kordoc 4.17.0 → 4.17.2

package/dist/index.d.cts

@@ -270,7 +270,8 @@ interface ParseOptions {
      * `images` 는 비고, 블록의 `imageData` 도 떼며, HWP5 `inlineImages` 도 무시한다.
      * 그림 자리 표시(`![image](…)`)는 마크다운·블록에 그대로 남아 위치는 알 수 있다.
      * PDF 는 PNG 인코딩을 건너뛴다. 거르는 기준은 같되 메모리 보호용 128MB 누적 상한은 걸리지
-     * 않아, 그 상한 뒤의 그림도 자리 표시가 남는다.
+     * 않아, 그 상한 뒤의 그림도 자리 표시가 남는다. DOCX 는 그림 파트(`word/media`·`word/embeddings`)를
+     * 풀지 않고 ZIP 비압축 상한(100MB)에도 세지 않는다 (#108).
      *
      * 검색 색인처럼 글자만 필요한 호출자용이다. 그림이 많은 문서는 base64 로 불어난 이미지가
      * JSON 출력의 대부분을 차지했다 (PDF 실측 200MB → 9.5MB, CPU 시간 30% 감소). 본문 글자와

package/dist/index.d.ts

@@ -270,7 +270,8 @@ interface ParseOptions {
      * `images` 는 비고, 블록의 `imageData` 도 떼며, HWP5 `inlineImages` 도 무시한다.
      * 그림 자리 표시(`![image](…)`)는 마크다운·블록에 그대로 남아 위치는 알 수 있다.
      * PDF 는 PNG 인코딩을 건너뛴다. 거르는 기준은 같되 메모리 보호용 128MB 누적 상한은 걸리지
-     * 않아, 그 상한 뒤의 그림도 자리 표시가 남는다.
+     * 않아, 그 상한 뒤의 그림도 자리 표시가 남는다. DOCX 는 그림 파트(`word/media`·`word/embeddings`)를
+     * 풀지 않고 ZIP 비압축 상한(100MB)에도 세지 않는다 (#108).
      *
      * 검색 색인처럼 글자만 필요한 호출자용이다. 그림이 많은 문서는 base64 로 불어난 이미지가
      * JSON 출력의 대부분을 차지했다 (PDF 실측 200MB → 9.5MB, CPU 시간 30% 감소). 본문 글자와

package/dist/index.js

@@ -16,7 +16,7 @@ import {
   mapPuaText,
   markNonLayoutTable,
   noteSuffix
-} from "./chunk-YZ5HHS6A.js";
+} from "./chunk-SDRKHEVT.js";
 import {
   parsePageRange
 } from "./chunk-V4BH5SRQ.js";
@@ -30,7 +30,7 @@ import {
   tidyScriptTags,
   unframeLayoutTables,
   wrapScript
-} from "./chunk-MJBCXW5R.js";
+} from "./chunk-UX3HHAWH.js";
 import {
   KordocError,
   OPTIONAL_DEP_INSTALL_HINT,
@@ -47,7 +47,7 @@ import {
   toArrayBuffer,
   unescapeHtml,
   unzipLimitBytes
-} from "./chunk-ML4NHUR7.js";
+} from "./chunk-YGKC6E6A.js";
 
 // src/plain-markdown.ts
 function toPlainMarkdown(md2) {
@@ -746,6 +746,7 @@ async function extractHwpxStyles(zip, decompressed) {
       continue;
     }
   }
+  if (result.borderEdges.size > 0 && !result.borderEdges.has("0")) result.borderEdges.set("0", { t: false, b: false, l: false, r: false });
   return result;
 }
 function parseCharProperties(doc, map) {
@@ -5657,7 +5658,9 @@ function parseCell(records, lhIdx, end, ctx) {
   if (isHeader) cell2.isHeader = true;
   if (rec.data.length >= 24 && rec.data.readUInt32LE(20) > A4_HEIGHT) cell2.pageSpanning = true;
   if (rec.data.length >= 34) {
-    const edges = ctx.docInfo?.borderEdges?.[rec.data.readUInt16LE(32) - 1];
+    const borderFillId = rec.data.readUInt16LE(32);
+    const definitions = ctx.docInfo?.borderEdges;
+    const edges = borderFillId === 0 && definitions?.length ? { t: false, b: false, l: false, r: false } : definitions?.[borderFillId - 1];
     if (edges) cell2.edges = edges;
   }
   return cell2;
@@ -20233,6 +20236,21 @@ var ListCounter = class {
   }
 };
 
