kordoc 4.0.6 → 4.0.8

package/dist/index.js

@@ -13,6 +13,7 @@ import {
   convertTableToText,
   dedupeRunningHeaders,
   flattenLayoutTables,
+  inlineImagesIntoMarkdown,
   isPathTraversal,
   mapPuaText,
   normalizeSectionHref,
@@ -20,7 +21,7 @@ import {
   sanitizeHref,
   stripDtd,
   toArrayBuffer
-} from "./chunk-HOGJ5ALY.js";
+} from "./chunk-OQPILS7B.js";
 import {
   parsePageRange
 } from "./chunk-GE43BE46.js";
@@ -2711,628 +2712,807 @@ function gongmunDepthFromIndent(para2, left, ctx) {
   return depth >= 1 && depth < 8 ? depth : null;
 }
 
-// src/hwpx/images.ts
-function imageExtToMime(ext) {
-  switch (ext.toLowerCase()) {
-    case "jpg":
-    case "jpeg":
-      return "image/jpeg";
-    case "png":
-      return "image/png";
-    case "gif":
-      return "image/gif";
-    case "bmp":
-      return "image/bmp";
-    case "tif":
-    case "tiff":
-      return "image/tiff";
-    case "wmf":
-      return "image/wmf";
-    case "emf":
-      return "image/emf";
-    case "svg":
-      return "image/svg+xml";
-    default:
-      return "application/octet-stream";
+// src/hwp5/record.ts
+import { inflateRawSync, inflateSync } from "zlib";
+var TAG_PARA_HEADER = 66;
+var TAG_PARA_TEXT = 67;
+var TAG_CHAR_SHAPE = 68;
+var TAG_CTRL_HEADER = 71;
+var TAG_LIST_HEADER = 72;
+var TAG_TABLE = 77;
+var TAG_SHAPE_COMPONENT = 76;
+var TAG_SHAPE_COMPONENT_PICTURE = 85;
+var TAG_SHAPE_COMPONENT_CONTAINER = 86;
+var TAG_EQEDIT = 88;
+var TAG_BIN_DATA = 18;
+var TAG_DOC_CHAR_SHAPE = 21;
+var TAG_NUMBERING = 23;
+var TAG_BULLET = 24;
+var TAG_DOC_PARA_SHAPE = 25;
+var TAG_DOC_STYLE = 26;
+var CHAR_LINE = 0;
+var CHAR_SECTION_BREAK = 10;
+var CHAR_PARA = 13;
+var CHAR_TAB = 9;
+var CHAR_HYPHEN = 30;
+var CHAR_NBSP = 31;
+var CHAR_FIXED_NBSP = 24;
+var CHAR_FIXED_WIDTH = 25;
+var FLAG_COMPRESSED = 1 << 0;
+var FLAG_ENCRYPTED = 1 << 1;
+var FLAG_DISTRIBUTION = 1 << 2;
+var FLAG_DRM = 1 << 4;
+var MAX_RECORDS = 5e5;
+function readRecords(data) {
+  const records = [];
+  let offset = 0;
+  while (offset + 4 <= data.length && records.length < MAX_RECORDS) {
+    const header = data.readUInt32LE(offset);
+    offset += 4;
+    const tagId = header & 1023;
+    const level = header >> 10 & 1023;
+    let size = header >> 20 & 4095;
+    if (size === 4095) {
+      if (offset + 4 > data.length) break;
+      size = data.readUInt32LE(offset);
+      offset += 4;
+    }
+    if (offset + size > data.length) break;
+    records.push({ tagId, level, size, data: data.subarray(offset, offset + size) });
+    offset += size;
   }
+  return records;
 }
-function mimeToExt(mime) {
-  if (mime.includes("jpeg")) return "jpg";
-  if (mime.includes("png")) return "png";
-  if (mime.includes("gif")) return "gif";
-  if (mime.includes("bmp")) return "bmp";
-  if (mime.includes("tiff")) return "tif";
-  if (mime.includes("wmf")) return "wmf";
-  if (mime.includes("emf")) return "emf";
-  if (mime.includes("svg")) return "svg";
-  return "bin";
-}
-function collectImageBlocks(blocks, out, ownerCell, depth = 0) {
-  if (depth > MAX_XML_DEPTH) return;
-  for (const block of blocks) {
-    if (block.type === "image") {
-      out.push({ block, ownerCell });
-    } else if (block.type === "table" && block.table) {
-      for (const row of block.table.cells) {
-        for (const cell2 of row) {
-          if (cell2.blocks?.length) collectImageBlocks(cell2.blocks, out, cell2, depth + 1);
-        }
-      }
+var MAX_DECOMPRESS_SIZE2 = 100 * 1024 * 1024;
+function decompressStream(data) {
+  const opts = { maxOutputLength: MAX_DECOMPRESS_SIZE2 };
+  if (data.length >= 2 && data[0] === 120) {
+    try {
+      return inflateSync(data, opts);
+    } catch {
     }
   }
+  return inflateRawSync(data, opts);
 }
-async function extractImagesFromZip(zip, blocks, decompressed, warnings) {
-  const images = [];
-  let imageIndex = 0;
-  const imageBlocks = [];
-  collectImageBlocks(blocks, imageBlocks);
-  const resolved = /* @__PURE__ */ new Map();
-  for (const { block, ownerCell } of imageBlocks) {
-    if (block.type !== "image" || !block.text) continue;
-    const ref = block.text;
-    let img = resolved.get(ref);
-    if (img === void 0) {
-      img = null;
-      const candidates = [
-        `BinData/${ref}`,
-        `Contents/BinData/${ref}`,
-        ref
-        // 절대 경로일 수도 있음
-      ];
-      let resolvedPath = null;
-      if (!ref.includes(".")) {
-        const prefixes = [`BinData/${ref}`, `Contents/BinData/${ref}`];
-        for (const prefix of prefixes) {
-          const match = zip.file(new RegExp(`^${prefix.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}\\.[a-zA-Z0-9]+$`));
-          if (match.length > 0) {
-            resolvedPath = match[0].name;
-            break;
-          }
+function parseFileHeader(data) {
+  if (data.length < 40) throw new KordocError("FileHeader\uAC00 \uB108\uBB34 \uC9E7\uC2B5\uB2C8\uB2E4 (\uCD5C\uC18C 40\uBC14\uC774\uD2B8)");
+  const sig = data.subarray(0, 32).toString("utf8").replace(/\0+$/, "");
+  return {
+    signature: sig,
+    versionMajor: data[35],
+    flags: data.readUInt32LE(36)
+  };
+}
+function readHwpString(data, offset) {
+  if (offset + 2 > data.length) return { value: "", next: data.length };
+  const len = data.readUInt16LE(offset);
+  const start = offset + 2;
+  const end = start + len * 2;
+  if (len === 0 || end > data.length) return { value: "", next: start };
+  return { value: data.subarray(start, end).toString("utf16le"), next: end };
+}
+function parseDocInfo(records) {
+  const charShapes = [];
+  const paraShapes = [];
+  const styles = [];
+  const binData = [];
+  const numberings = [];
+  const bullets = [];
+  for (const rec of records) {
+    if (rec.tagId === TAG_DOC_PARA_SHAPE && rec.data.length >= 4) {
+      const attr1 = rec.data.readUInt32LE(0);
+      const headType = attr1 >>> 23 & 3;
+      const paraLevel = attr1 >>> 25 & 7;
+      const numberingId = rec.data.length >= 32 ? rec.data.readUInt16LE(30) : 0;
+      paraShapes.push({ headType, paraLevel, numberingId });
+    }
+    if (rec.tagId === TAG_BIN_DATA && rec.data.length >= 2) {
+      const attr = rec.data.readUInt16LE(0);
+      const typeBits = attr & 15;
+      if (typeBits === 0) {
+        binData.push({ kind: "link", storageId: 0, extension: "" });
+      } else {
+        const storageId = rec.data.length >= 4 ? rec.data.readUInt16LE(2) : 0;
+        const { value: extension } = readHwpString(rec.data, 4);
+        binData.push({ kind: typeBits === 2 ? "storage" : "embed", storageId, extension });
+      }
+    }
+    if (rec.tagId === TAG_NUMBERING && rec.data.length >= 14) {
+      const levelFormats = [];
+      const numberFormats = [];
+      const startNumbers = [1, 1, 1, 1, 1, 1, 1];
+      let offset = 0;
+      for (let level = 0; level < 7; level++) {
+        if (offset + 12 > rec.data.length) {
+          levelFormats.push("");
+          numberFormats.push(0);
+          continue;
         }
+        const attr = rec.data.readUInt32LE(offset);
+        numberFormats.push(attr >>> 5 & 15);
+        offset += 12;
+        const { value, next } = readHwpString(rec.data, offset);
+        levelFormats.push(value);
+        offset = next;
       }
-      const allCandidates = resolvedPath ? [resolvedPath, ...candidates] : candidates;
-      for (const path of allCandidates) {
-        if (isPathTraversal(path)) continue;
-        const file = zip.file(path);
-        if (!file) continue;
-        try {
-          const data = await file.async("uint8array");
-          decompressed.total += data.length;
-          if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
-          const ext = path.includes(".") ? path.split(".").pop() || "png" : "png";
-          const mimeType = imageExtToMime(ext);
-          imageIndex++;
-          const filename = `image_${String(imageIndex).padStart(3, "0")}.${mimeToExt(mimeType)}`;
-          img = { filename, data, mimeType };
-          images.push(img);
-          break;
-        } catch (err) {
-          if (err instanceof KordocError) throw err;
+      let baseStart = 1;
+      if (offset + 2 <= rec.data.length) {
+        baseStart = rec.data.readUInt16LE(offset) || 1;
+        offset += 2;
