kordoc 4.21.15 → 4.21.17

package/README.md

@@ -12,7 +12,7 @@ HWP 3.x·5.x, HWPX, HWPML, PDF, XLS·XLSX, DOCX, PPTX, PNG·JPG·WebP를 Markdow
 [English](README-EN.md) · [상세 사용법](docs/usage.md) · [벤치마크](docs/benchmarks.md) · [변경 이력](CHANGELOG.md)
 
 - 📊 **PDF 공개 벤치 종합 0.963** — opendataloader-bench 200문서. 공개 12개 파서 비교 1위(2026-09-29 기록). [측정 조건](docs/benchmarks.md)
-- 🇰🇷 **한국 공문서 표 구조 100%** — 원본 HWPX 2,424문서의 보이는 표 **10,342개 전부 구조 일치**. [4.21.15 검증](docs/benchmarks.md)
+- 🇰🇷 **한국 공문서 표 구조 100%** — 원본 HWPX 2,424문서의 보이는 표 **10,342개 전부 구조 일치**. [4.21.17 검증](docs/benchmarks.md)
 
 쓸모 있었다면 GitHub ⭐ 하나 눌러주세요. 다른 사람이 이 도구를 찾는 데 도움이 됩니다.
 
@@ -100,27 +100,29 @@ await writeFile("보고서.hwpx", Buffer.from(hwpx))
 
 ## 검증 결과
 
-**4.21.15 · 2026-10-10 배포 검증**. 고정 코퍼스와 참조 기준으로 측정한 결과입니다.
+**4.21.17 · 2026-10-11 배포 검증**. 고정 코퍼스와 참조 기준으로 측정한 결과입니다.
 
 | 대상 | 규모 | 결과 |
 | --- | --- | --- |
 | HWPX | 2,424문서 · 10,342표 | 표 구조 일치 **10,342/10,342 · 100%** |
 | HWP 5.x ↔ HWPX | 1,130쌍 · 4,315표 | 짝 문서의 표 구조 일치 **100%** |
-| PDF 글 | 744쌍 | 글자 재현율 **99.84%** · 정확도 **99.64%** · 읽기 순서 **99.16%** · 어절 F1 **98.98%** |
+| PDF 글 | 744쌍 | 글자 재현율 **99.84%** · 정확도 **99.64%** · 읽기 순서 **99.16%** · 어절 F1 **98.99%** |
 | PDF 표 | 708쌍 · 2,331표 | 표 탐지 **99.83%** · 구조 일치 **97.98%** · 칸 F1 **0.991** |
 | PDF 전체 | 1,911문서 중 1,729건 채점 | 글 커버리지 **99.80%** |
 | OCR | 53문서 · 102쪽 | CER **0.0412** · 글자 재현율 **99.02%** |
 
 표 구조 점수와 셀 내용·화면 재현은 별도 지표입니다. 채점 범위·제외 기준·전 벤치 문서별 비교는 [벤치마크 상세](docs/benchmarks.md)에 있습니다.
 
-외부 PDF 벤치 **opendataloader-bench 200문서**는 4.21.15 배포 측정에서 기본값 종합 **0.963**, OCR 끔 **0.939**입니다. [옵션별 결과와 재현 방법](docs/benchmarks.md#pdf--markdown--opendataloader-bench)
+외부 PDF 벤치 **opendataloader-bench 200문서**는 4.21.17 배포 측정에서 기본값 종합 **0.963**, OCR 끔 **0.939**입니다. [옵션별 결과와 재현 방법](docs/benchmarks.md#pdf--markdown--opendataloader-bench)
 
 ## 최근 업데이트
 
-최신 배포: **[4.21.15](https://github.com/chrisryugj/kordoc/releases/tag/v4.21.15)** · 2026-10-10
+최신 배포: **[4.21.17](https://github.com/chrisryugj/kordoc/releases/tag/v4.21.17)** · 2026-10-11
 
