fix: Korean keyboard unclickable + German ß — universal diacritic normalization

frontend/src/game.ts

@@ -631,7 +631,18 @@ export const createGameApp = () => {
 
             keyDown(event: KeyboardEvent | { key: string }): void {
                 if (this.animating) return;
-                const key = event.key;
+
+                // Physical keyboard → character mapping (bypasses IME for Korean, etc.)
+                let key = event.key;
+                const physicalKeyMap = this.config?.physical_key_map;
+                if (physicalKeyMap && 'code' in event) {
+                    const code = event.shiftKey ? `Shift${event.code}` : event.code;
+                    const mapped = physicalKeyMap[code];
+                    if (mapped) {
+                        event.preventDefault();
+                        key = mapped;
+                    }
+                }
 
                 if (key === 'Escape') {
                     this.showHelpModal = false;

frontend/src/types/index.ts

@@ -61,6 +61,11 @@ export interface LanguageConfig {
      *  Example for Greek: { "σ": "ς" }
      */
     final_form_map?: Record<string, string>;
+    /** Optional physical key → character map for bypassing IME composition.
+     *  Maps KeyboardEvent.code values to characters.
+     *  Example for Korean: { "KeyQ": "ㅂ", "ShiftKeyQ": "ㅃ", ... }
+     */
+    physical_key_map?: Record<string, string>;
 }
 
 // =============================================================================

tests/conftest.py

@@ -156,5 +156,30 @@ def load_keyboard(lang_code: str) -> list | None:
     return data
 
 
+def load_all_keyboard_chars(lang_code: str) -> set[str]:
+    """Load all keyboard characters across ALL layouts for a language.
+
+    For coverage tests, a character is typeable if it appears on any layout.
+    """
+    keyboard_file = LANGUAGES_DIR / lang_code / f"{lang_code}_keyboard.json"
+    if not keyboard_file.exists():
+        return set()
+    with open(keyboard_file, encoding="utf-8") as f:
+        data = json.load(f)
+
+    control_keys = {"⇨", "⟹", "⌫", "↵", "ENTER", "DEL"}
+    chars = set()
+
+    if isinstance(data, dict) and "layouts" in data:
+        for layout in data["layouts"].values():
+            for row in layout.get("rows", []):
+                chars.update(k for k in row if k not in control_keys)
+    elif isinstance(data, list):
+        for row in data:
+            chars.update(k for k in row if k not in control_keys)
+
+    return chars
+
+
 # Make language codes available for parametrize
 ALL_LANGUAGES = get_all_language_codes()

tests/test_language_config.py

@@ -9,6 +9,7 @@
 from tests.conftest import (
     ALL_LANGUAGES,
     get_diacritic_base_chars,
+    load_all_keyboard_chars,
     load_keyboard,
     load_language_config,
     load_word_list,
@@ -153,7 +154,7 @@ class TestKeyboardConfig:
     """Tests for keyboard configuration."""
 
     # Languages with known keyboard coverage gaps (complex scripts, incomplete keyboards)
-    KEYBOARD_COVERAGE_XFAIL = {"ko", "de"}
+    KEYBOARD_COVERAGE_XFAIL: set[str] = set()
 
     @pytest.mark.parametrize("lang", ALL_LANGUAGES)
     def test_keyboard_covers_all_characters(self, lang):
@@ -178,12 +179,8 @@ def test_keyboard_covers_all_characters(self, lang):
         for word in words:
             word_chars.update(word)
 
-        # Flatten keyboard to get all keys
-        keyboard_chars = set()
-        for row in keyboard:
-            for key in row:
-                if key not in ("⇨", "⟹", "⌫", "ENTER", "DEL"):
-                    keyboard_chars.add(key)
+        # Check all layouts — a character is typeable if it appears on any layout
+        keyboard_chars = load_all_keyboard_chars(lang)
 
         # Get diacritic mapping - chars that can be typed via base char
         diacritic_map = get_diacritic_base_chars(lang)

tests/test_word_lists.py

@@ -14,6 +14,7 @@
 from tests.conftest import (
     ALL_LANGUAGES,
     get_diacritic_base_chars,
+    load_all_keyboard_chars,
     load_blocklist,
     load_characters,
     load_daily_words,
@@ -124,7 +125,7 @@ class TestKeyboardCoverage:
     """Tests for keyboard coverage of word characters."""
 