+      }
+      for (let level = 0; level < 7; level++) {
+        if (offset + 4 <= rec.data.length) {
+          startNumbers[level] = rec.data.readUInt32LE(offset) || 1;
+          offset += 4;
+        } else {
+          startNumbers[level] = baseStart;
         }
       }
-      if (!img) warnings?.push({ page: block.pageNumber, message: `\uC774\uBBF8\uC9C0 \uD30C\uC77C \uC5C6\uC74C: ${ref}`, code: "SKIPPED_IMAGE" });
-      resolved.set(ref, img);
+      numberings.push({ levelFormats, numberFormats, startNumbers });
     }
-    if (!img) {
-      block.type = "paragraph";
-      block.text = `[\uC774\uBBF8\uC9C0: ${ref}]`;
-      if (ownerCell) ownerCell.text = ownerCell.text.replace(`![image](${ref})`, `[\uC774\uBBF8\uC9C0: ${ref}]`);
-      continue;
+    if (rec.tagId === TAG_BULLET && rec.data.length >= 14) {
+      const code = rec.data.readUInt16LE(12);
+      bullets.push({ char: code > 0 ? String.fromCharCode(code) : "\u2022" });
     }
-    block.text = img.filename;
-    block.imageData = { data: img.data, mimeType: img.mimeType, filename: ref };
-    if (ownerCell) ownerCell.text = ownerCell.text.replace(`![image](${ref})`, `![image](${img.filename})`);
-  }
-  return images;
-}
-
-// src/hwpx/metadata.ts
-import JSZip2 from "jszip";
-
-// src/hwpx/zip-sections.ts
-import { inflateRawSync } from "zlib";
-function extractFromBrokenZip(buffer) {
-  const data = new Uint8Array(buffer);
-  const view = new DataView(buffer);
-  let pos = 0;
-  const blocks = [];
-  const warnings = [
-    { code: "BROKEN_ZIP_RECOVERY", message: "\uC190\uC0C1\uB41C ZIP \uAD6C\uC870 \u2014 Local File Header \uAE30\uBC18 \uBCF5\uAD6C \uBAA8\uB4DC" }
-  ];
-  let totalDecompressed = 0;
-  let entryCount = 0;
-  let sectionNum = 0;
-  const shared = createSectionShared();
-  while (pos < data.length - 30) {
-    if (data[pos] !== 80 || data[pos + 1] !== 75 || data[pos + 2] !== 3 || data[pos + 3] !== 4) {
-      pos++;
-      while (pos < data.length - 30) {
-        if (data[pos] === 80 && data[pos + 1] === 75 && data[pos + 2] === 3 && data[pos + 3] === 4) break;
-        pos++;
+    if (rec.tagId === TAG_DOC_CHAR_SHAPE && rec.data.length >= 18) {
+      if (rec.data.length >= 50) {
+        const fontSize = rec.data.readUInt32LE(42);
+        const attrFlags = rec.data.readUInt32LE(46);
+        charShapes.push({ fontSize, attrFlags });
+      } else {
+        charShapes.push({ fontSize: 0, attrFlags: 0 });
       }
-      continue;
     }
-    if (++entryCount > MAX_ZIP_ENTRIES) break;
-    const method = view.getUint16(pos + 8, true);
-    const compSize = view.getUint32(pos + 18, true);
-    const nameLen = view.getUint16(pos + 26, true);
-    const extraLen = view.getUint16(pos + 28, true);
-    if (nameLen > 1024 || extraLen > 65535) {
-      pos += 30 + nameLen + extraLen;
-      continue;
-    }
-    const fileStart = pos + 30 + nameLen + extraLen;
-    if (fileStart + compSize > data.length) break;
-    if (compSize === 0 && method !== 0) {
-      pos = fileStart;
-      continue;
-    }
-    const nameBytes = data.slice(pos + 30, pos + 30 + nameLen);
-    const name = new TextDecoder().decode(nameBytes);
-    if (isPathTraversal(name)) {
-      pos = fileStart + compSize;
-      continue;
-    }
-    const fileData = data.slice(fileStart, fileStart + compSize);
-    pos = fileStart + compSize;
-    if (!name.toLowerCase().includes("section") || !name.endsWith(".xml")) continue;
-    try {
-      let content;
-      if (method === 0) {
-        content = new TextDecoder().decode(fileData);
-      } else if (method === 8) {
-        const decompressed = inflateRawSync(Buffer.from(fileData), { maxOutputLength: MAX_DECOMPRESS_SIZE });
-        content = new TextDecoder().decode(decompressed);
-      } else {
-        continue;
+    if (rec.tagId === TAG_DOC_STYLE && rec.data.length >= 8) {
+      try {
+        let offset = 0;
+        const nameLen = rec.data.readUInt16LE(offset);
+        offset += 2;
+        const nameBytes = nameLen * 2;
+        const name = nameBytes > 0 && offset + nameBytes <= rec.data.length ? rec.data.subarray(offset, offset + nameBytes).toString("utf16le") : "";
+        offset += nameBytes;
+        let nameKo = "";
+        if (offset + 2 <= rec.data.length) {
+          const nameKoLen = rec.data.readUInt16LE(offset);
+          offset += 2;
+          const nameKoBytes = nameKoLen * 2;
+          if (nameKoBytes > 0 && offset + nameKoBytes <= rec.data.length) {
+            nameKo = rec.data.subarray(offset, offset + nameKoBytes).toString("utf16le");
+          }
+          offset += nameKoBytes;
+        }
+        const type = offset < rec.data.length ? rec.data.readUInt8(offset) : 0;
+        offset += 1;
+        offset += 1;
+        offset += 2;
+        const paraShapeId = offset + 2 <= rec.data.length ? rec.data.readUInt16LE(offset) : 0;
+        offset += 2;
+        const charShapeId = offset + 2 <= rec.data.length ? rec.data.readUInt16LE(offset) : 0;
+        styles.push({ name, nameKo, charShapeId, paraShapeId, type });
+      } catch {
       }
-      totalDecompressed += content.length * 2;
-      if (totalDecompressed > MAX_DECOMPRESS_SIZE) throw new KordocError("\uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC");
-      sectionNum++;
-      blocks.push(...parseSectionXml(content, void 0, warnings, sectionNum, shared));
-    } catch {
-      continue;
     }
   }
-  if (blocks.length === 0) throw new KordocError("\uC190\uC0C1\uB41C HWPX\uC5D0\uC11C \uC139\uC158 \uB370\uC774\uD130\uB97C \uBCF5\uAD6C\uD560 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
-  applyPageText(blocks, shared);
-  const markdown = blocksToMarkdown(blocks);
-  return { markdown, blocks, warnings: warnings.length > 0 ? warnings : void 0 };
+  return { charShapes, paraShapes, styles, binData, numberings, bullets };
 }
-async function readKordocLayout(zip) {
-  const file = zip.file("Contents/content.hpf") ?? zip.file("content.hpf");
-  if (!file) return null;
-  try {
-    const xml = await file.async("text");
-    if (!/<opf:meta\b[^>]*name="generator"[^>]*content="kordoc"/.test(xml)) return null;
-    return xml.match(/<opf:meta\b[^>]*name="kordoc-layout"[^>]*content="([^"]*)"/)?.[1] ?? null;
-  } catch {
-    return null;
-  }
+function createParaTextState() {
+  return { text: "", ctrlIdx: 0, fieldStack: [], fieldRanges: [] };
 }
-async function resolveSectionPaths(zip) {
-  const manifestPaths = ["Contents/content.hpf", "content.hpf"];
-  for (const mp of manifestPaths) {
-    const mpLower = mp.toLowerCase();
-    const file = zip.file(mp) || Object.values(zip.files).find((f) => f.name.toLowerCase() === mpLower) || null;
-    if (!file) continue;
-    const xml = await file.async("text");
-    const paths = parseSectionPathsFromManifest(xml);
-    if (paths.length > 0) return paths;
-  }
-  const sectionFiles = zip.file(/[Ss]ection\d+\.xml$/);
-  return sectionFiles.map((f) => f.name).sort(compareSectionPaths);
+function isExtendedOnlyCtrlChar(ch) {
+  return ch >= 1 && ch <= 3 || ch >= 11 && ch <= 12 || ch >= 14 && ch <= 18 || ch >= 21 && ch <= 23;
 }
-function parseSectionPathsFromManifest(xml) {
-  const parser = createXmlParser();
-  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
-  const items = doc.getElementsByTagName("opf:item");
-  const spine = doc.getElementsByTagName("opf:itemref");
-  const idToHref = /* @__PURE__ */ new Map();
-  for (let i = 0; i < items.length; i++) {
-    const item = items[i];
-    const id = item.getAttribute("id") || "";
-    const href = normalizeSectionHref(item.getAttribute("href") || "");
-    if (id && href) idToHref.set(id, href);
-  }
-  if (spine.length > 0) {
-    const ordered = [];
-    for (let i = 0; i < spine.length; i++) {
-      const href = idToHref.get(spine[i].getAttribute("idref") || "");
-      if (href) ordered.push(href);
+function appendParaText(state, data, resolveControl) {
+  let result = "";
+  let i = 0;
+  const base = state.text.length;
+  const resolveAt = (byteOffset, extended) => {
+    const ctrlId = data.readUInt32LE(byteOffset);
+    const idx = extended ? state.ctrlIdx : -1;
+    const replacement = resolveControl?.(idx, ctrlId);
+    if (replacement) result += replacement;
