kordoc 4.21.21 → 4.21.23

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**. [4.21.21 검증](docs/benchmarks.md)
+- 🇰🇷 **한국 공문서 표 구조 100%** — 원본 HWPX 2,424문서의 보이는 표 **10,342개 전부 구조 일치**. [4.21.22 검증](docs/benchmarks.md)
 
 쓸모 있었다면 GitHub ⭐ 하나 눌러주세요. 다른 사람이 이 도구를 찾는 데 도움이 됩니다.
 
@@ -100,7 +100,7 @@ await writeFile("보고서.hwpx", Buffer.from(hwpx))
 
 ## 검증 결과
 
-**4.21.21 · 2026-10-11 배포 검증**. 고정 코퍼스와 참조 기준으로 측정한 결과입니다.
+**4.21.22 · 2026-10-11 배포 검증**. 고정 코퍼스와 참조 기준으로 측정한 결과입니다.
 
 | 대상 | 규모 | 결과 |
 | --- | --- | --- |
@@ -113,14 +113,16 @@ await writeFile("보고서.hwpx", Buffer.from(hwpx))
 
 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 4.21.21 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 4.21.22 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: **[4.21.21](https://github.com/chrisryugj/kordoc/releases/tag/v4.21.21)** · 2026-10-11
+최신 배포: **[4.21.23](https://github.com/chrisryugj/kordoc/releases/tag/v4.21.23)** · 2026-10-11
 
 | 버전 | 주요 변경 |
 | --- | --- |
+| 4.21.23 | HWP·HWPX 삽입 그림 속 글 OCR — `ocr: true` 일 때 그림 블록 뒤에 그림 속 글(#143, 기본 산출 불변) |
+| 4.21.22 | PDF 양쪽 정렬 영문 참고문헌 줄을 열로 읽어 낱말이 뒤섞이던 것 고침 (web052 읽기 순서 → 1, score 커버리지 2문서 → 1) |
 | 4.21.21 | PDF 쪽 테두리 틀 안 바닥글 걷기 · 첨부 이름표("<붙임2>"·"<별지 서식 제3호>")·강의 번호를 머리말로 지우던 것 고침 (PDF 글 정확도 99.64 → 99.67%, 어절 F1 99.00 → 99.02%) |
 | 4.21.20 | PDF 괘선 없는 표 칸의 어절 가운데 꺾임을 붙여 씀("남⏎편이") · 채우기·패치 서식 보존(바뀐 문단만 줄 배치 캐시 삭제·강조 표지·균등 띄어쓰기·라벨 칸 방향) (PDF 어절 F1 98.987 → 98.998%) |
 | 4.21.19 | PDF 위아래로 쌓인 표와 사이 글줄을 한 표로 묶던 괘선 격자 버리기 — 지역 머리·각주가 제 표와 떨어지던 것 (PDF 읽기 순서 99.16 → 99.23%) |

package/dist/{-ZYF6HX4M.js → -FLEBKE6V.js}

@@ -54,25 +54,10 @@ import {
   unknownFontWarnings,
   usesGaejosikMunche,
   validateHwpx
-} from "./chunk-YM3K3LCD.js";
-import {
-  blocksToPages
-} from "./chunk-2ZGP7A5T.js";
-import {
-  applySplices,
-  buildParagraphSplices,
-  buildRangeSplices,
-  placeSealHwpx,
-  scanSectionXml
-} from "./chunk-HWIW3WK2.js";
-import {
-  DEFAULT_REDACT_RULES,
-  redactMarkdown,
-  redactText
-} from "./chunk-3WSNLH4B.js";
+} from "./chunk-4F6OVO2A.js";
 import {
   blocksToChunks
-} from "./chunk-7YAJCXCD.js";
+} from "./chunk-7MTKOSAF.js";
 import {
   SPACE_EM_FIXED,
   SPACE_EM_FONT,
@@ -97,20 +82,35 @@ import {
   renderSceneToHtml,
   simulateWrap,
   simulateWrapKeepWord
-} from "./chunk-QHNVH4TC.js";
-import "./chunk-7FCNOKOR.js";
-import "./chunk-PTF4NCFM.js";
+} from "./chunk-2S37CQUJ.js";
+import "./chunk-RKWJMA2B.js";
+import {
+  blocksToPages
+} from "./chunk-JLG7KK4M.js";
+import "./chunk-VKVDMH7X.js";
 import "./chunk-VQFA55IC.js";
-import "./chunk-BG6LUCQL.js";
+import "./chunk-YA6UGARL.js";
 import {
   blocksToMarkdown,
   flattenLayoutTables,
   hasStructuredCellContent
-} from "./chunk-RBRHO7WL.js";
-import "./chunk-YHPJTLCA.js";
+} from "./chunk-GPIMKPEI.js";
+import {
+  applySplices,
+  buildParagraphSplices,
+  buildRangeSplices,
+  placeSealHwpx,
+  scanSectionXml
+} from "./chunk-JA4ZWCYL.js";
+import "./chunk-U6D7LUCR.js";
 import {
   VERSION
-} from "./chunk-FZMGF6XU.js";
+} from "./chunk-TAU45SJK.js";
+import {
+  DEFAULT_REDACT_RULES,
+  redactMarkdown,
+  redactText
+} from "./chunk-3WSNLH4B.js";
 export {
   AGENCY_COMMON_NAME,
   BUILTIN_TEMPLATES,
@@ -204,4 +204,4 @@ export {
   usesGaejosikMunche,
   validateHwpx
 };
-//# sourceMappingURL=-ZYF6HX4M.js.map
+//# sourceMappingURL=-FLEBKE6V.js.map

