formatferry 1.0.16 → 1.0.18

package/dist/index.js

@@ -22126,6 +22126,30 @@ function escapeHtml(str) {
 function computeFontSize(transform) {
   return Math.sqrt(transform[2] ** 2 + transform[3] ** 2);
 }
+var CJK_REGEX2 = /[\u4E00-\u9FFF\u3400-\u4DBF\u3000-\u303F\u3040-\u309F\u30A0-\u30FF\uAC00-\uD7AF\uF900-\uFAFF\uFF00-\uFF60\uFFE0-\uFFEF]/;
+function containsCJK(str) {
+  return CJK_REGEX2.test(str);
+}
+function isCJKChar(ch) {
+  const cp = ch.codePointAt(0);
+  return cp >= 19968 && cp <= 40959 || // CJK Unified Ideographs
+  cp >= 13312 && cp <= 19903 || // CJK Extension A
+  cp >= 12288 && cp <= 12351 || // CJK Symbols and Punctuation
+  cp >= 12352 && cp <= 12447 || // Hiragana
+  cp >= 12448 && cp <= 12543 || // Katakana
+  cp >= 44032 && cp <= 55215 || // Hangul Syllables
+  cp >= 63744 && cp <= 64255 || // CJK Compatibility Ideographs
+  // Fullwidth Forms — exclude halfwidth Katakana (FF61-FF9F) which are narrow
+  cp >= 65280 && cp <= 65376 || // Fullwidth Latin + punctuation
+  cp >= 65504 && cp <= 65519;
+}
+function visualCharCount(str) {
+  let count = 0;
+  for (const ch of str) {
+    count += isCJKChar(ch) ? 2 : 1;
+  }
+  return count;
+}
 function normalizeLigatures(text) {
   return text.normalize("NFC").replace(/ff/g, "ff").replace(/fi/g, "fi").replace(/fl/g, "fl").replace(/ffi/g, "ffi").replace(/ffl/g, "ffl");
 }
@@ -22137,8 +22161,8 @@ function isSafeUrl(url) {
 }
 function detectFontStyle(fontName) {
   if (!fontName) return null;
-  const bold = /(Bold|Black|Heavy|Semibold|Demi)/i.test(fontName) && !/Blackletter/i.test(fontName);
-  const italic = /(Italic|Oblique|Slanted)/i.test(fontName);
+  const bold = /(Bold|Black|Heavy|Semibold|Demi|太|中太)/i.test(fontName) && !/Blackletter/i.test(fontName);
+  const italic = /(Italic|Oblique|Slanted|Kaiti|楷体|斜)/i.test(fontName);
   if (bold && italic) return "bold-italic";
   if (bold) return "bold";
   if (italic) return "italic";
@@ -22188,7 +22212,7 @@ function textItemToPositioned(item) {
   const ty = item.transform[5];
   const rawHeight = item.height ?? Math.abs(item.transform[3]) ?? 10;
   const height = rawHeight > 0 ? rawHeight : 10;
-  const width = item.width ?? item.str.length * (height * 0.5);
+  const width = item.width ?? visualCharCount(item.str) * (height * 0.5);
   const descent = height * DESCENDER_BUDGET_FACTOR;
   const y = ty - descent;
   const paddedHeight = height + descent;
@@ -22276,7 +22300,10 @@ function itemsToLinearText(items) {
       rows.push([it]);
     }
   }
-  const widths = items.map((i) => i.str.length > 0 ? i.width / i.str.length : 0).filter((w) => w > 0);
+  const widths = items.map((i) => {
+    const vc = visualCharCount(i.str);
+    return vc > 0 ? i.width / vc : 0;
+  }).filter((w) => w > 0);
   const medianCharWidth = widths.length ? [...widths].sort((a, b) => a - b)[Math.floor(widths.length / 2)] || 1 : 1;
   const labelGap = medianCharWidth * FORM_LABEL_GAP_FACTOR;
   const NBSP2 = "\xA0\xA0";
@@ -22339,7 +22366,10 @@ function renderTextBlockEnhanced(items, medianPageFontSize, outlineTitles) {
     row.sort((a, b) => a.x - b.x);
     const fsSamples = row.map((it) => it.fontSize ?? 0).filter((f) => f > 0);
     const rowFs = fsSamples.length > 0 ? [...fsSamples].sort((a, b) => a - b)[Math.floor(fsSamples.length / 2)] : medianPageFontSize;
-    const rowWidths = row.map((it) => it.str.length > 0 ? it.width / it.str.length : 0).filter((w) => w > 0);
+    const rowWidths = row.map((it) => {
+      const vc = visualCharCount(it.str);
+      return vc > 0 ? it.width / vc : 0;
+    }).filter((w) => w > 0);
     const rowMedianCharWidth = rowWidths.length ? [...rowWidths].sort((a, b) => a - b)[Math.floor(rowWidths.length / 2)] || 1 : 1;
     const rowLabelGap = rowMedianCharWidth * FORM_LABEL_GAP_FACTOR;
     let plainText = "";
@@ -22422,8 +22452,14 @@ function repairRowHyphenation(rows) {
       const nextRow = rows[i + 1];
       const lastItem = [...row].sort((a, b) => b.right - a.right)[0];
       const firstNext = [...nextRow].sort((a, b) => a.x - b.x)[0];
-      if (lastItem && firstNext && /-$/.test(lastItem.str) && new RegExp("^\\p{Ll}", "u").test(firstNext.str)) {
-        const fixed = { ...lastItem, str: lastItem.str.slice(0, -1) };
+      if (lastItem && firstNext && /-$/.test(lastItem.str) && !containsCJK(lastItem.str) && new RegExp("^\\p{Ll}", "u").test(firstNext.str)) {
+        const hyphenWidth = lastItem.width / lastItem.str.length * 0.4;
+        const fixed = {
+          ...lastItem,
+          str: lastItem.str.slice(0, -1),
+          width: lastItem.width - hyphenWidth,
+          right: lastItem.right - hyphenWidth
+        };
         result.push([...row.map((it) => it === lastItem ? fixed : it), ...nextRow]);
         i += 2;
         continue;
@@ -22435,8 +22471,12 @@ function repairRowHyphenation(rows) {
   return result;
 }
 function detectListMarker(plainText) {
+  const cjkBulletMatch = /^[・〇☆※]\s*(.+)/u.exec(plainText);
+  if (cjkBulletMatch && cjkBulletMatch[1].length > 0) return { kind: "ul", bodyText: cjkBulletMatch[1] };
   const bulletMatch = /^[•◦◆▪▸–\-*·]\s+(.+)/u.exec(plainText);
   if (bulletMatch) return { kind: "ul", bodyText: bulletMatch[1] };
+  const cjkNumberedMatch = /^[\u2460-\u24FF\u3248-\u325F\u3200-\u321E㊀㊁㊂㊃㊄㊅㊆㊇㊈㊉]\s*(.+)/u.exec(plainText);
+  if (cjkNumberedMatch) return { kind: "ol", bodyText: cjkNumberedMatch[1] };
   const numberedMatch = /^\d+[.)]\s+(.+)/.exec(plainText);
   if (numberedMatch) return { kind: "ol", bodyText: numberedMatch[1] };
   const letteredMatch = /^[a-z][.)]\s+(.+)/.exec(plainText);
@@ -22547,7 +22587,6 @@ function extractVectorLines(fnArray, argsArray, pageWidth, pageHeight) {
   let pathStartX = 0;
   let pathStartY = 0;
   let pending = [];
-  const isPathOp = (op) => op === PDF_OP.moveTo || op === PDF_OP.lineTo || op === PDF_OP.closePath || op === PDF_OP.rectangle;
   const isStrokeOp = (op) => op === PDF_OP.stroke || op === PDF_OP.closeStroke || op === PDF_OP.fillStroke;
   const flushPending = () => {
     for (const p of pending) {
@@ -22573,6 +22612,7 @@ function extractVectorLines(fnArray, argsArray, pageWidth, pageHeight) {
     const op = fnArray[i];
     const args = argsArray[i] ?? [];
     if (op === PDF_OP.moveTo) {
+      pending = [];
       curX = args[0] ?? 0;
       curY = args[1] ?? 0;
       pathStartX = curX;
@@ -22598,8 +22638,6 @@ function extractVectorLines(fnArray, argsArray, pageWidth, pageHeight) {
       curY = pathStartY;
     } else if (isStrokeOp(op)) {
       flushPending();
-    } else if (!isPathOp(op)) {
-      pending = [];
     }
   }
   return lines;
@@ -24270,7 +24308,7 @@ function extractArticle(html, domProvider) {
 }
 