+    if (extended) state.ctrlIdx++;
+  };
+  while (i + 1 < data.length) {
+    const ch = data.readUInt16LE(i);
+    i += 2;
+    switch (ch) {
+      // ── char 타입 (2바이트만, 확장 데이터 없음) ──
+      case CHAR_LINE:
+        result += "\n";
+        break;
+      case CHAR_SECTION_BREAK: {
+        if (i + 16 <= data.length && data.readUInt16LE(i) === 11) {
+          resolveAt(i + 2, true);
+          i += 16;
+          break;
+        }
+        result += "\n";
+        break;
+      }
+      case CHAR_PARA:
+        break;
+      // 문단 끝
+      case CHAR_HYPHEN:
+        result += "-";
+        break;
+      case CHAR_NBSP:
+        result += " ";
+        break;
+      case CHAR_FIXED_NBSP:
+        result += "\xA0";
+        break;
+      // 진짜 NBSP
+      case CHAR_FIXED_WIDTH:
+        result += " ";
+        break;
+      // 고정폭 공백
+      // ── inline 타입 (2바이트 + 14바이트 확장) ──
+      case CHAR_TAB:
+        result += "	";
+        if (i + 14 <= data.length) i += 14;
+        break;
+      default:
+        if (ch >= 1 && ch <= 31) {
+          const isExtended = isExtendedOnlyCtrlChar(ch);
+          const isInline = ch >= 4 && ch <= 9 || ch >= 19 && ch <= 20;
+          if ((isExtended || isInline) && i + 14 <= data.length) {
+            if (ch === 3) {
+              state.fieldStack.push({ start: base + result.length, ctrlIdx: state.ctrlIdx });
+            } else if (ch === 4) {
+              const open = state.fieldStack.pop();
+              if (open) {
+                state.fieldRanges.push({ start: open.start, end: base + result.length, ctrlIdx: open.ctrlIdx });
+              }
+            }
+            resolveAt(i, isExtended);
+            i += 14;
+          }
+        } else if (ch >= 32) {
+          if (ch >= 55296 && ch <= 56319 && i + 1 < data.length) {
+            const lo = data.readUInt16LE(i);
+            if (lo >= 56320 && lo <= 57343) {
+              i += 2;
+              const codePoint = (ch - 55296 << 10) + (lo - 56320) + 65536;
+              result += String.fromCodePoint(codePoint);
+              break;
+            }
+          }
+          result += String.fromCharCode(ch);
+        }
+        break;
     }
-    if (ordered.length > 0) return ordered;
   }
-  return Array.from(idToHref.values()).sort(compareSectionPaths);
+  state.text += result;
+}
+function extractEquationText(data) {
+  if (data.length < 6) return null;
+  const scriptLength = data.readUInt16LE(4);
+  const scriptStart = 6;
+  const scriptEnd = scriptStart + scriptLength * 2;
+  if (scriptLength <= 0 || scriptEnd > data.length) return null;
+  const equation = data.subarray(scriptStart, scriptEnd).toString("utf16le").replace(/\0+/g, "").trim();
+  return equation || null;
 }
 
-// src/hwpx/metadata.ts
-async function extractHwpxMetadata(zip, metadata, decompressed) {
+// src/hwp5/images.ts
+function detectImageMime(data) {
+  if (data.length < 4) return null;
+  if (data[0] === 137 && data[1] === 80 && data[2] === 78 && data[3] === 71) return "image/png";
+  if (data[0] === 255 && data[1] === 216 && data[2] === 255) return "image/jpeg";
+  if (data[0] === 71 && data[1] === 73 && data[2] === 70) return "image/gif";
+  if (data[0] === 66 && data[1] === 77) return "image/bmp";
+  if (data[0] === 215 && data[1] === 205 && data[2] === 198 && data[3] === 154) return "image/wmf";
+  if (data[0] === 1 && data[1] === 0 && data[2] === 0 && data[3] === 0) return "image/emf";
+  return null;
+}
+function normalizeBinPayload(data) {
+  if (detectImageMime(data)) return data;
   try {
-    const metaPaths = ["meta.xml", "META-INF/meta.xml", "docProps/core.xml"];
-    for (const mp of metaPaths) {
-      const file = zip.file(mp) || Object.values(zip.files).find((f) => f.name.toLowerCase() === mp.toLowerCase()) || null;
-      if (!file) continue;
-      const xml = await file.async("text");
-      if (decompressed) {
-        decompressed.total += xml.length * 2;
-        if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
-      }
-      parseDublinCoreMetadata(xml, metadata);
-      if (metadata.title || metadata.author) return;
-    }
+    const inflated = decompressStream(data);
+    if (inflated.length > 0) return inflated;
   } catch {
   }
+  return data;
 }
-function parseDublinCoreMetadata(xml, metadata) {
-  const parser = createXmlParser();
-  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
-  if (!doc.documentElement) return;
-  const getText = (tagNames) => {
-    for (const tag of tagNames) {
-      const els = doc.getElementsByTagName(tag);
-      if (els.length > 0) {
-        const text = els[0].textContent?.trim();
-        if (text) return text;
+var BIN_ENTRY_RE = /(?:^|\/)BIN([0-9A-Fa-f]{4,8})(?:\.[^./\\]*)?$/;
+function collectImageBlocks(blocks, out) {
+  for (const b of blocks) {
+    if (b.type === "image") out.push(b);
+    if (b.table) {
+      for (const row of b.table.cells) {
+        for (const cell2 of row) {
+          if (cell2.blocks) collectImageBlocks(cell2.blocks, out);
+        }
       }
     }
-    return void 0;
-  };
-  metadata.title = metadata.title || getText(["dc:title", "title"]);
-  metadata.author = metadata.author || getText(["dc:creator", "creator", "cp:lastModifiedBy"]);
-  metadata.description = metadata.description || getText(["dc:description", "description", "dc:subject", "subject"]);
-  metadata.createdAt = metadata.createdAt || getText(["dcterms:created", "meta:creation-date"]);
-  metadata.modifiedAt = metadata.modifiedAt || getText(["dcterms:modified", "meta:date"]);
-  const keywords = getText(["dc:keyword", "cp:keywords", "meta:keyword"]);
-  if (keywords && !metadata.keywords) {
-    metadata.keywords = keywords.split(/[,;]/).map((k) => k.trim()).filter(Boolean);
+    if (b.children) collectImageBlocks(b.children, out);
   }
 }
-
-// src/hwpx/parser.ts
-async function parseHwpxDocument(buffer, options) {
-  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE, MAX_ZIP_ENTRIES);
-  let zip;
-  try {
-    zip = await JSZip3.loadAsync(buffer);
-  } catch {
-    return extractFromBrokenZip(buffer);
-  }
-  const actualEntryCount = Object.keys(zip.files).length;
-  if (actualEntryCount > MAX_ZIP_ENTRIES) {
-    throw new KordocError("ZIP \uC5D4\uD2B8\uB9AC \uC218 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+function forEachTableCell(blocks, fn) {
+  for (const b of blocks) {
+    if (b.table) {
+      for (const row of b.table.cells) {
+        for (const cell2 of row) {
+          fn(cell2);
+          if (cell2.blocks) forEachTableCell(cell2.blocks, fn);
+        }
+      }
+    }
+    if (b.children) forEachTableCell(b.children, fn);
   }
-  const manifestFile = zip.file("META-INF/manifest.xml");
-  if (manifestFile) {
-    const manifestXml = await manifestFile.async("text");
-    if (isEncryptedHwpx(manifestXml)) {
-      if (isComFallbackAvailable() && options?.filePath) {
-        const { pages, pageCount, warnings: warnings2 } = extractTextViaCom(options.filePath);
-        if (pages.some((p) => p && p.trim().length > 0)) {
-          return comResultToParseResult(pages, pageCount, warnings2);
+}
+var CELL_IMAGE_SENTINEL_RE = /!\[image\]\(hwp5bin:(\d+)\)/g;
+function resolveCellImageSentinels(blocks, renamed) {
+  forEachTableCell(blocks, (cell2) => {
+    if (!cell2.text.includes("hwp5bin:")) return;
+    cell2.text = cell2.text.replace(CELL_IMAGE_SENTINEL_RE, (_m, idStr) => {
+      const filename = renamed.get(Number(idStr));
+      return filename ? `![image](${filename})` : "[\uC774\uBBF8\uC9C0]";
+    });
+  });
+}
+function resolveImageBlocks(binDataMap, blocks, warnings, sweepUnreferenced) {
+  const imageBlocks = [];
+  collectImageBlocks(blocks, imageBlocks);
+  const images = [];
+  const renamed = /* @__PURE__ */ new Map();
+  const resolved = /* @__PURE__ */ new Map();
+  let imageIndex = 0;
+  for (const block of imageBlocks) {
+    if (!block.text) continue;
+    const storageId = parseInt(block.text, 10);
+    if (isNaN(storageId)) continue;
+    let img = resolved.get(storageId);
+    if (img === void 0) {
+      const bin = binDataMap.get(storageId);
+      if (!bin) {
+        warnings.push({ page: block.pageNumber, message: `BinData ${storageId} \uC5C6\uC74C`, code: "SKIPPED_IMAGE" });