package/dist/{chunk-QHNVH4TC.js → chunk-2S37CQUJ.js}

@@ -1,7 +1,14 @@
 #!/usr/bin/env node
+import {
+  loadSharp,
+  rasterizePageSvg
+} from "./chunk-RKWJMA2B.js";
 import {
   inlineImagesIntoMarkdown
-} from "./chunk-7FCNOKOR.js";
+} from "./chunk-VKVDMH7X.js";
+import {
+  parsePageRange
+} from "./chunk-VQFA55IC.js";
 import {
   CELL_EDGES,
   HEADING_RATIO_H1,
@@ -9,20 +16,7 @@ import {
   HEADING_RATIO_H3,
   tidyScriptTags,
   unframeLayoutTables
-} from "./chunk-PTF4NCFM.js";
-import {
-  parsePageRange
-} from "./chunk-VQFA55IC.js";
-import {
-  DEFAULT_CHAR,
-  DEFAULT_PARA_GEOM,
-  MAX_DECOMPRESS_SIZE,
-  MAX_ZIP_ENTRIES,
-  createXmlParser,
-  findChildByLocalName,
-  hwpFaceToCssStack,
-  parseRenderStyles
-} from "./chunk-BG6LUCQL.js";
+} from "./chunk-YA6UGARL.js";
 import {
   MAX_COLS,
   MAX_ROWS,
@@ -33,18 +27,24 @@ import {
   flattenLayoutTables,
   mapPuaText,
   markNonLayoutTable
-} from "./chunk-RBRHO7WL.js";
+} from "./chunk-GPIMKPEI.js";
 import {
-  loadSharp,
-  rasterizePageSvg
-} from "./chunk-YHPJTLCA.js";
+  DEFAULT_CHAR,
+  DEFAULT_PARA_GEOM,
+  MAX_DECOMPRESS_SIZE,
+  MAX_ZIP_ENTRIES,
+  createXmlParser,
+  findChildByLocalName,
+  hwpFaceToCssStack,
+  parseRenderStyles
+} from "./chunk-U6D7LUCR.js";
 import {
   KordocError,
   escapeHtml,
   precheckZipSize,
   sanitizeHref,
   toArrayBuffer
-} from "./chunk-FZMGF6XU.js";
+} from "./chunk-TAU45SJK.js";
 
 // src/render/svg-render.ts
 import JSZip from "jszip";
@@ -7804,4 +7804,4 @@ export {
   cropRect,
   extractRenderedRegions
 };
-//# sourceMappingURL=chunk-QHNVH4TC.js.map
+//# sourceMappingURL=chunk-2S37CQUJ.js.map

