feat: search use both languages and remove diacrites

web/src/lib/i18n/matching.ts

@@ -0,0 +1,66 @@
+/**
+ * Нормализация строки для поиска: нижний регистр, ё→е, снятие диакритики (NFD).
+ */
+export function normalizeForSearch(value: string): string {
+	return value
+		.toLowerCase()
+		.replaceAll('ё', 'е')
+		.normalize('NFD')
+		.replace(/\p{M}/gu, '');
+}
+
+export type SearchDoc = {
+	id: string;
+	titles: string[];
+	descriptions: string[];
+};
+
+const SUBSEQUENCE_BASE = 20;
+
+function bestTitleScore(titles: string[], q: string): number | null {
+	let best: number | null = null;
+	for (const raw of titles) {
+		const title = normalizeForSearch(raw);
+		if (title.startsWith(q)) return 100;
+		const at = title.indexOf(q);
+		if (at >= 0) {
+			const s = 80 - Math.min(at, 40);
+			if (best === null || s > best) best = s;
+			continue;
+		}
+		let pos = 0;
+		let sub = SUBSEQUENCE_BASE;
+		for (const ch of q) {
+			pos = title.indexOf(ch, pos);
+			if (pos < 0) {
+				sub = -1;
+				break;
+			}
+			sub -= 1;
+			pos += 1;
+		}
+		if (sub >= 0 && (best === null || sub > best)) best = sub;
+	}
+	return best;
+}
+
+/**
+ * Скоринг документа против нормализованного запроса.
+ * Пустой запрос — нейтральный балл; отсутствие совпадения — null.
+ */
+export function scoreDoc(doc: SearchDoc, normalizedQuery: string): number | null {
+	if (normalizedQuery.length === 0) return 1;
+	if (normalizeForSearch(doc.id).includes(normalizedQuery)) {
+		const at = normalizeForSearch(doc.id).indexOf(normalizedQuery);
+		const byId = 60 - Math.min(at, 40);
+		const byTitle = bestTitleScore(doc.titles, normalizedQuery);
+		if (byTitle !== null && byTitle >= 80) return byTitle;
+		return Math.max(byId, byTitle ?? -1);
+	}
+	const byTitle = bestTitleScore(doc.titles, normalizedQuery);
+	if (byTitle !== null && byTitle >= 20) return byTitle;
+	for (const raw of doc.descriptions) {
+		if (normalizeForSearch(raw).includes(normalizedQuery)) return 30;
+	}
+	return byTitle !== null && byTitle > 0 ? byTitle : null;
+}