 | 버전 | 주요 변경 |
 | --- | --- |
+| 4.21.17 | PDF 공백 없는 꽉 찬 줄 끝 + 영문 괄호 — "공적개발원조⏎(ODA)"·"하네스⏎(safety …)" 를 붙여 씀 (PDF 어절 F1 98.98 → 98.99%) |
+| 4.21.16 | PDF 표 쪽 넘김 갈린 행의 수량 이어짐 — 머리 칸 "30%" / 다음 쪽 "미만" 을 한 칸으로; 공문서 autoFit.safety(다른 조판기용 한 줄 판정 여유) (법령 별표 PDF 표 97.40 → 97.69%) |
 | 4.21.15 | PDF 줄 끝 공백 글리프로 어절 경계 꺾임 가르기 — 원문이 띄운 줄 꺾임("이루어⏎진다면")을 붙이던 것 (PDF 어절 F1 98.89 → 98.98%) |
 | 4.21.14 | 한컴 PDF 유니코드 없는 기호 되살리기 — 자동 글머리표·괄호(U+F000)를 글꼴 윤곽 모양으로 "▸"·"□"·"《》"·"↓" 등으로; 텍스트층 커버리지 채점 기준 변경 (PDF 어절 F1 98.87 → 98.89%·글 재현율 99.83 → 99.84%) |
 | 4.21.13 | PDF 법령 별표 쪽 넘김 행 — 번호 열이 다음 차례면 새 행, "가)" 항목도 다음 항목 머리로, 쪽을 넘는 긴 칸은 두 행을 덮는 한 칸으로; parse-worker 가 닫힌 stdout 에서 조용히 끝남(#142) (법령 별표 PDF 표 구조 일치 96.82 → 97.40%, 문서별 하락 0) |

package/dist/{-HYTPLFLG.js → -GS6LEDZ6.js}

@@ -54,17 +54,17 @@ import {
   unknownFontWarnings,
   usesGaejosikMunche,
   validateHwpx
-} from "./chunk-HIXDQ43L.js";
+} from "./chunk-6F6L63RG.js";
 import {
   blocksToPages
-} from "./chunk-VFGH2R4P.js";
+} from "./chunk-L3VCBSFW.js";
 import {
   applySplices,
   buildParagraphSplices,
   buildRangeSplices,
   placeSealHwpx,
   scanSectionXml
-} from "./chunk-7JCSAOTM.js";
+} from "./chunk-7ASI3XKM.js";
 import {
   DEFAULT_REDACT_RULES,
   redactMarkdown,
@@ -72,7 +72,7 @@ import {
 } from "./chunk-3WSNLH4B.js";
 import {
   blocksToChunks
-} from "./chunk-ER3N3QPL.js";
+} from "./chunk-FWQHYJZ2.js";
 import {
   SPACE_EM_FIXED,
   SPACE_EM_FONT,
@@ -97,20 +97,20 @@ import {
   renderSceneToHtml,
   simulateWrap,
   simulateWrapKeepWord
-} from "./chunk-SQFCH4AY.js";
+} from "./chunk-FMUBLLIZ.js";
 import "./chunk-7FCNOKOR.js";
-import "./chunk-M4LXFLAV.js";
+import "./chunk-TZU7PXV2.js";
 import "./chunk-VQFA55IC.js";
-import "./chunk-OUBHY6W3.js";
+import "./chunk-Q4GXF444.js";
 import {
   blocksToMarkdown,
   flattenLayoutTables,
   hasStructuredCellContent
-} from "./chunk-65GJ2MMG.js";
-import "./chunk-FGC3EEQD.js";
+} from "./chunk-QLL7VG6L.js";
+import "./chunk-TUSREHO3.js";
 import {
   VERSION
-} from "./chunk-NHMRAR3I.js";
+} from "./chunk-PEQ5K2JO.js";
 export {
   AGENCY_COMMON_NAME,
   BUILTIN_TEMPLATES,
@@ -204,4 +204,4 @@ export {
   usesGaejosikMunche,
   validateHwpx
 };
-//# sourceMappingURL=-HYTPLFLG.js.map
+//# sourceMappingURL=-GS6LEDZ6.js.map

package/dist/{chunk-ZCGVIGG5.js → chunk-3PYAUHWY.js}

@@ -3,7 +3,7 @@ import {
   ensureModelsIn,
   getModelStatusIn,
   getModelsDir
-} from "./chunk-6CVYB634.js";
+} from "./chunk-J2FMAYNJ.js";
 