     # Languages with known keyboard coverage gaps (complex scripts)
-    KEYBOARD_COVERAGE_XFAIL: set[str] = {"ko", "de"}
+    KEYBOARD_COVERAGE_XFAIL: set[str] = set()
 
     @pytest.mark.parametrize("lang", ALL_LANGUAGES)
     def test_keyboard_covers_all_word_characters(self, lang):
@@ -146,13 +147,8 @@ def test_keyboard_covers_all_word_characters(self, lang):
         if not keyboard or all(len(row) == 0 for row in keyboard):
             pytest.skip(f"{lang}: Empty keyboard (app will auto-generate)")
 
-        # Extract all characters from keyboard (load_keyboard returns normalized rows)
-        keyboard_chars = set()
-        for row in keyboard:
-            for key in row:
-                # Skip control keys
-                if key not in ["⇨", "⟹", "⌫", "↵", "ENTER", "DEL"]:
-                    keyboard_chars.add(key)
+        # Check all layouts — a character is typeable if it appears on any layout
+        keyboard_chars = load_all_keyboard_chars(lang)
 
         if not keyboard_chars:
             pytest.skip(f"{lang}: Empty keyboard layout")

webapp/app.py

@@ -672,12 +672,28 @@ def __init__(self, language_code, word_list, keyboard_layout=None):
 
         self.characters = language_characters[language_code]
         # remove chars that aren't used to reduce bloat a bit
-        characters_used = []
+        characters_used = set()
         for word in self.word_list:
-            characters_used += list(word)
-        characters_used = list(set(characters_used))
+            characters_used.update(word)
         self.characters = [char for char in self.characters if char in characters_used]
 
+        # Include diacritic base characters whose variants appear in the word list.
+        # This allows keyboards using base forms (e.g., Compatibility Jamo) to work
+        # with word lists using variant forms (e.g., Hangul Jamo).
+        #
+        # Languages using this mechanism:
+        #   - Korean (ko): Compatibility Jamo ↔ Hangul Jamo (different Unicode blocks)
+        #   - German (de): s ↔ ß (sharp S treated as variant of s)
+        #   - European languages: base letters ↔ accented variants (a ↔ ä, o ↔ ö)
+        #
+        # Future: languages with composition-based scripts (e.g., Devanagari for Hindi,
+        # Thai, Khmer) may need similar keyboard↔wordlist normalization if added.
+        diacritic_map = self.config.get("diacritic_map", {})
+        chars_set = set(self.characters)
+        for base, variants in diacritic_map.items():
+            if base not in chars_set and any(v in characters_used for v in variants):
+                self.characters.append(base)
+
         keyboard_config = keyboards.get(language_code, {"default": None, "layouts": {}})
         self.keyboard_layouts = self._build_keyboard_layouts(keyboard_config)
         self.keyboard_layout_name = self._select_keyboard_layout(
@@ -776,6 +792,11 @@ def _build_key_diacritic_hints(self):
             for key in row:
                 keyboard_keys.add(key.lower())
 
+        # Hide diacritic hints when the map is used for encoding normalization
+        # (e.g., Korean Jamo) rather than player-visible accent variants.
+        if self.config.get("hide_diacritic_hints"):
+            return {}
+
         hints = {}
         for base_char, variants in diacritic_map.items():
             if base_char.lower() in keyboard_keys:

webapp/data/languages/de/language_config.json

@@ -13,7 +13,7 @@
         "u": [
             "ü"
         ],
-        "ss": [
+        "s": [
             "ß"
         ]
     },