+// src/docx/symbol-font.ts
+var SYMBOL_LO = " !\u2200#\u2203%&\u220B()\u2217+,\u2212./0123456789:;<=>?\u2245\u0391\u0392\u03A7\u0394\u0395\u03A6\u0393\u0397\u0399\u03D1\u039A\u039B\u039C\u039D\u039F\u03A0\u0398\u03A1\u03A3\u03A4\u03A5\u03C2\u03A9\u039E\u03A8\u0396[\u2234]\u22A5_\0\u03B1\u03B2\u03C7\u03B4\u03B5\u03C6\u03B3\u03B7\u03B9\u03D5\u03BA\u03BB\u03BC\u03BD\u03BF\u03C0\u03B8\u03C1\u03C3\u03C4\u03C5\u03D6\u03C9\u03BE\u03C8\u03B6{|}\u223C";
+var SYMBOL_HI = "\u20AC\u03D2\u2032\u2264\u2044\u221E\u0192\u2663\u2666\u2665\u2660\u2194\u2190\u2191\u2192\u2193\xB0\xB1\u2033\u2265\xD7\u221D\u2202\u2022\xF7\u2260\u2261\u2248\u2026\u23D0\u23AF\u21B5\u2135\u2111\u211C\u2118\u2297\u2295\u2205\u2229\u222A\u2283\u2287\u2284\u2282\u2286\u2208\u2209\u2220\u2207\xAE\xA9\u2122\u220F\u221A\u22C5\xAC\u2227\u2228\u21D4\u21D0\u21D1\u21D2\u21D3\u25CA\u27E8\xAE\xA9\u2122\u2211\u239B\u239C\u239D\u23A1\u23A2\u23A3\u23A7\u23A8\u23A9\u23AA\0\u27E9\u222B\u2320\u23AE\u2321\u239E\u239F\u23A0\u23A4\u23A5\u23A6\u23AB\u23AC\u23AD";
+var DINGBAT_FONT_RE = /^\s*(?:wingdings|webdings|marlett|zapf\s*dingbats)/i;
+function symbolChar(font, char) {
+  const code = parseInt(char ?? "", 16);
+  if (!(code >= 32 && code <= 65535) || DINGBAT_FONT_RE.test(font ?? "")) return "";
+  if (/^\s*symbol\s*$/i.test(font ?? "")) {
+    const lo = code & 255;
+    const c = lo >= 32 && lo <= 126 ? SYMBOL_LO[lo - 32] : lo >= 160 && lo <= 254 ? SYMBOL_HI[lo - 160] : "";
+    return c === "\0" ? "" : c;
+  }
+  return code >= 61440 && code <= 61695 ? "" : String.fromCharCode(code);
+}
+
 // src/docx/parser.ts
 import { DOMParser as DOMParser3 } from "@xmldom/xmldom";
 
