@dev-pi2pie/word-counter 0.1.3-canary.1 → 0.1.3

package/dist/esm/bin.mjs

@@ -421,7 +421,7 @@ var require_picocolors = /* @__PURE__ */ __commonJSMin(((exports, module) => {
 //#endregion
 //#region src/cli/program/version-embedded.ts
 var import_picocolors = /* @__PURE__ */ __toESM(require_picocolors(), 1);
-const EMBEDDED_PACKAGE_VERSION = "0.1.3-canary.1";
+const EMBEDDED_PACKAGE_VERSION = "0.1.3";
 
 //#endregion
 //#region src/cli/program/version.ts
@@ -1760,7 +1760,7 @@ function detectLatinLocale(char, context) {
 	}
 	return DEFAULT_LOCALE;
 }
-function detectLocaleForChar(char, previousLocale, options = {}, context = resolveLocaleDetectContext(options)) {
+function detectLocaleForChar(char, previousLocale, options = {}, context = resolveLocaleDetectContext(options), allowLatinLocaleCarry = true, allowJapaneseHanCarry = true) {
 	if (regex.hiragana.test(char) || regex.katakana.test(char)) return "ja";
 	if (regex.hangul.test(char)) return "ko";
 	if (regex.arabic.test(char)) return "ar";
@@ -1768,13 +1768,13 @@ function detectLocaleForChar(char, previousLocale, options = {}, context = resol
 	if (regex.devanagari.test(char)) return "hi";
 	if (regex.thai.test(char)) return "th";
 	if (regex.han.test(char)) {
-		if (previousLocale && previousLocale.startsWith("ja")) return previousLocale;
+		if (allowJapaneseHanCarry && previousLocale && previousLocale.startsWith("ja")) return previousLocale;
 		return context.hanHint ?? DEFAULT_HAN_TAG;
 	}
 	if (regex.latin.test(char)) {
 		const hintedLocale = detectLatinLocale(char, context);
 		if (hintedLocale !== DEFAULT_LOCALE) return hintedLocale;
-		if (previousLocale && isLatinLocale(previousLocale, context) && previousLocale !== DEFAULT_LOCALE) return previousLocale;
+		if (allowLatinLocaleCarry && previousLocale && isLatinLocale(previousLocale, context) && previousLocale !== DEFAULT_LOCALE) return previousLocale;
 		if (context.latinHint) return context.latinHint;
 		return DEFAULT_LOCALE;
 	}
@@ -1783,32 +1783,59 @@ function detectLocaleForChar(char, previousLocale, options = {}, context = resol
 
 //#endregion
 //#region src/wc/segment.ts
+const HARD_BOUNDARY_REGEX = /[\r\n,.!?;:，、。！？；：．｡､]/u;
+const LATIN_PROMOTION_BREAK_REGEX = /[\s,.!?;:，、。！？；：．｡､]/u;
 function segmentTextByLocale(text, options = {}) {
 	const context = resolveLocaleDetectContext(options);
 	const chunks = [];
 	let currentLocale = DEFAULT_LOCALE;
 	let buffer = "";
 	let bufferHasScript = false;
+	let sawCarryBoundary = false;
+	const updateCarryBoundaryState = (detected, char) => {
+		if (detected !== null) {
+			sawCarryBoundary = false;
+			return;
+		}
+		if (HARD_BOUNDARY_REGEX.test(char)) sawCarryBoundary = true;
+	};
 	for (const char of text) {
-		const detected = detectLocaleForChar(char, currentLocale, options, context);
+		const detected = detectLocaleForChar(char, currentLocale, options, context, !sawCarryBoundary, !sawCarryBoundary);
 		const targetLocale = detected ?? currentLocale;
 		if (buffer === "") {
 			currentLocale = targetLocale;
 			buffer = char;
 			bufferHasScript = detected !== null;
+			updateCarryBoundaryState(detected, char);
 			continue;
 		}
 		if (detected !== null && !bufferHasScript) {
 			currentLocale = targetLocale;
 			buffer += char;
 			bufferHasScript = true;
+			updateCarryBoundaryState(detected, char);
 			continue;
 		}
 		if (targetLocale !== currentLocale && detected !== null) {
 			if (currentLocale === DEFAULT_LOCALE && isLatinLocale(targetLocale, context)) {
+				const promotionBreakIndex = findLastLatinPromotionBreakIndex(buffer);
+				if (promotionBreakIndex === -1) {
+					currentLocale = targetLocale;
+					buffer += char;
+					bufferHasScript = true;
+					updateCarryBoundaryState(detected, char);
+					continue;
+				}
+				const prefix = buffer.slice(0, promotionBreakIndex + 1);
+				const suffix = buffer.slice(promotionBreakIndex + 1);
+				if (prefix.length > 0) chunks.push({
+					locale: currentLocale,
+					text: prefix
+				});
 				currentLocale = targetLocale;
-				buffer += char;
+				buffer = `${suffix}${char}`;
 				bufferHasScript = true;
+				updateCarryBoundaryState(detected, char);
 				continue;
 			}
 			chunks.push({
@@ -1818,10 +1845,12 @@ function segmentTextByLocale(text, options = {}) {
 			currentLocale = targetLocale;
 			buffer = char;
 			bufferHasScript = true;
+			updateCarryBoundaryState(detected, char);
 			continue;
 		}
 		buffer += char;
 		if (detected !== null) bufferHasScript = true;
+		updateCarryBoundaryState(detected, char);
 	}
 	if (buffer.length > 0) chunks.push({
 		locale: currentLocale,
@@ -1829,6 +1858,14 @@ function segmentTextByLocale(text, options = {}) {
 	});
 	return mergeAdjacentChunks(chunks);
 }
+function findLastLatinPromotionBreakIndex(buffer) {
+	for (let index = buffer.length - 1; index >= 0; index -= 1) {
+		const char = buffer[index];
+		if (!char) continue;
+		if (LATIN_PROMOTION_BREAK_REGEX.test(char)) return index;
+	}
+	return -1;
+}
 function mergeAdjacentChunks(chunks) {
 	if (chunks.length === 0) return chunks;
 	const merged = [];