korean: ㄳ from two consonants, 가 from a vowel typed first; index jamo by scanning

libhangul, and so fcitx5-hangul, joins two lone consonants into the
compound jong they make (rt → ㄳ, split again by a vowel: rtk → ㄱ사),
and by default reorders a vowel typed before its consonant (kr → 가);
strans committed the first jamo and started over.  Backspace on a lone
ㄳ leaves ㄱ, as on a syllable.

The three jamo index tables and their -1 macros were the cho, jung, and
jong arrays inverted by hand, bounded by a Jrange the callers had to
respect; a linear scan of the arrays says the same in ten lines.

ko.c

@@ -4,8 +4,6 @@
 enum
 {
 	Sbase	= 0xAC00,
-	Jbase	= 0x3131,
-	Jrange	= 51,
 	Ncho	= 19,
 	Njung	= 21,
 	Njong	= 28,
@@ -35,41 +33,21 @@ static Rune jong[] = {
 	L'ㅌ', L'ㅍ', L'ㅎ',
 };
 
-static int choidx[Jrange] = {
-	[L'ㄱ'-Jbase] = 1,  [L'ㄲ'-Jbase] = 2,  [L'ㄴ'-Jbase] = 3,
-	[L'ㄷ'-Jbase] = 4,  [L'ㄸ'-Jbase] = 5,  [L'ㄹ'-Jbase] = 6,
-	[L'ㅁ'-Jbase] = 7,  [L'ㅂ'-Jbase] = 8,  [L'ㅃ'-Jbase] = 9,
-	[L'ㅅ'-Jbase] = 10, [L'ㅆ'-Jbase] = 11, [L'ㅇ'-Jbase] = 12,
-	[L'ㅈ'-Jbase] = 13, [L'ㅉ'-Jbase] = 14, [L'ㅊ'-Jbase] = 15,
-	[L'ㅋ'-Jbase] = 16, [L'ㅌ'-Jbase] = 17, [L'ㅍ'-Jbase] = 18,
-	[L'ㅎ'-Jbase] = 19,
-};
+/* A jamo's index in cho, jung, or jong, or -1; jong[0] is no jong. */
+static int
+idx(Rune *t, int n, Rune r)
+{
+	int i;
 
-static int jungidx[Jrange] = {
-	[L'ㅏ'-Jbase] = 1,  [L'ㅐ'-Jbase] = 2,  [L'ㅑ'-Jbase] = 3,
-	[L'ㅒ'-Jbase] = 4,  [L'ㅓ'-Jbase] = 5,  [L'ㅔ'-Jbase] = 6,
-	[L'ㅕ'-Jbase] = 7,  [L'ㅖ'-Jbase] = 8,  [L'ㅗ'-Jbase] = 9,
-	[L'ㅘ'-Jbase] = 10, [L'ㅙ'-Jbase] = 11, [L'ㅚ'-Jbase] = 12,
-	[L'ㅛ'-Jbase] = 13, [L'ㅜ'-Jbase] = 14, [L'ㅝ'-Jbase] = 15,
-	[L'ㅞ'-Jbase] = 16, [L'ㅟ'-Jbase] = 17, [L'ㅠ'-Jbase] = 18,
-	[L'ㅡ'-Jbase] = 19, [L'ㅢ'-Jbase] = 20, [L'ㅣ'-Jbase] = 21,
-};
+	for(i = 0; i < n; i++)
+		if(t[i] == r)
+			return i;
+	return -1;
+}
 