+        resolved.set(storageId, null);
+      } else {
+        const mime = detectImageMime(bin.data);
+        if (!mime) {
+          warnings.push({ page: block.pageNumber, message: `BinData ${storageId}: \uC54C \uC218 \uC5C6\uB294 \uC774\uBBF8\uC9C0 \uD615\uC2DD`, code: "SKIPPED_IMAGE" });
+          resolved.set(storageId, null);
+        } else {
+          imageIndex++;
+          const ext = mime.includes("jpeg") ? "jpg" : mime.includes("png") ? "png" : mime.includes("gif") ? "gif" : mime.includes("bmp") ? "bmp" : "bin";
+          img = { filename: `image_${String(imageIndex).padStart(3, "0")}.${ext}`, data: new Uint8Array(bin.data), mime };
+          resolved.set(storageId, img);
+          images.push({ filename: img.filename, data: img.data, mimeType: img.mime });
+          renamed.set(storageId, img.filename);
         }
       }
-      throw new KordocError("DRM \uC554\uD638\uD654\uB41C HWPX \uD30C\uC77C\uC785\uB2C8\uB2E4. Windows + \uD55C\uCEF4 \uC624\uD53C\uC2A4 \uC124\uCE58 \uC2DC \uC790\uB3D9 \uCD94\uCD9C\uB429\uB2C8\uB2E4.");
+      img = resolved.get(storageId);
     }
+    if (!img) {
+      const bin = binDataMap.get(storageId);
+      block.type = "paragraph";
+      block.text = bin ? `[\uC774\uBBF8\uC9C0: ${bin.name}]` : `[\uC774\uBBF8\uC9C0: BinData ${storageId}]`;
+      continue;
+    }
+    block.text = img.filename;
+    block.imageData = { data: img.data, mimeType: img.mime, filename: binDataMap.get(storageId).name };
   }
-  const decompressed = { total: 0 };
-  const metadata = {};
-  await extractHwpxMetadata(zip, metadata, decompressed);
-  const styleMap = await extractHwpxStyles(zip, decompressed);
-  const warnings = [];
-  const sectionPaths = await resolveSectionPaths(zip);
-  if (sectionPaths.length === 0) throw new KordocError("HWPX\uC5D0\uC11C \uC139\uC158 \uD30C\uC77C\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
-  metadata.pageCount = sectionPaths.length;
-  const pageFilter = options?.pages ? parsePageRange(options.pages, sectionPaths.length) : null;
-  const totalTarget = pageFilter ? pageFilter.size : sectionPaths.length;
-  const blocks = [];
-  const shared = createSectionShared();
-  shared.kordocLayout = await readKordocLayout(zip);
-  let parsedSections = 0;
-  for (let si = 0; si < sectionPaths.length; si++) {
-    if (pageFilter && !pageFilter.has(si + 1)) continue;
-    const file = zip.file(sectionPaths[si]);
-    if (!file) continue;
-    try {
-      const xml = await file.async("text");
-      decompressed.total += xml.length * 2;
-      if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
-      blocks.push(...parseSectionXml(xml, styleMap, warnings, si + 1, shared));
-      parsedSections++;
-      options?.onProgress?.(parsedSections, totalTarget);
-    } catch (secErr) {
-      if (secErr instanceof KordocError) throw secErr;
-      warnings.push({ page: si + 1, message: `\uC139\uC158 ${si + 1} \uD30C\uC2F1 \uC2E4\uD328: ${secErr instanceof Error ? secErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
+  if (sweepUnreferenced) {
+    for (const [storageId, bin] of [...binDataMap.entries()].sort((a, b) => a[0] - b[0])) {
+      if (resolved.has(storageId)) continue;
+      const mime = detectImageMime(bin.data);
+      if (!mime) continue;
+      imageIndex++;
+      const ext = mime.includes("jpeg") ? "jpg" : mime.includes("png") ? "png" : mime.includes("gif") ? "gif" : mime.includes("bmp") ? "bmp" : "bin";
+      const filename = `image_${String(imageIndex).padStart(3, "0")}.${ext}`;
+      const data = new Uint8Array(bin.data);
+      images.push({ filename, data, mimeType: mime });
+      blocks.push({ type: "image", text: filename, imageData: { data, mimeType: mime, filename: bin.name } });
     }
   }
-  applyPageText(blocks, shared);
-  const images = await extractImagesFromZip(zip, blocks, decompressed, warnings);
-  detectHwpxHeadings(blocks, styleMap);
-  const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
-  const markdown = blocksToMarkdown(blocks);
-  return { markdown, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0, images: images.length > 0 ? images : void 0 };
+  resolveCellImageSentinels(blocks, renamed);
+  return images;
 }
-
-// src/hwp5/record.ts
-import { inflateRawSync as inflateRawSync2, inflateSync } from "zlib";
-var TAG_PARA_HEADER = 66;
-var TAG_PARA_TEXT = 67;
-var TAG_CHAR_SHAPE = 68;
-var TAG_CTRL_HEADER = 71;
-var TAG_LIST_HEADER = 72;
-var TAG_TABLE = 77;
-var TAG_SHAPE_COMPONENT = 76;
-var TAG_SHAPE_COMPONENT_PICTURE = 85;
-var TAG_SHAPE_COMPONENT_CONTAINER = 86;
-var TAG_EQEDIT = 88;
-var TAG_BIN_DATA = 18;
-var TAG_DOC_CHAR_SHAPE = 21;
-var TAG_NUMBERING = 23;
-var TAG_BULLET = 24;
-var TAG_DOC_PARA_SHAPE = 25;
-var TAG_DOC_STYLE = 26;
-var CHAR_LINE = 0;
-var CHAR_SECTION_BREAK = 10;
-var CHAR_PARA = 13;
-var CHAR_TAB = 9;
-var CHAR_HYPHEN = 30;
-var CHAR_NBSP = 31;
-var CHAR_FIXED_NBSP = 24;
-var CHAR_FIXED_WIDTH = 25;
-var FLAG_COMPRESSED = 1 << 0;
-var FLAG_ENCRYPTED = 1 << 1;
-var FLAG_DISTRIBUTION = 1 << 2;
-var FLAG_DRM = 1 << 4;
-var MAX_RECORDS = 5e5;
-function readRecords(data) {
-  const records = [];
-  let offset = 0;
-  while (offset + 4 <= data.length && records.length < MAX_RECORDS) {
-    const header = data.readUInt32LE(offset);
-    offset += 4;
-    const tagId = header & 1023;
-    const level = header >> 10 & 1023;
-    let size = header >> 20 & 4095;
-    if (size === 4095) {
-      if (offset + 4 > data.length) break;
-      size = data.readUInt32LE(offset);
-      offset += 4;
+function extractHwp5Images(fileIndex, blocks, warnings, sweepUnreferenced) {
+  const binDataMap = /* @__PURE__ */ new Map();
+  if (fileIndex) {
+    for (const entry of fileIndex) {
+      if (!entry?.name || !entry.content) continue;
+      const match = entry.name.match(BIN_ENTRY_RE);
+      if (!match) continue;
+      const idx = parseInt(match[1], 16);
+      const data = normalizeBinPayload(Buffer.from(entry.content));
+      binDataMap.set(idx, { data, name: entry.name });
     }
-    if (offset + size > data.length) break;
-    records.push({ tagId, level, size, data: data.subarray(offset, offset + size) });
-    offset += size;
   }
-  return records;
+  if (binDataMap.size === 0) {
+    resolveCellImageSentinels(blocks, /* @__PURE__ */ new Map());
+    return [];
+  }
+  return resolveImageBlocks(binDataMap, blocks, warnings, sweepUnreferenced);
 }
-var MAX_DECOMPRESS_SIZE2 = 100 * 1024 * 1024;
-function decompressStream(data) {
-  const opts = { maxOutputLength: MAX_DECOMPRESS_SIZE2 };
-  if (data.length >= 2 && data[0] === 120) {
-    try {
-      return inflateSync(data, opts);
-    } catch {
-    }
+function extractHwp5ImagesLenient(lcfb, blocks, warnings, sweepUnreferenced) {
+  const binDataMap = /* @__PURE__ */ new Map();
+  const binRe = /^BIN([0-9A-Fa-f]{4,8})(?:\.|$)/;
+  for (const e of lcfb.entries()) {
+    const match = e.name.match(binRe);
+    if (!match) continue;
+    const idx = parseInt(match[1], 16);
+    const raw = lcfb.findStream(e.name);
+    if (!raw) continue;
+    binDataMap.set(idx, { data: normalizeBinPayload(raw), name: e.name });
+  }
+  if (binDataMap.size === 0) {
+    resolveCellImageSentinels(blocks, /* @__PURE__ */ new Map());
+    return [];
   }
-  return inflateRawSync2(data, opts);
+  return resolveImageBlocks(binDataMap, blocks, warnings, sweepUnreferenced);
 }
-function parseFileHeader(data) {
-  if (data.length < 40) throw new KordocError("FileHeader\uAC00 \uB108\uBB34 \uC9E7\uC2B5\uB2C8\uB2E4 (\uCD5C\uC18C 40\uBC14\uC774\uD2B8)");
-  const sig = data.subarray(0, 32).toString("utf8").replace(/\0+$/, "");
-  return {
-    signature: sig,
-    versionMajor: data[35],
-    flags: data.readUInt32LE(36)
-  };
+
+// src/hwpx/images.ts
+function imageExtToMime(ext) {
+  switch (ext.toLowerCase()) {
+    case "jpg":
+    case "jpeg":
+      return "image/jpeg";
+    case "png":
+      return "image/png";
+    case "gif":
+      return "image/gif";
+    case "bmp":
+      return "image/bmp";
+    case "tif":
+    case "tiff":
+      return "image/tiff";
+    case "wmf":
+      return "image/wmf";
+    case "emf":
+      return "image/emf";
+    case "svg":
+      return "image/svg+xml";
