in the Lao block
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u0E80b"));
+}
+
+TEST(IDNA_is_valid_u_label, zwj_without_virama_context_rejected) {
+  // ZWJ between two non-virama letters fails ContextJ A.2
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u200Db"));
+}
+
+TEST(IDNA_is_valid_u_label, zwnj_with_no_joining_context_rejected) {
+  // ZWNJ between two ASCII letters fails ContextJ A.1
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u200Cb"));
+}
+
+TEST(IDNA_is_valid_u_label, middle_dot_not_flanked_by_l_rejected) {
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u00B7b"));
+}
+
+TEST(IDNA_is_valid_u_label, greek_keraia_followed_by_non_greek_rejected) {
+  // U+0375 must be followed by Greek script
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"a\u0375b"));
+}
+
+TEST(IDNA_is_valid_u_label, arabic_indic_digit_mixed_with_extended_rejected) {
+  // U+0660 and U+06F0 must not appear together
+  EXPECT_FALSE(sourcemeta::core::idna_is_valid_u_label(U"\u0627\u0660\u06F0"));
+}
+
+TEST(IDNA_is_valid_u_label, single_hyphen_in_position_3_only) {
+  // Only position 3 is hyphen, position 4 is not, so allowed
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"ab-cdef"));
+}
+
+TEST(IDNA_is_valid_u_label, single_hyphen_in_position_4_only) {
+  // Only position 4 is hyphen, position 3 is not, so allowed
+  EXPECT_TRUE(sourcemeta::core::idna_is_valid_u_label(U"abc-def"));
+}