-static int jongidx[Jrange] = {
-	[L'ㄱ'-Jbase] = 2,  [L'ㄲ'-Jbase] = 3,  [L'ㄳ'-Jbase] = 4,
-	[L'ㄴ'-Jbase] = 5,  [L'ㄵ'-Jbase] = 6,  [L'ㄶ'-Jbase] = 7,
-	[L'ㄷ'-Jbase] = 8,  [L'ㄹ'-Jbase] = 9,  [L'ㄺ'-Jbase] = 10,
-	[L'ㄻ'-Jbase] = 11, [L'ㄼ'-Jbase] = 12, [L'ㄽ'-Jbase] = 13,
-	[L'ㄾ'-Jbase] = 14, [L'ㄿ'-Jbase] = 15, [L'ㅀ'-Jbase] = 16,
-	[L'ㅁ'-Jbase] = 17, [L'ㅂ'-Jbase] = 18, [L'ㅄ'-Jbase] = 19,
-	[L'ㅅ'-Jbase] = 20, [L'ㅆ'-Jbase] = 21, [L'ㅇ'-Jbase] = 22,
-	[L'ㅈ'-Jbase] = 23, [L'ㅊ'-Jbase] = 24, [L'ㅋ'-Jbase] = 25,
-	[L'ㅌ'-Jbase] = 26, [L'ㅍ'-Jbase] = 27, [L'ㅎ'-Jbase] = 28,
-};
-
-#define Choidx(r)	(choidx[(r) - Jbase] - 1)
-#define Jungidx(r)	(jungidx[(r) - Jbase] - 1)
-#define Jongidx(r)	(jongidx[(r) - Jbase] - 1)
+#define Choidx(r)	idx(cho, Ncho, r)
+#define Jungidx(r)	idx(jung, Njung, r)
+#define Jongidx(r)	idx(jong, Njong, r)
 
 static Rune jamomap[128] = {
 	['r'] = L'ㄱ', ['R'] = L'ㄲ',
@@ -221,12 +199,23 @@ transko(Im *im, Rune c)
 			sputr(&e.next, compose(ci, ji, 0));
 			return e;
 		}
-		if(Jungidx(last) >= 0 && ji >= 0){
-			comb = combine(cvow, nelem(cvow), last, jm);
-			if(comb){
-				sputr(&e.next, comb);
-				return e;
-			}
+		/* A vowel typed first is reordered under its consonant. */
+		if(Jungidx(last) >= 0 && Choidx(jm) >= 0){
+			sputr(&e.next, compose(Choidx(jm), Jungidx(last), 0));
+			return e;
+		}
+		/* Two vowels or two consonants may be one jamo, ㅘ or ㄳ. */
+		comb = combine(cvow, nelem(cvow), last, jm);
+		if(comb == 0)
+			comb = combine(cjong, nelem(cjong), last, jm);
+		if(comb){
+			sputr(&e.next, comb);
+			return e;
+		}
+		if(ji >= 0 && splitpair(cjong, nelem(cjong), last, &stay, &next)){
+			sputr(&e.s, stay);
+			sputr(&e.next, compose(Choidx(next), ji, 0));
+			return e;
 		}
 		e.s = im->pre;
 		sputr(&e.next, jm);
@@ -294,12 +283,10 @@ backko(Im *im)
 
 	last = slastr(&im->pre);
 	if(!issyl(last)){
-		if(splitpair(cvow, nelem(cvow), last, &stay, &next)){
-			spopr(&im->pre);
-			sputr(&im->pre, stay);
-			return;
-		}
 		spopr(&im->pre);
+		if(splitpair(cvow, nelem(cvow), last, &stay, &next) ||
+		   splitpair(cjong, nelem(cjong), last, &stay, &next))
+			sputr(&im->pre, stay);
 		return;
 	}
 	decompose(last, &c, &j, &jo);

tests/ko_test.c

@@ -25,6 +25,12 @@ korean_sequences(struct ct *t)
 		{ "rr", "ㄱㄱ" },
 		{ "Rk", "까" },
 		{ "rk1", "가1" },
+		{ "rt", "ㄳ" },
+		{ "rtk", "ㄱ사" },
+		{ "rtt", "ㄳㅅ" },
+		{ "kr", "가" },
+		{ "hkr", "과" },
+		{ "kk", "ㅏㅏ" },
 	};
 	Str out;
 	int i;
@@ -81,6 +87,7 @@ korean_backspace(struct ct *t)
 {
 	static const struct { char *before, *after; } cases[] = {
 		{ "ㄱ", "" },
+		{ "ㄳ", "ㄱ" },
 		{ "가", "ㄱ" },
 		{ "과", "고" }, { "괘", "고" }, { "괴", "고" },
 		{ "궈", "구" }, { "궤", "구" }, { "귀", "구" },