package/dist/{chunk-YM3K3LCD.js → chunk-4F6OVO2A.js}

@@ -1,23 +1,4 @@
 #!/usr/bin/env node
-import {
-  blocksToPages
-} from "./chunk-2ZGP7A5T.js";
-import {
-  allLinesegRemovalSplices,
-  applySplices,
-  buildParagraphSplices,
-  buildRangeSplices,
-  changedLinesegRemovalSplices,
-  decodeXmlEntities,
-  escapeXmlText,
-  findElementEnd,
-  markerRunSplices,
-  paraTText,
-  paraTextPureT,
-  patchZipEntries,
-  readZipEntries,
-  scanSectionXml
-} from "./chunk-HWIW3WK2.js";
 import {
   CONVERT_MAP,
   FLAG_COMPRESSED,
@@ -68,10 +49,17 @@ import {
   romanNumeral,
   simulateWrap,
   takeTocLeaders
-} from "./chunk-QHNVH4TC.js";
+} from "./chunk-2S37CQUJ.js";
 import {
+  blocksToPages
+} from "./chunk-JLG7KK4M.js";
+import {
+  bmpToPng,
   inlineImagesIntoMarkdown
-} from "./chunk-7FCNOKOR.js";
+} from "./chunk-VKVDMH7X.js";
+import {
+  parsePageRange
+} from "./chunk-VQFA55IC.js";
 import {
   CELL_EDGES,
   HEADING_RATIO_H1,
@@ -83,10 +71,38 @@ import {
   tidyScriptTags,
   unframeLayoutTables,
   wrapScript
-} from "./chunk-PTF4NCFM.js";
+} from "./chunk-YA6UGARL.js";
 import {
-  parsePageRange
-} from "./chunk-VQFA55IC.js";
+  MAX_COLS,
+  MAX_ROWS,
+  MAX_TABLE_CELLS,
+  blocksToMarkdown,
+  buildTable,
+  convertTableToText,
+  escapeGfm,
+  escapeHtmlCellText,
+  escapeLiteralDollar,
+  flattenLayoutTables,
+  hasStructuredCellContent,
+  mapPuaText,
+  noteSuffix
+} from "./chunk-GPIMKPEI.js";
+import {
+  allLinesegRemovalSplices,
+  applySplices,
+  buildParagraphSplices,
+  buildRangeSplices,
+  changedLinesegRemovalSplices,
+  decodeXmlEntities,
+  escapeXmlText,
+  findElementEnd,
+  markerRunSplices,
+  paraTText,
+  paraTextPureT,
+  patchZipEntries,
+  readZipEntries,
+  scanSectionXml
+} from "./chunk-JA4ZWCYL.js";
 import {
   MAX_DECOMPRESS_SIZE,
   MAX_XML_DEPTH,
@@ -103,22 +119,7 @@ import {
   localName,
   rawTextContent,
   readZipEntry
-} from "./chunk-BG6LUCQL.js";
-import {
-  MAX_COLS,
-  MAX_ROWS,
-  MAX_TABLE_CELLS,
-  blocksToMarkdown,
-  buildTable,
-  convertTableToText,
-  escapeGfm,
-  escapeHtmlCellText,
-  escapeLiteralDollar,
-  flattenLayoutTables,
-  hasStructuredCellContent,
-  mapPuaText,
-  noteSuffix
-} from "./chunk-RBRHO7WL.js";
+} from "./chunk-U6D7LUCR.js";
 import {
   KordocError,
   classifyError,
@@ -134,7 +135,7 @@ import {
   toArrayBuffer,
   unescapeHtml,
   unzipLimitBytes
-} from "./chunk-FZMGF6XU.js";
+} from "./chunk-TAU45SJK.js";
 
 // src/table/visual.ts
 import { readFile as readFile2 } from "fs/promises";
@@ -19952,6 +19953,66 @@ function collectAllTables(scan) {
   return out;
 }
 