@@ -20240,8 +20258,12 @@ import { DOMParser as DOMParser3 } from "@xmldom/xmldom";
 function runToLatex(r) {
   let out = "";
   for (const t of childrenByLocalName(r, "t")) out += t.textContent ?? "";
-  return out;
+  return out.replace(/[#%]/g, "\\$&");
+}
+function join(a, b) {
+  return /\\[A-Za-z]+$/.test(a) && /^[A-Za-z]/.test(b) ? a + " " + b : a + b;
 }
+var EQ_NUMBER_RE = /\s*\\#\s*(?:\\left\((.+)\\right\)|\(([^()]*)\))\s*$/;
 var FUNC_NAMES = /* @__PURE__ */ new Set([
   "sin",
   "cos",
@@ -20361,7 +20383,7 @@ function grp(body) {
 function childrenToLatex(parent) {
   let out = "";
   for (const ch of elementChildren(parent)) {
-    out += nodeToLatex(ch);
+    out = join(out, nodeToLatex(ch));
   }
   return out;
 }
@@ -20486,7 +20508,7 @@ function nodeToLatex(el) {
       }
       const items = childrenByLocalName(el, "e").map(childrenToLatex);
       const body = items.join(sep);
-      return "\\left" + mapDelim(beg, true) + body + "\\right" + mapDelim(end, false);
+      return join("\\left" + mapDelim(beg, true), body) + "\\right" + mapDelim(end, false);
     }
     // 행렬
     case "m": {
@@ -20555,10 +20577,16 @@ function nodeToLatex(el) {
     // group character (over/underset 비슷)
     case "groupChr":
       return childrenToLatex(findChildByLocalName(el, "e") ?? el);
-    // box/borderBox/phantom/eqArr/… 는 자식 본문만 유지
+    // 수식 배열 — 한 줄짜리 끝의 수식 번호 `#(43)` 는 \tag{43} (#106). 여러 줄은 종전대로 이어 붙인다
+    case "eqArr": {
+      const rows = childrenByLocalName(el, "e");
+      const one = rows.length === 1 ? childrenToLatex(rows[0]) : "";
+      const m = one.match(EQ_NUMBER_RE);
+      return m ? one.slice(0, m.index) + " \\tag{" + (m[1] ?? m[2]).trim() + "}" : childrenToLatex(el);
+    }
+    // box/borderBox/phantom/… 는 자식 본문만 유지
     case "borderBox":
     case "phant":
-    case "eqArr":
       return childrenToLatex(el);
     // 최상위 컨테이너
     case "oMath":
@@ -20598,6 +20626,7 @@ function isDisplayMath(el) {
 