+    default:
+      return "application/octet-stream";
+  }
 }
-function readHwpString(data, offset) {
-  if (offset + 2 > data.length) return { value: "", next: data.length };
-  const len = data.readUInt16LE(offset);
-  const start = offset + 2;
-  const end = start + len * 2;
-  if (len === 0 || end > data.length) return { value: "", next: start };
-  return { value: data.subarray(start, end).toString("utf16le"), next: end };
+function mimeToExt(mime) {
+  if (mime.includes("jpeg")) return "jpg";
+  if (mime.includes("png")) return "png";
+  if (mime.includes("gif")) return "gif";
+  if (mime.includes("bmp")) return "bmp";
+  if (mime.includes("tiff")) return "tif";
+  if (mime.includes("wmf")) return "wmf";
+  if (mime.includes("emf")) return "emf";
+  if (mime.includes("svg")) return "svg";
+  return "bin";
 }
-function parseDocInfo(records) {
-  const charShapes = [];
-  const paraShapes = [];
-  const styles = [];
-  const binData = [];
-  const numberings = [];
-  const bullets = [];
-  for (const rec of records) {
-    if (rec.tagId === TAG_DOC_PARA_SHAPE && rec.data.length >= 4) {
-      const attr1 = rec.data.readUInt32LE(0);
-      const headType = attr1 >>> 23 & 3;
-      const paraLevel = attr1 >>> 25 & 7;
-      const numberingId = rec.data.length >= 32 ? rec.data.readUInt16LE(30) : 0;
-      paraShapes.push({ headType, paraLevel, numberingId });
-    }
-    if (rec.tagId === TAG_BIN_DATA && rec.data.length >= 2) {
-      const attr = rec.data.readUInt16LE(0);
-      const typeBits = attr & 15;
-      if (typeBits === 0) {
-        binData.push({ kind: "link", storageId: 0, extension: "" });
-      } else {
-        const storageId = rec.data.length >= 4 ? rec.data.readUInt16LE(2) : 0;
-        const { value: extension } = readHwpString(rec.data, 4);
-        binData.push({ kind: typeBits === 2 ? "storage" : "embed", storageId, extension });
+function collectImageBlocks2(blocks, out, ownerCell, depth = 0) {
+  if (depth > MAX_XML_DEPTH) return;
+  for (const block of blocks) {
+    if (block.type === "image") {
+      out.push({ block, ownerCell });
+    } else if (block.type === "table" && block.table) {
+      for (const row of block.table.cells) {
+        for (const cell2 of row) {
+          if (cell2.blocks?.length) collectImageBlocks2(cell2.blocks, out, cell2, depth + 1);
+        }
       }
     }
-    if (rec.tagId === TAG_NUMBERING && rec.data.length >= 14) {
-      const levelFormats = [];
-      const numberFormats = [];
-      const startNumbers = [1, 1, 1, 1, 1, 1, 1];
-      let offset = 0;
-      for (let level = 0; level < 7; level++) {
-        if (offset + 12 > rec.data.length) {
-          levelFormats.push("");
-          numberFormats.push(0);
-          continue;
+  }
+}
+async function extractImagesFromZip(zip, blocks, decompressed, warnings, sweepUnreferenced) {
+  const images = [];
+  let imageIndex = 0;
+  const usedPaths = /* @__PURE__ */ new Set();
+  const imageBlocks = [];
+  collectImageBlocks2(blocks, imageBlocks);
+  const resolved = /* @__PURE__ */ new Map();
+  for (const { block, ownerCell } of imageBlocks) {
+    if (block.type !== "image" || !block.text) continue;
+    const ref = block.text;
+    let img = resolved.get(ref);
+    if (img === void 0) {
+      img = null;
+      const candidates = [
+        `BinData/${ref}`,
+        `Contents/BinData/${ref}`,
+        ref
+        // 절대 경로일 수도 있음
+      ];
+      let resolvedPath = null;
+      if (!ref.includes(".")) {
+        const prefixes = [`BinData/${ref}`, `Contents/BinData/${ref}`];
+        for (const prefix of prefixes) {
+          const match = zip.file(new RegExp(`^${prefix.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}\\.[a-zA-Z0-9]+$`));
+          if (match.length > 0) {
+            resolvedPath = match[0].name;
+            break;
+          }
         }
-        const attr = rec.data.readUInt32LE(offset);
-        numberFormats.push(attr >>> 5 & 15);
-        offset += 12;
-        const { value, next } = readHwpString(rec.data, offset);
-        levelFormats.push(value);
-        offset = next;
       }
-      let baseStart = 1;
-      if (offset + 2 <= rec.data.length) {
-        baseStart = rec.data.readUInt16LE(offset) || 1;
-        offset += 2;
+      const allCandidates = resolvedPath ? [resolvedPath, ...candidates] : candidates;
+      for (const path of allCandidates) {
+        if (isPathTraversal(path)) continue;
+        const file = zip.file(path);
+        if (!file) continue;
+        try {
+          const data = await file.async("uint8array");
+          decompressed.total += data.length;
+          if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+          const ext = path.includes(".") ? path.split(".").pop() || "png" : "png";
+          const mimeType = imageExtToMime(ext);
+          imageIndex++;
+          const filename = `image_${String(imageIndex).padStart(3, "0")}.${mimeToExt(mimeType)}`;
+          img = { filename, data, mimeType };
+          images.push(img);
+          usedPaths.add(path);
+          break;
+        } catch (err) {
+          if (err instanceof KordocError) throw err;
+        }
       }
-      for (let level = 0; level < 7; level++) {
-        if (offset + 4 <= rec.data.length) {
-          startNumbers[level] = rec.data.readUInt32LE(offset) || 1;
-          offset += 4;
-        } else {
-          startNumbers[level] = baseStart;
+      if (!img) warnings?.push({ page: block.pageNumber, message: `\uC774\uBBF8\uC9C0 \uD30C\uC77C \uC5C6\uC74C: ${ref}`, code: "SKIPPED_IMAGE" });
+      resolved.set(ref, img);
+    }
+    if (!img) {
+      block.type = "paragraph";
+      block.text = `[\uC774\uBBF8\uC9C0: ${ref}]`;
+      if (ownerCell) ownerCell.text = ownerCell.text.replace(`![image](${ref})`, `[\uC774\uBBF8\uC9C0: ${ref}]`);
+      continue;
+    }
+    block.text = img.filename;
+    block.imageData = { data: img.data, mimeType: img.mimeType, filename: ref };
+    if (ownerCell) ownerCell.text = ownerCell.text.replace(`![image](${ref})`, `![image](${img.filename})`);
+  }
+  if (sweepUnreferenced) {
+    const binEntries = zip.file(/(?:^|\/)BinData\//i);
+    for (const file of binEntries) {
+      if (file.dir || usedPaths.has(file.name) || isPathTraversal(file.name)) continue;
+      try {
+        const data = await file.async("uint8array");
+        decompressed.total += data.length;
+        if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+        const ext = file.name.includes(".") ? file.name.split(".").pop() || "" : "";
+        let mimeType = imageExtToMime(ext);
+        if (mimeType === "application/octet-stream") {
+          const sniffed = detectImageMime(data);
+          if (!sniffed) continue;
+          mimeType = sniffed;
         }
+        imageIndex++;
+        const filename = `image_${String(imageIndex).padStart(3, "0")}.${mimeToExt(mimeType)}`;
+        const img = { filename, data, mimeType };
+        images.push(img);
+        blocks.push({ type: "image", text: filename, imageData: { data, mimeType, filename: file.name } });
+      } catch (err) {
+        if (err instanceof KordocError) throw err;
+        warnings?.push({ message: `\uC774\uBBF8\uC9C0 \uCD94\uCD9C \uC2E4\uD328: ${file.name}`, code: "SKIPPED_IMAGE" });
       }
-      numberings.push({ levelFormats, numberFormats, startNumbers });
     }
-    if (rec.tagId === TAG_BULLET && rec.data.length >= 14) {
-      const code = rec.data.readUInt16LE(12);
-      bullets.push({ char: code > 0 ? String.fromCharCode(code) : "\u2022" });
+  }
+  return images;
+}
+
+// src/hwpx/metadata.ts
+import JSZip2 from "jszip";
+
+// src/hwpx/zip-sections.ts
+import { inflateRawSync as inflateRawSync2 } from "zlib";
+function extractFromBrokenZip(buffer) {
+  const data = new Uint8Array(buffer);
+  const view = new DataView(buffer);
+  let pos = 0;
+  const blocks = [];
+  const warnings = [
+    { code: "BROKEN_ZIP_RECOVERY", message: "\uC190\uC0C1\uB41C ZIP \uAD6C\uC870 \u2014 Local File Header \uAE30\uBC18 \uBCF5\uAD6C \uBAA8\uB4DC" }
+  ];
+  let totalDecompressed = 0;
+  let entryCount = 0;
+  let sectionNum = 0;
+  const shared = createSectionShared();
+  while (pos < data.length - 30) {