webapp/data/languages/ko/ko_blocklist.txt

@@ -0,0 +1,139 @@
+# Korean compound jongseong blocklist
+# These words contain compound final consonants (겹받침) like ㄺ, ㄻ, ㄼ, etc.
+# that are not on the default keyboard layout. They remain valid guesses
+# (playable via the Full keyboard layout) but are excluded from daily selection.
+#
+# Future improvement: decompose compound jongseong into individual jamo and
+# use 6 cells per word (like 꼬들/kordle.kr does), which would eliminate the
+# need for compound keys entirely. See ko_keyboard.json for details.
+# Total: 129 words
+
+갉다
+갉아
+값싸
+값져
+값해
+곪다
+곪아
+곯다
+곯려
+곯아
+굵기
+굵다
+굵어
+굶겨
+굶다
+굶어
+긁다
+긁어
+긁혀
+기슭
+까닭
+꿇다
+꿇려
+꿇어
+끊겨
+끊다
+끊어
+끊여
+끓다
+끓어
+끓여
+낡다
+낡아
+넓다
+넓어
+넓이
+넓혀
+늙다
+늙어
+닭띠
+닮다
+닮아
+닳다
+닳아
+떫다
+떫어
+뚫다
+뚫려
+뚫어
+많다
+많아
+많이
+맑다
+맑아
+묽다
+묽어
+밝기
+밝다
+밝아
+밝혀
+밝히
+밟다
+밟아
+밟혀
+붉다
+붉어
+붉혀
+삶다
+삶아
+섧다
+섧어
+수탉
+싫다
+싫어
+앉다
+앉아
+앉혀
+않다
+않아
+앓다
+앓아
+얇다
+얇아
+얹다
+얹어
+얹혀
+얽다
+얽매
+얽어
+얽혀
+없다
+없애
+없어
+없이
+여덟
+엷다
+엷어
+옭다
+옭매
+옭아
+옮겨
+옮다
+옮아
+옳다
+옳소
+옳아
+외곬
+읊다
+읊어
+읽기
+읽다
+읽어
+읽혀
+잃다
+잃어
+젊다
+젊어
+제값
+짊다
+짊어
+짧다
+짧아
+칡차
+핥다
+핥아
+훑다
+훑어
+흙내
+흙비

webapp/data/languages/ko/ko_keyboard.json

@@ -6,15 +6,19 @@
             "rows": [
                 ["ㅂ", "ㅈ", "ㄷ", "ㄱ", "ㅅ", "ㅛ", "ㅕ", "ㅑ", "ㅐ", "ㅔ"],
                 ["ㅁ", "ㄴ", "ㅇ", "ㄹ", "ㅎ", "ㅗ", "ㅓ", "ㅏ", "ㅣ"],
+                ["ㅃ", "ㅉ", "ㄸ", "ㄲ", "ㅆ", "ㅒ", "ㅖ"],
+                ["ㅘ", "ㅙ", "ㅚ", "ㅝ", "ㅞ", "ㅟ", "ㅢ"],
                 ["⇨", "ㅋ", "ㅌ", "ㅊ", "ㅍ", "ㅠ", "ㅜ", "ㅡ", "⌫"]
             ]
         },
-        "korean_double": {
-            "label": "With Double Consonants",
+        "korean_full": {
+            "label": "Full (전체 자모)",
             "rows": [
-                ["ㅂ", "ㅃ", "ㅈ", "ㅉ", "ㄷ", "ㄸ", "ㄱ", "ㄲ", "ㅅ", "ㅆ", "ㅛ", "ㅕ", "ㅑ", "ㅐ", "ㅒ", "ㅔ", "ㅖ"],
+                ["ㅂ", "ㅃ", "ㅈ", "ㅉ", "ㄷ", "ㄸ", "ㄱ", "ㄲ", "ㅅ", "ㅆ"],
+                ["ㅛ", "ㅕ", "ㅑ", "ㅐ", "ㅒ", "ㅔ", "ㅖ", "ㅘ", "ㅙ", "ㅚ"],
                 ["ㅁ", "ㄴ", "ㅇ", "ㄹ", "ㅎ", "ㅗ", "ㅓ", "ㅏ", "ㅣ"],
-                ["⇨", "ㅋ", "ㅌ", "ㅊ", "ㅍ", "ㅠ", "ㅜ", "ㅡ", "⌫"]
+                ["ㅝ", "ㅞ", "ㅟ", "ㅢ", "ㅋ", "ㅌ", "ㅊ", "ㅍ", "ㅠ", "ㅜ", "ㅡ"],
+                ["⇨", "ㄵ", "ㄶ", "ㄺ", "ㄻ", "ㄼ", "ㄽ", "ㄾ", "ㄿ", "ㅀ", "ㅄ", "⌫"]
             ]
         }
     }