 // src/ocr/models.ts
 import { join } from "path";
@@ -93,4 +93,4 @@ export {
   parseCharacterDict,
   ocrModelPath
 };
-//# sourceMappingURL=chunk-ZCGVIGG5.js.map
+//# sourceMappingURL=chunk-3PYAUHWY.js.map

package/dist/{chunk-HIXDQ43L.js → chunk-6F6L63RG.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   blocksToPages
-} from "./chunk-VFGH2R4P.js";
+} from "./chunk-L3VCBSFW.js";
 import {
   allLinesegRemovalSplices,
   applySplices,
@@ -16,7 +16,7 @@ import {
   patchZipEntries,
   readZipEntries,
   scanSectionXml
-} from "./chunk-7JCSAOTM.js";
+} from "./chunk-7ASI3XKM.js";
 import {
   CONVERT_MAP,
   FLAG_COMPRESSED,
@@ -67,7 +67,7 @@ import {
   romanNumeral,
   simulateWrap,
   takeTocLeaders
-} from "./chunk-SQFCH4AY.js";
+} from "./chunk-FMUBLLIZ.js";
 import {
   inlineImagesIntoMarkdown
 } from "./chunk-7FCNOKOR.js";
@@ -82,7 +82,7 @@ import {
   tidyScriptTags,
   unframeLayoutTables,
   wrapScript
-} from "./chunk-M4LXFLAV.js";
+} from "./chunk-TZU7PXV2.js";
 import {
   parsePageRange
 } from "./chunk-VQFA55IC.js";
@@ -102,7 +102,7 @@ import {
   localName,
   rawTextContent,
   readZipEntry
-} from "./chunk-OUBHY6W3.js";
+} from "./chunk-Q4GXF444.js";
 import {
   MAX_COLS,
   MAX_ROWS,
@@ -117,7 +117,7 @@ import {
   hasStructuredCellContent,
   mapPuaText,
   noteSuffix
-} from "./chunk-65GJ2MMG.js";
+} from "./chunk-QLL7VG6L.js";
 import {
   KordocError,
   classifyError,
@@ -133,7 +133,7 @@ import {
   toArrayBuffer,
   unescapeHtml,
   unzipLimitBytes
-} from "./chunk-NHMRAR3I.js";
+} from "./chunk-PEQ5K2JO.js";
 