+    if (data[pos] !== 80 || data[pos + 1] !== 75 || data[pos + 2] !== 3 || data[pos + 3] !== 4) {
+      pos++;
+      while (pos < data.length - 30) {
+        if (data[pos] === 80 && data[pos + 1] === 75 && data[pos + 2] === 3 && data[pos + 3] === 4) break;
+        pos++;
+      }
+      continue;
+    }
+    if (++entryCount > MAX_ZIP_ENTRIES) break;
+    const method = view.getUint16(pos + 8, true);
+    const compSize = view.getUint32(pos + 18, true);
+    const nameLen = view.getUint16(pos + 26, true);
+    const extraLen = view.getUint16(pos + 28, true);
+    if (nameLen > 1024 || extraLen > 65535) {
+      pos += 30 + nameLen + extraLen;
+      continue;
+    }
+    const fileStart = pos + 30 + nameLen + extraLen;
+    if (fileStart + compSize > data.length) break;
+    if (compSize === 0 && method !== 0) {
+      pos = fileStart;
+      continue;
+    }
+    const nameBytes = data.slice(pos + 30, pos + 30 + nameLen);
+    const name = new TextDecoder().decode(nameBytes);
+    if (isPathTraversal(name)) {
+      pos = fileStart + compSize;
+      continue;
     }
-    if (rec.tagId === TAG_DOC_CHAR_SHAPE && rec.data.length >= 18) {
-      if (rec.data.length >= 50) {
-        const fontSize = rec.data.readUInt32LE(42);
-        const attrFlags = rec.data.readUInt32LE(46);
-        charShapes.push({ fontSize, attrFlags });
+    const fileData = data.slice(fileStart, fileStart + compSize);
+    pos = fileStart + compSize;
+    if (!name.toLowerCase().includes("section") || !name.endsWith(".xml")) continue;
+    try {
+      let content;
+      if (method === 0) {
+        content = new TextDecoder().decode(fileData);
+      } else if (method === 8) {
+        const decompressed = inflateRawSync2(Buffer.from(fileData), { maxOutputLength: MAX_DECOMPRESS_SIZE });
+        content = new TextDecoder().decode(decompressed);
       } else {
-        charShapes.push({ fontSize: 0, attrFlags: 0 });
-      }
-    }
-    if (rec.tagId === TAG_DOC_STYLE && rec.data.length >= 8) {
-      try {
-        let offset = 0;
-        const nameLen = rec.data.readUInt16LE(offset);
-        offset += 2;
-        const nameBytes = nameLen * 2;
-        const name = nameBytes > 0 && offset + nameBytes <= rec.data.length ? rec.data.subarray(offset, offset + nameBytes).toString("utf16le") : "";
-        offset += nameBytes;
-        let nameKo = "";
-        if (offset + 2 <= rec.data.length) {
-          const nameKoLen = rec.data.readUInt16LE(offset);
-          offset += 2;
-          const nameKoBytes = nameKoLen * 2;
-          if (nameKoBytes > 0 && offset + nameKoBytes <= rec.data.length) {
-            nameKo = rec.data.subarray(offset, offset + nameKoBytes).toString("utf16le");
-          }
-          offset += nameKoBytes;
-        }
-        const type = offset < rec.data.length ? rec.data.readUInt8(offset) : 0;
-        offset += 1;
-        offset += 1;
-        offset += 2;
-        const paraShapeId = offset + 2 <= rec.data.length ? rec.data.readUInt16LE(offset) : 0;
-        offset += 2;
-        const charShapeId = offset + 2 <= rec.data.length ? rec.data.readUInt16LE(offset) : 0;
-        styles.push({ name, nameKo, charShapeId, paraShapeId, type });
-      } catch {
+        continue;
       }
+      totalDecompressed += content.length * 2;
+      if (totalDecompressed > MAX_DECOMPRESS_SIZE) throw new KordocError("\uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC");
+      sectionNum++;
+      blocks.push(...parseSectionXml(content, void 0, warnings, sectionNum, shared));
+    } catch {
+      continue;
     }
   }
-  return { charShapes, paraShapes, styles, binData, numberings, bullets };
+  if (blocks.length === 0) throw new KordocError("\uC190\uC0C1\uB41C HWPX\uC5D0\uC11C \uC139\uC158 \uB370\uC774\uD130\uB97C \uBCF5\uAD6C\uD560 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
+  applyPageText(blocks, shared);
+  const markdown = blocksToMarkdown(blocks);
+  return { markdown, blocks, warnings: warnings.length > 0 ? warnings : void 0 };
 }
-function createParaTextState() {
-  return { text: "", ctrlIdx: 0, fieldStack: [], fieldRanges: [] };
+async function readKordocLayout(zip) {
+  const file = zip.file("Contents/content.hpf") ?? zip.file("content.hpf");
+  if (!file) return null;
+  try {
+    const xml = await file.async("text");
+    if (!/<opf:meta\b[^>]*name="generator"[^>]*content="kordoc"/.test(xml)) return null;
+    return xml.match(/<opf:meta\b[^>]*name="kordoc-layout"[^>]*content="([^"]*)"/)?.[1] ?? null;
+  } catch {
+    return null;
+  }
 }
-function isExtendedOnlyCtrlChar(ch) {
-  return ch >= 1 && ch <= 3 || ch >= 11 && ch <= 12 || ch >= 14 && ch <= 18 || ch >= 21 && ch <= 23;
+async function resolveSectionPaths(zip) {
+  const manifestPaths = ["Contents/content.hpf", "content.hpf"];
+  for (const mp of manifestPaths) {
+    const mpLower = mp.toLowerCase();
+    const file = zip.file(mp) || Object.values(zip.files).find((f) => f.name.toLowerCase() === mpLower) || null;
+    if (!file) continue;
+    const xml = await file.async("text");
+    const paths = parseSectionPathsFromManifest(xml);
+    if (paths.length > 0) return paths;
+  }
+  const sectionFiles = zip.file(/[Ss]ection\d+\.xml$/);
+  return sectionFiles.map((f) => f.name).sort(compareSectionPaths);
 }
-function appendParaText(state, data, resolveControl) {
-  let result = "";
-  let i = 0;
-  const base = state.text.length;
-  const resolveAt = (byteOffset, extended) => {
-    const ctrlId = data.readUInt32LE(byteOffset);
-    const idx = extended ? state.ctrlIdx : -1;
-    const replacement = resolveControl?.(idx, ctrlId);
-    if (replacement) result += replacement;
-    if (extended) state.ctrlIdx++;
+function parseSectionPathsFromManifest(xml) {
+  const parser = createXmlParser();
+  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
+  const items = doc.getElementsByTagName("opf:item");
+  const spine = doc.getElementsByTagName("opf:itemref");
+  const idToHref = /* @__PURE__ */ new Map();
+  for (let i = 0; i < items.length; i++) {
+    const item = items[i];
+    const id = item.getAttribute("id") || "";
+    const href = normalizeSectionHref(item.getAttribute("href") || "");
+    if (id && href) idToHref.set(id, href);
+  }
+  if (spine.length > 0) {
+    const ordered = [];
+    for (let i = 0; i < spine.length; i++) {
+      const href = idToHref.get(spine[i].getAttribute("idref") || "");
+      if (href) ordered.push(href);
+    }
+    if (ordered.length > 0) return ordered;
+  }
+  return Array.from(idToHref.values()).sort(compareSectionPaths);
+}
+
+// src/hwpx/metadata.ts
+async function extractHwpxMetadata(zip, metadata, decompressed) {
+  try {
+    const metaPaths = ["meta.xml", "META-INF/meta.xml", "docProps/core.xml"];
+    for (const mp of metaPaths) {
+      const file = zip.file(mp) || Object.values(zip.files).find((f) => f.name.toLowerCase() === mp.toLowerCase()) || null;
+      if (!file) continue;
+      const xml = await file.async("text");
+      if (decompressed) {
+        decompressed.total += xml.length * 2;
+        if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+      }
+      parseDublinCoreMetadata(xml, metadata);
+      if (metadata.title || metadata.author) return;
+    }
+  } catch {
+  }
+}
+function parseDublinCoreMetadata(xml, metadata) {
+  const parser = createXmlParser();
+  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
+  if (!doc.documentElement) return;
+  const getText = (tagNames) => {
+    for (const tag of tagNames) {
+      const els = doc.getElementsByTagName(tag);
+      if (els.length > 0) {
+        const text = els[0].textContent?.trim();
+        if (text) return text;
+      }
+    }
+    return void 0;
   };
-  while (i + 1 < data.length) {
-    const ch = data.readUInt16LE(i);
-    i += 2;
-    switch (ch) {
-      // ── char 타입 (2바이트만, 확장 데이터 없음) ──
-      case CHAR_LINE:
-        result += "\n";
-        break;
-      case CHAR_SECTION_BREAK: {
-        if (i + 16 <= data.length && data.readUInt16LE(i) === 11) {
-          resolveAt(i + 2, true);
-          i += 16;
-          break;
+  metadata.title = metadata.title || getText(["dc:title", "title"]);
+  metadata.author = metadata.author || getText(["dc:creator", "creator", "cp:lastModifiedBy"]);