 // src/docx/parser.ts
 var MAX_DECOMPRESS_SIZE5 = unzipLimitBytes(100 * 1024 * 1024);
+var MEDIA_PART_RE = /^word\/(?:media|embeddings)\//;
 function matchesLocal(el, localName3) {
   return el.localName === localName3 || (el.tagName?.endsWith(`:${localName3}`) ?? false);
 }
@@ -20615,7 +20644,7 @@ function effectiveChildElements(parent) {
           result.push(...effectiveChildElements(c));
         }
       }
-    } else if (matchesLocal(el, "ins") || matchesLocal(el, "smartTag")) {
+    } else if (matchesLocal(el, "ins") || matchesLocal(el, "smartTag") || matchesLocal(el, "dir") || matchesLocal(el, "bdo")) {
       result.push(...effectiveChildElements(el));
     } else {
       result.push(el);
@@ -20823,6 +20852,7 @@ function extractRun(r) {
     if (matchesLocal(el, "t")) text += el.textContent ?? "";
     else if (matchesLocal(el, "br") || matchesLocal(el, "cr")) text += "\n";
     else if (matchesLocal(el, "tab")) text += " ";
+    else if (matchesLocal(el, "sym")) text += symbolChar(getAttr(el, "font"), getAttr(el, "char"));
   }
   let bold = false;
   let italic = false;
@@ -20873,7 +20903,19 @@ function collectInline(p, footnotes, rels) {
     fieldInstr = "";
     fieldDisplay = "";
   };
+  const placedMath = /* @__PURE__ */ new Set();
+  const mathText = (om) => {
+    const latex = ommlElementToLatex(om);
+    return !latex ? "" : isDisplayMath(om) ? " $$" + latex + "$$ " : "$" + latex + "$";
+  };
   for (const el of effectiveChildElements(p)) {
+    if (matchesLocal(el, "oMath") || matchesLocal(el, "oMathPara")) {
+      placedMath.add(el);
+      const m = mathText(el);
+      if (fieldActive && fieldStage === "display") fieldDisplay += m;
+      else parts.push(m);
+      continue;
+    }
     if (matchesLocal(el, "hyperlink")) {
       const rId = getAttr(el, "id");
       const anchor = getAttr(el, "anchor");
@@ -20925,10 +20967,9 @@ function collectInline(p, footnotes, rels) {
   }
   flushField();
   for (const om of collectOmmlRoots(p)) {
-    const latex = ommlElementToLatex(om);
-    if (!latex) continue;
-    if (isDisplayMath(om)) parts.push(" $$" + latex + "$$ ");
-    else parts.push(" $" + latex + "$ ");
+    if (placedMath.has(om)) continue;
+    const m = mathText(om);
+    if (m) parts.push(isDisplayMath(om) ? m : " " + m + " ");
   }
   const text = tidyScriptTags(parts.join("")).replace(/[ \t]{2,}/g, " ").trim();
   return { text, bold, italic, footnoteText };
@@ -21080,7 +21121,7 @@ function collectCellText(tc2, styles, numbering, footnotes, rels, depth) {
   }
   return parts;
 }
-async function buildImageMap(zip, rels, doc, warnings) {
+async function buildImageMap(zip, rels, doc, warnings, readBytes) {
   const imageMap = /* @__PURE__ */ new Map();
   const images = [];
   let imgIdx = 0;
@@ -21101,7 +21142,8 @@ async function buildImageMap(zip, rels, doc, warnings) {
     const imgFile = zip.file(imgPath);
     if (!imgFile) continue;
     try {
-      const data = await imgFile.async("uint8array");
+      let data = null;
+      if (readBytes) data = await imgFile.async("uint8array");
       imgIdx++;
       const ext = imgPath.split(".").pop()?.toLowerCase() ?? "png";
       const mimeMap = {
@@ -21117,8 +21159,10 @@ async function buildImageMap(zip, rels, doc, warnings) {
         svg: "image/svg+xml"
       };
       const filename = `image_${String(imgIdx).padStart(3, "0")}.${ext}`;
-      const mimeType = mimeMap[ext] ?? detectImageMime(data) ?? "application/octet-stream";
-      images.push({ filename, data, mimeType, source: imgPath });
+      if (data) {
+        const mimeType = mimeMap[ext] ?? detectImageMime(data) ?? "application/octet-stream";
+        images.push({ filename, data, mimeType, source: imgPath });
+      }
       imageMap.set(embedId, filename);
     } catch (err) {
       warnings.push({
@@ -21199,7 +21243,8 @@ async function parseHeaderFooterBlocks(zip, doc, rels, kind, styles, numbering,
   return out;
 }
 async function parseDocxDocument(buffer, options) {
-  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE5);
+  const readImages = options?.images !== false;
+  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE5, void 0, { re: MEDIA_PART_RE, skip: !readImages });
   const zip = await JSZip5.loadAsync(buffer);
   const warnings = [];
   const docFile = zip.file("word/document.xml");
@@ -21253,7 +21298,7 @@ async function parseDocxDocument(buffer, options) {
   if (body.length === 0) {
     throw new KordocError("DOCX \uBCF8\uBB38(w:body)\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
   }
-  const { imageMap, images } = await buildImageMap(zip, rels, doc, warnings);
+  const { imageMap, images } = await buildImageMap(zip, rels, doc, warnings, readImages);
   const linkedImages = /* @__PURE__ */ new Set();
   const blocks = [];
   const bodyEl = body[0];
@@ -25024,11 +25069,10 @@ var FONTS = {
 function hasFontMetrics(face) {
   return !!face && Object.prototype.hasOwnProperty.call(FONTS, face.trim());
 }
-function fontAdvanceEm1000(face, cp) {
+function fontWidthFn(face) {
+  if (!hasFontMetrics(face)) return null;
   const m = FONTS[face.trim()];
-  if (!m) return null;
-  if (cp >= 32 && cp <= 126) return m.ascii[cp - 32];
-  return m.ex[cp] ?? m.cjk;
+  return (cp) => cp >= 32 && cp <= 126 ? m.ascii[cp - 32] : m.ex[cp] ?? m.cjk;
 }
 