 // src/table/visual.ts
 import { readFile as readFile2 } from "fs/promises";
@@ -20517,12 +20517,13 @@ function buildTitleBox(title, titleCharPr, barCharPr, bodyWidth, reg) {
 // src/hwpx/fit-line.ts
 var RATIOS = [100, 97, 95, 92, 90, 87, 85];
 var SPACINGS = [0, -3, -5];
-function safety(faceClass, approx) {
-  return faceClass.startsWith("font:") ? 0.98 : approx;
+function safety(faceClass, approx, limit) {
+  const k = faceClass.startsWith("font:") ? 0.98 : approx;
+  return limit === void 0 ? k : Math.min(k, limit);
 }
-function fitOneLine(text, font, pt, availHu, minPt = Math.max(pt - 3, 10), minRatio = 85) {
+function fitOneLine(text, font, pt, availHu, minPt = Math.max(pt - 3, 10), minRatio = 85, safetyLimit) {
   const faceClass = faceClassForGen(font);
-  const avail = availHu * safety(faceClass, 0.98);
+  const avail = availHu * safety(faceClass, 0.98, safetyLimit);
   const fits = (p, r, s) => measureTextWidth(text, p * 100, r, { spacingPct: s, faceClass }) <= avail;
   if (fits(pt, 100, 0)) return { pt, ratio: 100, spacing: 0, overflow: false };
   const ratios = RATIOS.filter((r) => r >= minRatio);
@@ -20557,10 +20558,10 @@ function worstLooseness(text, starts, firstW, contW, h, ratio, spacing, faceClas
   }
   return worst;
 }
-function fitParagraph(text, font, pt, firstW, contW, minRatio = 88, orphanRatio = ORPHAN_RATIO) {
+function fitParagraph(text, font, pt, firstW, contW, minRatio = 88, orphanRatio = ORPHAN_RATIO, safetyLimit) {
   const faceClass = faceClassForGen(font);
   const h = pt * 100;
-  const k = safety(faceClass, 0.95);
+  const k = safety(faceClass, 0.95, safetyLimit);
   const f = firstW * k, c = contW * k;
   const ladder = [[100, 0], ...SQUEEZE.filter(([r]) => r >= minRatio)];
   const base = simulateWrap(text, f, c, h, 100, "keep", { faceClass });
@@ -20614,10 +20615,10 @@ function fitParagraph(text, font, pt, firstW, contW, minRatio = 88, orphanRatio
   return best && (best.r !== 100 || best.sp !== 0) ? { ratio: best.r, spacing: best.sp } : null;
 }
 var LIST_BREAK = /* @__PURE__ */ new Set([..."\xB7,\u3001/"]);
-function fitCharBreaks(text, font, pt, firstW, contW, minRatio = 88) {
+function fitCharBreaks(text, font, pt, firstW, contW, minRatio = 88, safetyLimit) {
   const faceClass = faceClassForGen(font);
   const h = pt * 100;
-  const k = safety(faceClass, 0.95);
+  const k = safety(faceClass, 0.95, safetyLimit);
   if (!text.split(/ +/).some((w) => measureTextWidth(w, h, 100, { faceClass }) > contW * k)) return null;
   const prepared = prepareWrap(text, "charAll", { faceClass });
   let best = null;
@@ -21270,6 +21271,9 @@ function validateGongmunOptions(opts) {
   if (typeof opts.autoFit === "object" && opts.autoFit.minRatio !== void 0) {
     assertFiniteRange("autoFit.minRatio", opts.autoFit.minRatio, 50, 99);
   }
+  if (typeof opts.autoFit === "object" && opts.autoFit.safety !== void 0) {
+    assertFiniteRange("autoFit.safety", opts.autoFit.safety, 0.8, 1);
+  }
   if (opts.margins) {
     for (const side of ["top", "bottom", "left", "right"]) {
       const value = opts.margins[side];
@@ -21387,6 +21391,7 @@ function resolveGongmun(input) {
     margins: opts.margins ?? (ministry ? MINISTRY_MARGINS : preset === "report" || preset === "plan" || preset === "bangchim" ? SEOUL_REPORT_MARGINS : reportFamily ? GAEJOSIK_MARGINS : OFFICIAL_MARGINS),
     centerTitle: opts.centerTitle ?? true,
     autoFitMinRatio,
+    ...typeof opts.autoFit === "object" && opts.autoFit.safety !== void 0 ? { autoFitSafety: opts.autoFit.safety } : {},
     // 보도자료는 머리박스가 1페이지 최상단을 차지하는 서식이라 표지·목차와 양립 불가 —
     // 켜면 머리박스가 표지에 얹히고 25pt 제목·부제가 유실된다 (docHead 프리셋 게이팅과 동일 관례)
     cover: coverOn && preset !== "press" ? { date: coverOpts.date ?? null, org: coverOpts.org ?? "", ...coverOpts.dept ? { dept: coverOpts.dept } : {}, ...coverOpts.label ? { label: coverOpts.label } : {} } : null,
@@ -26104,7 +26109,7 @@ function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
     const textW2 = W2 - lay.left - lay.hang;
     let ratio = 100, spacing = 0, charBreaks = false;
     if (st.oneLine) {
-      const fit = fitOneLine(plain2(node.text), st.font, st.pt, textW2, st.pt, 90);
+      const fit = fitOneLine(plain2(node.text), st.font, st.pt, textW2, st.pt, 90, g.autoFitSafety);