+  metadata.description = metadata.description || getText(["dc:description", "description", "dc:subject", "subject"]);
+  metadata.createdAt = metadata.createdAt || getText(["dcterms:created", "meta:creation-date"]);
+  metadata.modifiedAt = metadata.modifiedAt || getText(["dcterms:modified", "meta:date"]);
+  const keywords = getText(["dc:keyword", "cp:keywords", "meta:keyword"]);
+  if (keywords && !metadata.keywords) {
+    metadata.keywords = keywords.split(/[,;]/).map((k) => k.trim()).filter(Boolean);
+  }
+}
+
+// src/hwpx/parser.ts
+async function parseHwpxDocument(buffer, options) {
+  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE, MAX_ZIP_ENTRIES);
+  let zip;
+  try {
+    zip = await JSZip3.loadAsync(buffer);
+  } catch {
+    return extractFromBrokenZip(buffer);
+  }
+  const actualEntryCount = Object.keys(zip.files).length;
+  if (actualEntryCount > MAX_ZIP_ENTRIES) {
+    throw new KordocError("ZIP \uC5D4\uD2B8\uB9AC \uC218 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+  }
+  const manifestFile = zip.file("META-INF/manifest.xml");
+  if (manifestFile) {
+    const manifestXml = await manifestFile.async("text");
+    if (isEncryptedHwpx(manifestXml)) {
+      if (isComFallbackAvailable() && options?.filePath) {
+        const { pages, pageCount, warnings: warnings2 } = extractTextViaCom(options.filePath);
+        if (pages.some((p) => p && p.trim().length > 0)) {
+          return comResultToParseResult(pages, pageCount, warnings2);
         }
-        result += "\n";
-        break;
       }
-      case CHAR_PARA:
-        break;
-      // 문단 끝
-      case CHAR_HYPHEN:
-        result += "-";
-        break;
-      case CHAR_NBSP:
-        result += " ";
-        break;
-      case CHAR_FIXED_NBSP:
-        result += "\xA0";
-        break;
-      // 진짜 NBSP
-      case CHAR_FIXED_WIDTH:
-        result += " ";
-        break;
-      // 고정폭 공백
-      // ── inline 타입 (2바이트 + 14바이트 확장) ──
-      case CHAR_TAB:
-        result += "	";
-        if (i + 14 <= data.length) i += 14;
-        break;
-      default:
-        if (ch >= 1 && ch <= 31) {
-          const isExtended = isExtendedOnlyCtrlChar(ch);
-          const isInline = ch >= 4 && ch <= 9 || ch >= 19 && ch <= 20;
-          if ((isExtended || isInline) && i + 14 <= data.length) {
-            if (ch === 3) {
-              state.fieldStack.push({ start: base + result.length, ctrlIdx: state.ctrlIdx });
-            } else if (ch === 4) {
-              const open = state.fieldStack.pop();
-              if (open) {
-                state.fieldRanges.push({ start: open.start, end: base + result.length, ctrlIdx: open.ctrlIdx });
-              }
-            }
-            resolveAt(i, isExtended);
-            i += 14;
-          }
-        } else if (ch >= 32) {
-          if (ch >= 55296 && ch <= 56319 && i + 1 < data.length) {
-            const lo = data.readUInt16LE(i);
-            if (lo >= 56320 && lo <= 57343) {
-              i += 2;
-              const codePoint = (ch - 55296 << 10) + (lo - 56320) + 65536;
-              result += String.fromCodePoint(codePoint);
-              break;
-            }
-          }
-          result += String.fromCharCode(ch);
-        }
-        break;
+      throw new KordocError("DRM \uC554\uD638\uD654\uB41C HWPX \uD30C\uC77C\uC785\uB2C8\uB2E4. Windows + \uD55C\uCEF4 \uC624\uD53C\uC2A4 \uC124\uCE58 \uC2DC \uC790\uB3D9 \uCD94\uCD9C\uB429\uB2C8\uB2E4.");
+    }
+  }
+  const decompressed = { total: 0 };
+  const metadata = {};
+  await extractHwpxMetadata(zip, metadata, decompressed);
+  const styleMap = await extractHwpxStyles(zip, decompressed);
+  const warnings = [];
+  const sectionPaths = await resolveSectionPaths(zip);
+  if (sectionPaths.length === 0) throw new KordocError("HWPX\uC5D0\uC11C \uC139\uC158 \uD30C\uC77C\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
+  metadata.pageCount = sectionPaths.length;
+  const pageFilter = options?.pages ? parsePageRange(options.pages, sectionPaths.length) : null;
+  const totalTarget = pageFilter ? pageFilter.size : sectionPaths.length;
+  const blocks = [];
+  const shared = createSectionShared();
+  shared.kordocLayout = await readKordocLayout(zip);
+  let parsedSections = 0;
+  for (let si = 0; si < sectionPaths.length; si++) {
+    if (pageFilter && !pageFilter.has(si + 1)) continue;
+    const file = zip.file(sectionPaths[si]);
+    if (!file) continue;
+    try {
+      const xml = await file.async("text");
+      decompressed.total += xml.length * 2;
+      if (decompressed.total > MAX_DECOMPRESS_SIZE) throw new KordocError("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
+      blocks.push(...parseSectionXml(xml, styleMap, warnings, si + 1, shared));
+      parsedSections++;
+      options?.onProgress?.(parsedSections, totalTarget);
+    } catch (secErr) {
+      if (secErr instanceof KordocError) throw secErr;
+      warnings.push({ page: si + 1, message: `\uC139\uC158 ${si + 1} \uD30C\uC2F1 \uC2E4\uD328: ${secErr instanceof Error ? secErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
     }
   }
-  state.text += result;
-}
-function extractEquationText(data) {
-  if (data.length < 6) return null;
-  const scriptLength = data.readUInt16LE(4);
-  const scriptStart = 6;
-  const scriptEnd = scriptStart + scriptLength * 2;
-  if (scriptLength <= 0 || scriptEnd > data.length) return null;
-  const equation = data.subarray(scriptStart, scriptEnd).toString("utf16le").replace(/\0+/g, "").trim();
-  return equation || null;
+  applyPageText(blocks, shared);
+  const images = await extractImagesFromZip(zip, blocks, decompressed, warnings, !pageFilter);
+  detectHwpxHeadings(blocks, styleMap);
+  const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
+  const markdown = blocksToMarkdown(blocks);
+  return { markdown, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0, images: images.length > 0 ? images : void 0 };
 }
 
 // src/hwp5/numbering.ts
@@ -3513,282 +3693,6 @@ function expandNumberingFormat(formatStr, counters, numbering) {
   return result;
 }
 
-// src/hwp5/images.ts
-function detectImageMime(data) {
-  if (data.length < 4) return null;
-  if (data[0] === 137 && data[1] === 80 && data[2] === 78 && data[3] === 71) return "image/png";
-  if (data[0] === 255 && data[1] === 216 && data[2] === 255) return "image/jpeg";
-  if (data[0] === 71 && data[1] === 73 && data[2] === 70) return "image/gif";
-  if (data[0] === 66 && data[1] === 77) return "image/bmp";
-  if (data[0] === 215 && data[1] === 205 && data[2] === 198 && data[3] === 154) return "image/wmf";
-  if (data[0] === 1 && data[1] === 0 && data[2] === 0 && data[3] === 0) return "image/emf";
-  return null;
-}
-function normalizeBinPayload(data) {
-  if (detectImageMime(data)) return data;
-  try {
-    const inflated = decompressStream(data);
-    if (inflated.length > 0) return inflated;
-  } catch {
-  }
-  return data;
-}
-var BIN_ENTRY_RE = /(?:^|\/)BIN([0-9A-Fa-f]{4,8})(?:\.[^./\\]*)?$/;
-function collectImageBlocks2(blocks, out) {
-  for (const b of blocks) {
-    if (b.type === "image") out.push(b);
-    if (b.table) {
-      for (const row of b.table.cells) {
-        for (const cell2 of row) {
-          if (cell2.blocks) collectImageBlocks2(cell2.blocks, out);
-        }
-      }
-    }
-    if (b.children) collectImageBlocks2(b.children, out);
-  }
-}
-function forEachTableCell(blocks, fn) {
-  for (const b of blocks) {
-    if (b.table) {
-      for (const row of b.table.cells) {
-        for (const cell2 of row) {
-          fn(cell2);
-          if (cell2.blocks) forEachTableCell(cell2.blocks, fn);
-        }
-      }
-    }
-    if (b.children) forEachTableCell(b.children, fn);
-  }
-}
-var CELL_IMAGE_SENTINEL_RE = /!\[image\]\(hwp5bin:(\d+)\)/g;
-function resolveCellImageSentinels(blocks, renamed) {
-  forEachTableCell(blocks, (cell2) => {
-    if (!cell2.text.includes("hwp5bin:")) return;
-    cell2.text = cell2.text.replace(CELL_IMAGE_SENTINEL_RE, (_m, idStr) => {
-      const filename = renamed.get(Number(idStr));
-      return filename ? `![image](${filename})` : "[\uC774\uBBF8\uC9C0]";
-    });
-  });
-}
-function resolveImageBlocks(binDataMap, blocks, warnings) {