 // src/hwpx/text-metrics.ts
@@ -25218,7 +25262,7 @@ function widthFnOf(faceClass) {
   if (faceClass === "gothic") return gothicWidthEm1000;
   if (faceClass?.startsWith("font:")) {
     const face = faceClass.slice(5);
-    return (cp) => fontAdvanceEm1000(face, cp) ?? gothicWidthEm1000(cp);
+    return fontWidthFn(face) ?? gothicWidthEm1000;
   }
   return charWidthEm1000;
 }
@@ -25269,7 +25313,8 @@ function simulateWrap(text, firstWidth, contWidth, height, ratioPct, mode = "kee
   const breakBefore = (unitPos, w) => {
     let bp = unitPos;
     const u = text[unitPos];
-    if (u !== void 0 && FORBID_START.has(u) && bp - 1 > lineStart() && text[bp - 1] !== " ") bp--;
+    const prevStart = (end) => end - (end >= 2 && /[\uDC00-\uDFFF]/.test(text[end - 1]) && /[\uD800-\uDBFF]/.test(text[end - 2]) ? 2 : 1);
+    if (u !== void 0 && FORBID_START.has(u) && prevStart(bp) > lineStart() && text[bp - 1] !== " ") bp = prevStart(bp);
     while (bp - 1 > lineStart() && FORBID_END.has(text[bp - 1])) bp--;
     if (bp <= lineStart()) bp = unitPos;
     starts.push(bp);
@@ -25747,28 +25792,49 @@ function fitParagraph(text, font, pt2, firstW, contW, minRatio = 88, orphanRatio
   const ladder = [[100, 0], ...SQUEEZE.filter(([r]) => r >= minRatio)];
   const base = simulateWrap(text, f, c, h, 100, "keep", { faceClass });
   if (base.lines < 2) return null;
+  const wraps = /* @__PURE__ */ new Map([["100:0", base]]);
+  const looseness = /* @__PURE__ */ new Map();
+  const wrap = (r, sp) => {
+    const key = `${r}:${sp}`;
+    let w = wraps.get(key);
+    if (!w) {
+      w = simulateWrap(text, f, c, h, r, "keep", { faceClass, spacingPct: sp });
+      wraps.set(key, w);
+    }
+    return w;
+  };
+  const looseAt = (r, sp, w) => {
+    const key = `${r}:${sp}`;
+    let loose = looseness.get(key);
+    if (loose === void 0) {
+      loose = worstLooseness(text, w.starts, f, c, h, r, sp, faceClass);
+      looseness.set(key, loose);
+    }
+    return loose;
+  };
   if (base.lastLineWidth <= c * PULL_TAIL) {
     let pull = null;
     for (const [r, sp] of ladder) {
       const amt = squeezeOf(r, sp);
       if (amt > PULL_MAX + 1e-9 || pull && amt >= pull.amt - 1e-9) continue;
-      const w = simulateWrap(text, f, c, h, r, "keep", { faceClass, spacingPct: sp });
-      if (w.lines < base.lines && worstLooseness(text, w.starts, f, c, h, r, sp, faceClass) <= LOOSE_LIMIT) pull = { r, sp, amt };
+      const w = wrap(r, sp);
+      if (w.lines < base.lines && looseAt(r, sp, w) <= LOOSE_LIMIT) pull = { r, sp, amt };
     }
     if (pull) return { ratio: pull.r, spacing: pull.sp };
   }
   let baseLines = 0;
   let best = null;
   for (const [r, sp] of ladder) {
-    const w = simulateWrap(text, f, c, h, r, "keep", { faceClass, spacingPct: sp });
+    const w = wrap(r, sp);
     if (!best) {
       if (w.lines < 2) return null;
       baseLines = w.lines;
     }
     if (w.lines > baseLines) continue;
     const orphan = w.lines > 1 && w.lastLineWidth <= contW * orphanRatio;
-    const loose = worstLooseness(text, w.starts, f, c, h, r, sp, faceClass);
+    const loose = looseAt(r, sp, w);
     const cost = Math.max(0, loose - LOOSE_LIMIT) + (orphan ? ORPHAN_COST : 0) + squeezeOf(r, sp) * COMPRESS_COST;
+    if (r === 100 && sp === 0 && cost === 0) return null;
     if (!best || cost < best.cost - 1e-9) best = { r, sp, cost };
   }