 // ../../src/shared/token-estimator.ts
-var CJK_REGEX2 = /[\u4e00-\u9fff\u3400-\u4dbf\u3000-\u303f\u3040-\u309f\u30a0-\u30ff\uac00-\ud7af]/g;
+var CJK_REGEX3 = /[\u4e00-\u9fff\u3400-\u4dbf\u3000-\u303f\u3040-\u309f\u30a0-\u30ff\uac00-\ud7af]/g;
 var MAX_INPUT_LENGTH = 1e6;
 function estimateTokens(text) {
   if (!text || text.length === 0) {
@@ -24279,7 +24317,7 @@ function estimateTokens(text) {
   if (text.length > MAX_INPUT_LENGTH) {
     return Math.round(text.length / 4);
   }
-  const cjkMatches = text.match(CJK_REGEX2);
+  const cjkMatches = text.match(CJK_REGEX3);
   const cjkCount = cjkMatches ? cjkMatches.length : 0;
   const otherCount = text.length - cjkCount;
   const cjkTokens = cjkCount / 1.5;
@@ -24639,6 +24677,9 @@ var ToonifyService = class {
       paragraphLines.push(line);
       endIndex++;
     }
+    if (paragraphLines.length === 0) {
+      return { block: null, endIndex: startIndex };
+    }
     return {
       block: { p: paragraphLines.join(" ").trim() },
       endIndex: endIndex - 1
@@ -25317,8 +25358,8 @@ import { readFileSync as readFileSync4 } from "fs";
 import { resolve, dirname as dirname3 } from "path";
 import { fileURLToPath } from "url";
 var getVersion = () => {
-  if ("1.0.16") {
-    return "1.0.16";
+  if ("1.0.18") {
+    return "1.0.18";
   }
   try {
     const __dirname = dirname3(fileURLToPath(import.meta.url));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "formatferry",
-  "version": "1.0.16",
+  "version": "1.0.18",
   "description": "FormatFerry CLI — Convert HTML, DOCX, PDF, XLSX, CSV, JSON, XML, PPTX to Markdown",
   "keywords": [
     "markdown",