+// src/ocr/embedded-image-ocr.ts
+var MIN_SIDE = 24;
+async function ocrEmbeddedImages(blocks, options, warnings) {
+  const ocr = options?.ocr;
+  if (ocr !== true && ocr !== "force" && typeof ocr !== "function") return false;
+  const { parseImageDocument } = await import("./image-ocr-BKRQVMW6.js");
+  let inserted = false, unreadable = 0;
+  let failed = null;
+  const walk = async (list, depth) => {
+    if (depth > 64) return;
+    for (let i = 0; i < list.length; i++) {
+      const b = list[i];
+      if (b.children) await walk(b.children, depth + 1);
+      if (b.table) {
+        for (const row of b.table.cells) for (const cell2 of row) if (cell2.blocks) await walk(cell2.blocks, depth + 1);
+      }
+      if (b.type !== "image" || !b.imageData || failed) continue;
+      const bytes = ocrBytes(b.imageData);
+      if (!bytes) {
+        unreadable++;
+        continue;
+      }
+      if (await tooSmall(bytes)) continue;
+      let found;
+      try {
+        const buf = bytes.buffer.slice(bytes.byteOffset, bytes.byteOffset + bytes.byteLength);
+        found = (await parseImageDocument(buf, options)).blocks;
+      } catch (err) {
+        failed = err instanceof Error ? err.message : String(err);
+        continue;
+      }
+      if (!found.length) continue;
+      const placed = found.map((f) => ({ ...f, pageNumber: b.pageNumber, bbox: void 0 }));
+      list.splice(i + 1, 0, ...placed);
+      i += placed.length;
+      inserted = true;
+    }
+  };
+  await walk(blocks, 0);
+  if (unreadable) warnings.push({ code: "SKIPPED_IMAGE", message: `${unreadable}\uAC1C \uADF8\uB9BC\uC740 OCR \uB85C \uC77D\uC744 \uC218 \uC5C6\uB294 \uD615\uC2DD(WMF\xB7EMF \uB4F1)\uC774\uB77C \uAC74\uB108\uB700` });
+  if (failed) warnings.push({ code: "OCR_FAILED", message: `\uC0BD\uC785 \uADF8\uB9BC OCR \uC2E4\uD328 \u2014 \uADF8\uB9BC \uC18D \uAE00 \uC5C6\uC774 \uACC4\uC18D: ${failed}` });
+  return inserted;
+}
+function ocrBytes(img) {
+  const mime = img.mimeType.toLowerCase();
+  if (mime.includes("wmf") || mime.includes("emf")) return null;
+  if (mime.includes("bmp")) return bmpToPng(img.data);
+  return img.data;
+}
+async function tooSmall(bytes) {
+  try {
+    const mod = await import("sharp");
+    const sharp = typeof mod === "function" ? mod : mod.default;
+    const { width, height } = await sharp(Buffer.from(bytes)).metadata();
+    return !!width && !!height && (width < MIN_SIDE || height < MIN_SIDE);
+  } catch {
+    return false;
+  }
+}
+
 // src/table/classifier.ts
 var DIAGRAM_KEYWORDS = /조직도|비상\s*연락망|연락망|체계도|업무\s*체계도|기구표|배치도|흐름도|추진\s*체계/u;
 var KEYWORD_GATE = 0.3;
@@ -31094,12 +31155,24 @@ async function dispatch(format2, buffer, opts) {
       return { success: false, fileType: "unknown", error: "\uC9C0\uC6D0\uD558\uC9C0 \uC54A\uB294 \uD30C\uC77C \uD615\uC2DD\uC785\uB2C8\uB2E4.", code: "UNSUPPORTED_FORMAT" };
   }
 }
+async function withEmbeddedImageOcr(blocks, markdown, warnings, images, options) {
+  const list = [...warnings ?? []];
+  if (!await ocrEmbeddedImages(blocks, options, list)) return { markdown, warnings: list.length ? list : void 0 };
+  let md = blocksToMarkdown(blocks, { tableFormat: options?.tableFormat });
+  if (options?.inlineImages && options.images !== false && images?.length) {
+    try {
+      md = inlineImagesIntoMarkdown(md, images, { compress: true });
+    } catch {
+    }
+  }
+  return { markdown: md, warnings: list.length ? list : void 0 };
+}
 function scriptsOff(r, off) {
   return off ? stripScriptTags(r) : r;
 }
 async function parseImage(buffer, options) {
   try {
-    const { parseImageDocument } = await import("./image-ocr-ITNY454I.js");
+    const { parseImageDocument } = await import("./image-ocr-BKRQVMW6.js");
     const { blocks, warnings } = await parseImageDocument(buffer, options);
     stripScriptTags({ blocks });
     return {
@@ -31124,7 +31197,9 @@ async function parseHwp3(buffer, options) {
 }
 async function parseHwpx(buffer, options) {
   try {
-    const { markdown, blocks, metadata, outline, warnings, images } = await parseHwpxDocument(buffer, options);
+    const doc = await parseHwpxDocument(buffer, options);
+    const { blocks, metadata, outline, images } = doc;
+    const { markdown, warnings } = await withEmbeddedImageOcr(blocks, doc.markdown, doc.warnings, void 0, options);
     return scriptsOff({ success: true, fileType: "hwpx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
   } catch (err) {
     return { success: false, fileType: "hwpx", error: sanitizeError(err), code: classifyError(err) };
@@ -31151,7 +31226,8 @@ async function parseHwp(buffer, options) {
       } catch {
       }
     }
-    return scriptsOff({ success: true, fileType: "hwp", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
+    const ocred = await withEmbeddedImageOcr(blocks, markdown, warnings, images, options);
+    return scriptsOff({ success: true, fileType: "hwp", markdown: ocred.markdown, blocks, metadata, outline, warnings: ocred.warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
   } catch (err) {
     return { success: false, fileType: "hwp", error: sanitizeError(err), code: classifyError(err) };
   }
@@ -31159,7 +31235,7 @@ async function parseHwp(buffer, options) {
 async function parsePdf(buffer, options) {
   let parsePdfDocument;
   try {
-    const mod = await import("./parser-KFCY3NIN.js");
+    const mod = await import("./parser-M45SFXWF.js");
     parsePdfDocument = mod.parsePdfDocument;
   } catch {
     return {
@@ -31389,4 +31465,4 @@ export {
   parseHwpml,
   fillForm
 };
-//# sourceMappingURL=chunk-YM3K3LCD.js.map
+//# sourceMappingURL=chunk-4F6OVO2A.js.map