   return best && (best.r !== 100 || best.sp !== 0) ? { ratio: best.r, spacing: best.sp } : null;
@@ -30472,7 +30538,7 @@ function legalMarkerDepth(marker) {
   return 7;
 }
 function parseLeadingMarker(text, keepMarkers = false) {
-  const t = text.replace(/^[\s　]+/, "").replace(/^\*\*([□■ㅇ○◦●❍◎▸※-])\s*(?=\S)/u, "$1 **");
+  const t = text.replace(/^[\s　]+/, "").replace(/^(\*\*|__)([□■ㅇ○◦●❍◎▸※-])\1(?=\s|$)/u, "$2").replace(/^(\*\*|__)([□■ㅇ○◦●❍◎▸※-])\s*(?=\S)/u, "$2 $1");
   const ref = REF_RE.exec(t);
   if (ref) return { kind: "ref", depth: 0, marker: ref[1], rest: t.slice(ref[0].length).trim() };
   if (keepMarkers) {
@@ -30942,7 +31008,7 @@ function chapterLabel(index, style) {
   return style === "roman" ? `${ROMAN[(index - 1) % 12]}.` : `${index}.`;
 }
 function plain2(text) {
-  return text.replace(/\*\*([^*]+)\*\*/g, "$1").replace(/\*([^*]+)\*/g, "$1").replace(/`([^`]+)`/g, "$1").replace(/!\[[^\]]*\]\([^)]*\)/g, "");
+  return parseInlineMarkdown(text).map((span2) => span2.text).join("");
 }
 function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
   resetTableIds();
@@ -31678,7 +31744,7 @@ function diffTableCells(a, b) {
 // src/form/templates.ts
 import { existsSync, readFileSync } from "fs";
 import { fileURLToPath } from "url";
-import { dirname, join } from "path";
+import { dirname, join as join2 } from "path";
 var BUILTIN_TEMPLATES = [
   {
     id: "gian",
@@ -31704,22 +31770,22 @@ function resolveBuiltinTemplate(name) {
 function builtinTemplateDir() {
   const here = dirname(fileURLToPath(import.meta.url));
   const candidates = [
-    join(here, "..", "..", "templates"),
+    join2(here, "..", "..", "templates"),
     // src/form/templates.ts → 패키지 루트
-    join(here, "..", "templates")
+    join2(here, "..", "templates")
     // dist/*.js (tsup 번들) → 패키지 루트
   ];
   for (const c of candidates) {
-    if (existsSync(join(c, BUILTIN_TEMPLATES[0].file))) return c;
+    if (existsSync(join2(c, BUILTIN_TEMPLATES[0].file))) return c;
   }
   throw new KordocError("\uB0B4\uC7A5 templates \uB514\uB809\uD1A0\uB9AC\uB97C \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4 (\uD328\uD0A4\uC9C0 \uC124\uCE58\uAC00 \uC190\uC0C1\uB418\uC5C8\uC744 \uC218 \uC788\uC2B5\uB2C8\uB2E4)");
 }
 function readBuiltinTemplate(t) {
-  const buf = readFileSync(join(builtinTemplateDir(), t.file));
+  const buf = readFileSync(join2(builtinTemplateDir(), t.file));
   return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);
 }
 function readBuiltinTemplateSample(t) {
-  return JSON.parse(readFileSync(join(builtinTemplateDir(), t.sampleFile), "utf-8"));
+  return JSON.parse(readFileSync(join2(builtinTemplateDir(), t.sampleFile), "utf-8"));
 }
 
 // src/form/seal.ts
@@ -40663,7 +40729,7 @@ function scriptsOff(r, off) {
 }
 async function parseImage(buffer, options) {
   try {
-    const { parseImageDocument } = await import("./image-ocr-QCT3A7X5.js");
+    const { parseImageDocument } = await import("./image-ocr-EREKAS47.js");
     const { blocks, warnings } = await parseImageDocument(buffer, options);
     stripScriptTags({ blocks });
     return {
@@ -40723,7 +40789,7 @@ async function parseHwp(buffer, options) {
 async function parsePdf(buffer, options) {
   let parsePdfDocument;
   try {
-    const mod = await import("./parser-CTRB7PKD.js");
+    const mod = await import("./parser-VH3F4EQU.js");
     parsePdfDocument = mod.parsePdfDocument;
   } catch {
     return {