Merge pull request #128 from contour-terminal/fix/issue-32-hangul-jamo-width

width: set conjoining Hangul V/T Jamo to zero width

Author: christianparpart

src/libunicode/tablegen/multistage_generator.cpp

@@ -526,6 +526,16 @@ void generateMultistageFiles(UcdParser const& parser, std::string const& outputD
         records[static_cast<size_t>(cp)].char_width =
             computeCharWidth(records[static_cast<size_t>(cp)], zeroWidthGCs, eawWide, eawFullwidth);
 
+    // Conjoining Hangul V/T Jamo must be width 0 so that decomposed syllables
+    // (L + V + T) sum to the same width as their precomposed forms (issue #32).
+    // Only V and T need override; L, LV, LVT already get width 2 from EAW=Wide.
+    for (auto const& r: parser.hangulSyllableType())
+    {
+        if (r.property == "V" || r.property == "T")
+            for (auto cp = r.first; cp <= r.last; ++cp)
+                records[static_cast<size_t>(cp)].char_width = 0;
+    }
+
     // ---- Generate multistage tables ----
     std::cout << "[tablegen]   Generating multistage tables (properties)...\n";
 

src/libunicode/tablegen/ucd_parser.cpp

@@ -194,6 +194,7 @@ void UcdParser::parseAll()
     // Phase 2: UCD API files
     loadGraphemeBreakProps();
     loadEastAsianWidths();
+    loadHangulSyllableType();
     loadEmojiProps();
     loadBidiMirrored();
     loadBidiMirroringGlyph();
@@ -443,6 +444,13 @@ void UcdParser::loadEastAsianWidths()
     _eastAsianWidths = loadGenericProperties(_ucdDir + "/EastAsianWidth.txt");
 }
 
+// ---- Hangul Syllable Type ----
+
+void UcdParser::loadHangulSyllableType()
+{
+    _hangulSyllableType = loadGenericProperties(_ucdDir + "/HangulSyllableType.txt");
+}
+
 // ---- Emoji Properties ----
 
 void UcdParser::loadEmojiProps()

src/libunicode/tablegen/ucd_parser.h

@@ -122,6 +122,9 @@ class UcdParser
     /// East Asian Width ranges, sorted by start codepoint.
     [[nodiscard]] auto const& eastAsianWidths() const noexcept { return _eastAsianWidths; }
 
+    /// Hangul Syllable Type ranges (L, V, T, LV, LVT), sorted by start codepoint.
+    [[nodiscard]] auto const& hangulSyllableType() const noexcept { return _hangulSyllableType; }
+
     /// Emoji properties grouped by property name.
     [[nodiscard]] auto const& emojiProps() const noexcept { return _emojiProps; }
 
@@ -174,6 +177,7 @@ class UcdParser
     void loadBlocks();
     void loadGraphemeBreakProps();
     void loadEastAsianWidths();
+    void loadHangulSyllableType();
     void loadEmojiProps();
     void loadBidiMirrored();
     void loadBidiMirroringGlyph();
@@ -229,6 +233,9 @@ class UcdParser
     // East Asian Width
     std::vector<PropertyRange> _eastAsianWidths;
 
+    // Hangul Syllable Type
+    std::vector<PropertyRange> _hangulSyllableType;
+
     // Emoji
     std::map<std::string, std::vector<PropertyRange>> _emojiProps;
 

src/libunicode/width_test.cpp

@@ -39,6 +39,48 @@ TEST_CASE("width.single_codepoint", "[width]")
     CHECK(unicode::width(U'\U0001F480') == 2); // 💀 :skull:
 }
 
+TEST_CASE("width.hangul_jamo", "[width]")
+{
+    // Leading Jamo (Choseong) — EAW=Wide → width 2
+    CHECK(unicode::width(U'\u1100') == 2); // HANGUL CHOSEONG KIYEOK
+    CHECK(unicode::width(U'\u115F') == 2); // HANGUL CHOSEONG FILLER
+    CHECK(unicode::width(U'\uA960') == 2); // HANGUL CHOSEONG TIKEUT-MIEUM (Extended-A)
+    CHECK(unicode::width(U'\uA97C') == 2); // HANGUL CHOSEONG SSANGYEORINHIEUH (Extended-A)
+
+    // Vowel Jamo (Jungseong) — conjoining → width 0
+    CHECK(unicode::width(U'\u1160') == 0); // HANGUL JUNGSEONG FILLER
+    CHECK(unicode::width(U'\u1161') == 0); // HANGUL JUNGSEONG A
+    CHECK(unicode::width(U'\u11A7') == 0); // HANGUL JUNGSEONG O-YAE
+    CHECK(unicode::width(U'\uD7B0') == 0); // HANGUL JUNGSEONG O-YEO (Extended-B)
+    CHECK(unicode::width(U'\uD7C6') == 0); // HANGUL JUNGSEONG ARAEA-E (Extended-B)
+
+    // Trailing Jamo (Jongseong) — conjoining → width 0
+    CHECK(unicode::width(U'\u11A8') == 0); // HANGUL JONGSEONG KIYEOK
+    CHECK(unicode::width(U'\u11FF') == 0); // HANGUL JONGSEONG SSANGNIEUN
+    CHECK(unicode::width(U'\uD7CB') == 0); // HANGUL JONGSEONG NIEUN-RIEUL (Extended-B)
+    CHECK(unicode::width(U'\uD7FB') == 0); // HANGUL JONGSEONG PHIEUPH-THIEUTH (Extended-B)
+
+    // Precomposed syllables — EAW=Wide → width 2
+    CHECK(unicode::width(U'\uAC00') == 2); // HANGUL SYLLABLE GA (first)
+    CHECK(unicode::width(U'\uD7A3') == 2); // HANGUL SYLLABLE HIH (last)
+
+    // Compatibility Jamo — EAW=Wide → width 2
+    CHECK(unicode::width(U'\u3131') == 2); // HANGUL LETTER KIYEOK
+
+    // Halfwidth Jamo — no conjoining → width 1
+    CHECK(unicode::width(U'\uFFA0') == 1); // HALFWIDTH HANGUL FILLER
+    CHECK(unicode::width(U'\uFFBE') == 1); // HALFWIDTH HANGUL LETTER KIYEOK
+}
+
+TEST_CASE("width.hangul_decomposed_syllable", "[width]")
+{
+    // L + V = grapheme cluster of width 2
+    CHECK(unicode::grapheme_cluster_width(U"\u1100\u1161"sv) == 2);
+
+    // L + V + T = grapheme cluster of width 2
+    CHECK(unicode::grapheme_cluster_width(U"\u1100\u1161\u11A8"sv) == 2);
+}
+
 TEST_CASE("grapheme_cluster_width.empty", "[width]")
 {
     CHECK(unicode::grapheme_cluster_width(std::u32string_view {}) == 0);