@furifont/dictionary 0.1.1 → 0.2.0

package/dist/index.d.mts

@@ -0,0 +1,5 @@
+import { Dictionary } from "@furifont/core";
+//#region src/index.d.ts
+/** Loads the compiled dictionary: JMdict and JMnedict place names, aligned by JmdictFurigana, with KANJIDIC2's on'yomi, JLPT levels and uses in Tatoeba's sentences. */
+export declare function loadDictionary(): Promise<Dictionary>;
+//#endregion

package/dist/index.mjs

@@ -0,0 +1,106 @@
+import * as v from "valibot";
+//#region src/overrides.ts
+/**
+* Readings that override the compiled data where it ranks them wrong, in inline notation. JMdict's
+* frequency tags belong to a written form, so across entries they cannot tell
+* two readings apart; and a form that drops its okurigana usually meets it in
+* running text, where the font cannot see it. Users' pins override these.
+*/
+const OVERRIDE_READINGS = [
+	"出[で]来[き]",
+	"何[なん]人[にん]",
+	"何[なん]時[じ]",
+	"三[さん]十[じゅう]日[にち]",
+	"外[そと]付[づ]",
+	"支[し]払[はら]",
+	"字[じ]",
+	"明[みん]朝[ちょう]"
+];
+/**
+* Kanji left bare when alone in a run, though one reading dominates the
+* dictionary: the reading depends on the word around it, which a font cannot
+* see past the run. 月 is つき in a poem and がつ in a date.
+*/
+const BARE_FORMS = [
+	"月",
+	"日",
+	"年"
+];
+//#endregion
+//#region src/index.ts
+const DataSchema = v.object({
+	sources: v.record(v.string(), v.string()),
+	onyomi: v.record(v.string(), v.array(v.string())),
+	entries: v.array(v.tuple([
+		v.string(),
+		v.string(),
+		v.number(),
+		v.string(),
+		v.number(),
+		v.optional(v.literal(1))
+	]))
+});
+const PlacesSchema = v.array(v.tuple([v.string(), v.string()]));
+const JlptSchema = v.array(v.tuple([
+	v.string(),
+	v.string(),
+	v.picklist([
+		1,
+		2,
+		3,
+		4,
+		5
+	])
+]));
+const TatoebaSchema = v.array(v.tuple([
+	v.string(),
+	v.string(),
+	v.pipe(v.number(), v.integer(), v.minValue(1))
+]));
+/**
+* A data file as a JSON module, which Node loads and bundlers split into a
+* lazy chunk. The path is a template so that TypeScript leaves the 11 MB file
+* untyped rather than inferring it.
+*/
+async function readData(name, schema) {
+	const { default: data } = await import(`../data/${name}.json`, { with: { type: "json" } });
+	return v.parse(schema, data);
+}
+/** Loads the compiled dictionary: JMdict and JMnedict place names, aligned by JmdictFurigana, with KANJIDIC2's on'yomi, JLPT levels and uses in Tatoeba's sentences. */
+async function loadDictionary() {
+	const [data, places, jlpt, tatoeba] = await Promise.all([
+		readData("dictionary", DataSchema),
+		readData("places", PlacesSchema),
+		readData("jlpt", JlptSchema),
+		readData("tatoeba", TatoebaSchema)
+	]);
+	const levels = new Map(jlpt.map(([form, furigana, level]) => [`${form}|${furigana}`, level]));
+	const uses = new Map(tatoeba.map(([form, furigana, n]) => [`${form}|${furigana}`, n]));
+	return {
+		sources: data.sources,
+		onyomi: data.onyomi,
+		defaults: OVERRIDE_READINGS,
+		bare: BARE_FORMS,
+		places: places.map(([form, furigana]) => ({
+			form,
+			furigana
+		})),
+		entries: data.entries.map(([form, furigana, nf, priority, order, conjugates]) => {
+			const level = levels.get(`${form}|${furigana}`);
+			const n = uses.get(`${form}|${furigana}`);
+			uses.delete(`${form}|${furigana}`);
+			return {
+				form,
+				furigana,
+				nf,
+				priority,
+				order,
+				...conjugates && { conjugatesReading: true },
+				...level && { jlpt: level },
+				...n && { tatoeba: n }
+			};
+		})
+	};
+}
+//#endregion
+export { loadDictionary };

package/dist/okurigana.d.mts

@@ -0,0 +1,8 @@
+//#region src/okurigana.d.ts
+/**
+ * Words of one kanji run and the kana after it, read as a pair: 来な is こ, 来ま
+ * is き, 一人も is ひとり. The longest key a text starts with decides; null marks
+ * a pair that reads two ways.
+ */
+export declare const OKURIGANA: ReadonlyMap<string, string | null>;
+//#endregion