engine: verbs and adjectives through SKK's okuri-ari entries

A sixth of kanji.dict is SKK's okuri-ari entries, keyed by a stem and
the letter of the okurigana that follows it — かk: 書 描 掛 —, and no
reading ever ends in a letter, so かく offered 確 and 各 but never 書く,
and no verb or adjective could be converted at all.  Now every split of a
complete reading is tried, longest stem first, with the okurigana put
back: かく adds 書く and 描く after the exact candidates, よむ gives
読む, あかい 赤い, おおきい 大きい, いった 言った, かいた 書いた.  A stem
that ends in っ writes it in kana too, so いt and いっt agree.

strans.c

@@ -110,6 +110,18 @@ cyclekouho(int delta)
 		im.sel = (im.sel + delta + im.nkouho) % im.nkouho;
 }
 
+static void
+addkouho(Str *s)
+{
+	int i;
+
+	for(i = 0; i < im.nkouho; i++)
+		if(scmp(&im.kouho[i], s) == 0)
+			return;
+	if(im.nkouho < Maxkouho)
+		im.kouho[im.nkouho++] = *s;
+}
+
 static int
 numberedkouho(u32int ks, u32int mod)
 {
@@ -238,6 +250,56 @@ redraw(void)
 	channbsend(drawc, &dc);
 }
 
+/* SKK's letter for the okurigana a kana begins: its row, or its vowel. */
+static Rune
+okuriletter(Rune r)
+{
+	static char *rows[] = {
+		"aあ", "iい", "uう", "eえ", "oお",
+		"kかきくけこ", "gがぎぐげご", "sさしすせそ", "zざじずぜぞ",
+		"tたちつてとっ", "dだぢづでど", "nなにぬねのん", "hはひふへほ",
+		"bばびぶべぼ", "pぱぴぷぺぽ", "mまみむめも", "yやゆよ",
+		"rらりるれろ", "wわ",
+	};
+	int i;
+
+	for(i = 0; i < nelem(rows); i++)
+		if(utfrune(rows[i]+1, r) != nil)
+			return rows[i][0];
+	return 0;
+}
+
+/*
+ * SKK keys a verb or adjective by its stem and the letter of the
+ * okurigana that follows (かk: 書, 描, ...).  Every split of the reading
+ * is tried, longest stem first, and the okurigana is put back; a っ
+ * ending the stem (いっt: 言, 行) is written in kana too.
+ */
+static void
+dictqokuri(Str *reading)
+{
+	Str key, okuri, kouho[Maxkouho];
+	Rune c;
+	int i, j, n;
+
+	for(i = reading->n - 1; i > 0; i--){
+		c = okuriletter(reading->r[i]);
+		if(c == 0)
+			continue;
+		key = *reading;
+		key.n = i;
+		sputr(&key, c);
+		n = dictlookup(im.l, &key, kouho, Maxkouho);
+		sclear(&okuri);
+		for(j = i - (reading->r[i-1] == L'っ'); j < reading->n; j++)
+			sputr(&okuri, reading->r[j]);
+		for(j = 0; j < n; j++){
+			sappend(&kouho[j], &okuri);
+			addkouho(&kouho[j]);
+		}
+	}
+}
+
 /* Kanji candidates for a complete reading; none is chosen yet. */
 static void
 dictqjp(void)
@@ -248,6 +310,7 @@ dictqjp(void)
 	if(!jpreading(im.l->map, &im.pre, &im.raw, &reading) || reading.n == 0)
 		return;
 	im.nkouho = dictlookup(im.l, &reading, im.kouho, Maxkouho);
+	dictqokuri(&reading);
 }
 
 static int
@@ -472,18 +535,6 @@ transstr(Lang *l, Str *raw, Str *out)
 	commitim(&q, out);
 }
 
-static void
-addkouho(Str *s)
-{
-	int i;
-
-	for(i = 0; i < im.nkouho; i++)
-		if(scmp(&im.kouho[i], s) == 0)
-			return;
-	if(im.nkouho < Maxkouho)
-		im.kouho[im.nkouho++] = *s;
-}
-
 /* The typed keys and their transliteration are both emoji queries. */
 static void
 emojiquery(void)

tests/engine_test.c

@@ -1200,6 +1200,32 @@ engine_japanese_candidates(struct ct *t)
 	checkstr(t, "typo mended", "かんじ", &shown);
 	CT_EQ_INT(t, 2, im.nkouho);
 	CT_CHECK(t, keystroke(Kesc, 0, &com));
+	/* SKK's okuri-ari entries convert verbs: stem plus okurigana. */
+	setdict(jp->dict, "かk", "書 描");
+	setdict(jp->dict, "かく", "確");
+	if(!typekeys(t, "kaku", &com))
+		goto cleanup;
+	if(CT_EQ_INT(t, 3, im.nkouho)){
+		checkstr(t, "exact candidate first", "確", &im.kouho[0]);
+		checkstr(t, "okurigana put back", "書く", &im.kouho[1]);
+		checkstr(t, "second okuri candidate", "描く", &im.kouho[2]);
+	}
+	CT_CHECK(t, keystroke(Kesc, 0, &com));
+	setdict(jp->dict, "おおk", "大");
+	if(!typekeys(t, "ookii", &com))
+		goto cleanup;
+	if(CT_EQ_INT(t, 1, im.nkouho))
+		checkstr(t, "two-kana okurigana", "大きい", &im.kouho[0]);
+	CT_CHECK(t, keystroke(Kesc, 0, &com));
+	setdict(jp->dict, "いt", "言 行");
+	setdict(jp->dict, "いっt", "言 行");
+	if(!typekeys(t, "itta", &com))
+		goto cleanup;
+	if(CT_EQ_INT(t, 2, im.nkouho)){
+		checkstr(t, "sokuon written in kana", "言った", &im.kouho[0]);
+		checkstr(t, "sokuon written in kana", "行った", &im.kouho[1]);
+	}
+	CT_CHECK(t, keystroke(Kesc, 0, &com));
 	/* Backspace looks a complete shorter reading up again. */
 	setdict(jp->dict, "かん", "缶");
 	if(!typekeys(t, "kanji", &com))