       if (fit.overflow) warnings.push(`\u25A1 \uD56D\uBAA9\uC774 \uD55C \uC904\uC5D0 \uB2F4\uAE30\uC9C0 \uC54A\uC544 \uB450 \uC904\uB85C \uAEBE\uC785\uB2C8\uB2E4 \u2014 \uBB38\uC7A5\uC744 \uC904\uC774\uC138\uC694(\uC7A5\uD3C9 90%\xB7\uC790\uAC04 -5 \uB85C\uB3C4 \uCD08\uACFC): "${node.text.slice(0, 30)}\u2026"`);
       else {
         ratio = fit.ratio;
@@ -26112,8 +26117,8 @@ function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
       }
     } else {
       const t = plain2(node.text);
-      const cb = fitCharBreaks(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio ?? 101);
-      const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio) : null);
+      const cb = fitCharBreaks(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio ?? 101, g.autoFitSafety);
+      const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio, void 0, g.autoFitSafety) : null);
       if (f) {
         ratio = f.ratio;
         spacing = f.spacing;
@@ -26142,8 +26147,8 @@ function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
     const left = isMinistry ? st.leadTa * taHu(st.pt) : leadLeft(node.depth, st.pt);
     const textW2 = W2 - left - lay.hang;
     const t = plain2(node.text);
-    const cb = fitCharBreaks(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio ?? 101);
-    const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio) : null);
+    const cb = fitCharBreaks(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio ?? 101, g.autoFitSafety);
+    const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, textW2, textW2, g.autoFitMinRatio, void 0, g.autoFitSafety) : null);
     const base = { font: st.font, pt: st.pt, bold: st.bold, ratio: f?.ratio ?? 100, spacing: f?.spacing ?? 0 };
     const paraId = reg.para({ align: st.align ?? "JUSTIFY", left, indent: -lay.hang, after: isMinistry ? 400 : 0, lineSp: isMinistry ? 130 : st.lineSp ?? scheme.lineSp, keepWord: !cb, autoTab: true, widowOrphan: true });
     const markerRun = markerRunXml(marker, reg.char({ font: st.font, pt: st.pt, bold: st.bold }), lay);
@@ -26172,8 +26177,8 @@ function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
     let ratio = 100, spacing = 0, charBreaks = false;
     if (!node.align) {
       const t = plain2(node.text);
-      const cb = fitCharBreaks(t, st.font, st.pt, W2, W2, g.autoFitMinRatio ?? 101);
-      const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, W2, W2, g.autoFitMinRatio) : null);
+      const cb = fitCharBreaks(t, st.font, st.pt, W2, W2, g.autoFitMinRatio ?? 101, g.autoFitSafety);
+      const f = cb ?? (g.autoFitMinRatio !== null ? fitParagraph(t, st.font, st.pt, W2, W2, g.autoFitMinRatio, void 0, g.autoFitSafety) : null);
       if (f) {
         ratio = f.ratio;
         spacing = f.spacing;
@@ -26215,7 +26220,7 @@ function buildGongmunSectionV5(blocks, gongmun, deps, theme) {
     }
     const label = chapterStyle === "none" ? "" : chapterLabel(node.index, chapterStyle === "number" ? "number" : "roman");
     const text = label ? `${label} ${node.text}` : node.text;
-    const fit = fitOneLine(plain2(text), st.font, st.pt, W2, st.pt - 2);
+    const fit = fitOneLine(plain2(text), st.font, st.pt, W2, st.pt - 2, void 0, g.autoFitSafety);
     const base = { font: st.font, pt: fit.pt, bold: st.bold, ratio: fit.ratio, spacing: fit.spacing };
     const before = prevKind === "start" || prevKind === "title" || prevKind === "summary" ? 0 : lineHu(scheme.body);
     const paraId = reg.para({ align: "LEFT", before, after: Math.round(lineHu(scheme.body) * 0.3), lineSp: st.lineSp ?? scheme.lineSp, keepWithNext: true });
@@ -30754,7 +30759,7 @@ function scriptsOff(r, off) {
 }
 async function parseImage(buffer, options) {
   try {
-    const { parseImageDocument } = await import("./image-ocr-DV5NNNZW.js");
+    const { parseImageDocument } = await import("./image-ocr-HLEL6TED.js");
     const { blocks, warnings } = await parseImageDocument(buffer, options);
     stripScriptTags({ blocks });
     return {
@@ -30814,7 +30819,7 @@ async function parseHwp(buffer, options) {
 async function parsePdf(buffer, options) {
   let parsePdfDocument;
   try {
-    const mod = await import("./parser-FSHPAFTU.js");
+    const mod = await import("./parser-WR6UBRI4.js");
     parsePdfDocument = mod.parsePdfDocument;
   } catch {
     return {
@@ -31044,4 +31049,4 @@ export {
   parseHwpml,
   fillForm
 };
-//# sourceMappingURL=chunk-HIXDQ43L.js.map
+//# sourceMappingURL=chunk-6F6L63RG.js.map