-  const imageBlocks = [];
-  collectImageBlocks2(blocks, imageBlocks);
-  if (imageBlocks.length === 0) return [];
-  const images = [];
-  const renamed = /* @__PURE__ */ new Map();
-  const resolved = /* @__PURE__ */ new Map();
-  let imageIndex = 0;
-  for (const block of imageBlocks) {
-    if (!block.text) continue;
-    const storageId = parseInt(block.text, 10);
-    if (isNaN(storageId)) continue;
-    let img = resolved.get(storageId);
-    if (img === void 0) {
-      const bin = binDataMap.get(storageId);
-      if (!bin) {
-        warnings.push({ page: block.pageNumber, message: `BinData ${storageId} \uC5C6\uC74C`, code: "SKIPPED_IMAGE" });
-        resolved.set(storageId, null);
-      } else {
-        const mime = detectImageMime(bin.data);
-        if (!mime) {
-          warnings.push({ page: block.pageNumber, message: `BinData ${storageId}: \uC54C \uC218 \uC5C6\uB294 \uC774\uBBF8\uC9C0 \uD615\uC2DD`, code: "SKIPPED_IMAGE" });
-          resolved.set(storageId, null);
-        } else {
-          imageIndex++;
-          const ext = mime.includes("jpeg") ? "jpg" : mime.includes("png") ? "png" : mime.includes("gif") ? "gif" : mime.includes("bmp") ? "bmp" : "bin";
-          img = { filename: `image_${String(imageIndex).padStart(3, "0")}.${ext}`, data: new Uint8Array(bin.data), mime };
-          resolved.set(storageId, img);
-          images.push({ filename: img.filename, data: img.data, mimeType: img.mime });
-          renamed.set(storageId, img.filename);
-        }
-      }
-      img = resolved.get(storageId);
-    }
-    if (!img) {
-      const bin = binDataMap.get(storageId);
-      block.type = "paragraph";
-      block.text = bin ? `[\uC774\uBBF8\uC9C0: ${bin.name}]` : `[\uC774\uBBF8\uC9C0: BinData ${storageId}]`;
-      continue;
-    }
-    block.text = img.filename;
-    block.imageData = { data: img.data, mimeType: img.mime, filename: binDataMap.get(storageId).name };
-  }
-  resolveCellImageSentinels(blocks, renamed);
-  return images;
-}
-function extractHwp5Images(fileIndex, blocks, warnings) {
-  const binDataMap = /* @__PURE__ */ new Map();
-  if (fileIndex) {
-    for (const entry of fileIndex) {
-      if (!entry?.name || !entry.content) continue;
-      const match = entry.name.match(BIN_ENTRY_RE);
-      if (!match) continue;
-      const idx = parseInt(match[1], 16);
-      const data = normalizeBinPayload(Buffer.from(entry.content));
-      binDataMap.set(idx, { data, name: entry.name });
-    }
-  }
-  if (binDataMap.size === 0) {
-    resolveCellImageSentinels(blocks, /* @__PURE__ */ new Map());
-    return [];
-  }
-  return resolveImageBlocks(binDataMap, blocks, warnings);
-}
-function extractHwp5ImagesLenient(lcfb, blocks, warnings) {
-  const binDataMap = /* @__PURE__ */ new Map();
-  const binRe = /^BIN([0-9A-Fa-f]{4,8})(?:\.|$)/;
-  for (const e of lcfb.entries()) {
-    const match = e.name.match(binRe);
-    if (!match) continue;
-    const idx = parseInt(match[1], 16);
-    const raw = lcfb.findStream(e.name);
-    if (!raw) continue;
-    binDataMap.set(idx, { data: normalizeBinPayload(raw), name: e.name });
-  }
-  if (binDataMap.size === 0) {
-    resolveCellImageSentinels(blocks, /* @__PURE__ */ new Map());
-    return [];
-  }
-  return resolveImageBlocks(binDataMap, blocks, warnings);
-}
-
-// src/image/transcode.ts
-import { deflateSync } from "zlib";
-var CRC_TABLE = (() => {
-  const table2 = new Uint32Array(256);
-  for (let n = 0; n < 256; n++) {
-    let c = n;
-    for (let k = 0; k < 8; k++) {
-      c = c & 1 ? 3988292384 ^ c >>> 1 : c >>> 1;
-    }
-    table2[n] = c >>> 0;
-  }
-  return table2;
-})();
-function crc32(bytes) {
-  let c = 4294967295;
-  for (let i = 0; i < bytes.length; i++) {
-    c = CRC_TABLE[(c ^ bytes[i]) & 255] ^ c >>> 8;
-  }
-  return (c ^ 4294967295) >>> 0;
-}
-var BI_RGB = 0;
-var MAX_DIM = 32767;
-var MAX_PIXELS = 36e6;
-function bmpToPng(bmp) {
-  if (bmp.length < 54) return null;
-  if (bmp[0] !== 66 || bmp[1] !== 77) return null;
-  const dv = new DataView(bmp.buffer, bmp.byteOffset, bmp.byteLength);
-  const dataOffset = dv.getUint32(10, true);
-  const headerSize = dv.getUint32(14, true);
-  if (headerSize < 40) return null;
-  const width = dv.getInt32(18, true);
-  const rawHeight = dv.getInt32(22, true);
-  const bitCount = dv.getUint16(28, true);
-  const compression = dv.getUint32(30, true);
-  if (compression !== BI_RGB) return null;
-  if (bitCount !== 24 && bitCount !== 32) return null;
-  if (width <= 0 || rawHeight === 0) return null;
-  if (width > MAX_DIM || Math.abs(rawHeight) > MAX_DIM) return null;
-  const topDown = rawHeight < 0;
-  const height = Math.abs(rawHeight);
-  if (width * height > MAX_PIXELS) return null;
-  const bytesPerPixel = bitCount >> 3;
-  const rowStride = width * bytesPerPixel + 3 & ~3;
-  if (dataOffset + rowStride * height > bmp.length) return null;
-  const rgba = new Uint8Array(width * height * 4);
-  let anyAlpha = 0;
-  for (let y = 0; y < height; y++) {
-    const srcRow = topDown ? y : height - 1 - y;
-    let src = dataOffset + srcRow * rowStride;
-    let dst = y * width * 4;
-    for (let x = 0; x < width; x++) {
-      rgba[dst] = bmp[src + 2];
-      rgba[dst + 1] = bmp[src + 1];
-      rgba[dst + 2] = bmp[src];
-      const a = bitCount === 32 ? bmp[src + 3] : 255;
-      rgba[dst + 3] = a;
-      anyAlpha |= a;
-      src += bytesPerPixel;
-      dst += 4;
-    }
-  }
-  if (bitCount === 32 && anyAlpha === 0) {
-    for (let i = 3; i < rgba.length; i += 4) rgba[i] = 255;
-  }
-  return encodePng(width, height, rgba);
-}
-var PNG_SIGNATURE = Uint8Array.of(137, 80, 78, 71, 13, 10, 26, 10);
-function chunk(type, data) {
-  const body = new Uint8Array(4 + data.length);
-  body[0] = type.charCodeAt(0);
-  body[1] = type.charCodeAt(1);
-  body[2] = type.charCodeAt(2);
-  body[3] = type.charCodeAt(3);
-  body.set(data, 4);
-  const out = new Uint8Array(8 + data.length + 4);
-  const dv = new DataView(out.buffer);
-  dv.setUint32(0, data.length, false);
-  out.set(body, 4);
-  dv.setUint32(8 + data.length, crc32(body), false);
-  return out;
-}
-function encodePng(width, height, rgba) {
-  const ihdr = new Uint8Array(13);
-  const iv = new DataView(ihdr.buffer);
-  iv.setUint32(0, width, false);
-  iv.setUint32(4, height, false);
-  ihdr[8] = 8;
-  ihdr[9] = 6;
-  const stride = width * 4;
-  const raw = new Uint8Array((stride + 1) * height);
-  for (let y = 0; y < height; y++) {
-    const rowStart = y * (stride + 1);
-    raw[rowStart] = 0;
-    raw.set(rgba.subarray(y * stride, y * stride + stride), rowStart + 1);
-  }
-  const idat = deflateSync(raw);
-  const ihdrChunk = chunk("IHDR", ihdr);
-  const idatChunk = chunk("IDAT", idat);
-  const iendChunk = chunk("IEND", new Uint8Array(0));
-  const out = new Uint8Array(PNG_SIGNATURE.length + ihdrChunk.length + idatChunk.length + iendChunk.length);
-  let o = 0;
-  out.set(PNG_SIGNATURE, o);
-  o += PNG_SIGNATURE.length;
-  out.set(ihdrChunk, o);
-  o += ihdrChunk.length;
-  out.set(idatChunk, o);
-  o += idatChunk.length;
-  out.set(iendChunk, o);
-  return out;
-}
-function escapeRegExp(s) {
-  return s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-}
-function inlineImagesIntoMarkdown(markdown, images, opts) {
-  const compress = opts?.compress !== false;
-  let out = markdown;
-  for (const img of images) {
-    let bytes = img.data;
-    let mime = img.mimeType;
-    if (compress && mime === "image/bmp") {
-      const png = bmpToPng(img.data);
-      if (png) {
-        bytes = png;
-        mime = "image/png";
-      }
-    }
-    const base64 = Buffer.from(bytes).toString("base64");
-    const dataUri = `data:${mime};base64,${base64}`;
-    const name = escapeRegExp(img.filename);
-    const mdRe = new RegExp(`!\\[image\\]\\((?:images/)?${name}\\)`, "g");
-    out = out.replace(mdRe, () => `![image](${dataUri})`);
-    const imgTagRe = new RegExp(`(<img\\b[^>]*\\bsrc=")(?:images/)?${name}(")`, "g");
-    out = out.replace(imgTagRe, (_m, pre, post) => `${pre}${dataUri}${post}`);
-  }
-  return out;
-}
-
 // src/hwp5/aes.ts
 var S_BOX = new Uint8Array([
   99,
@@ -4836,7 +4740,7 @@ function parseHwp5Document(buffer, options) {
     }
   }
   const blocks = [...doc.headerBlocks, ...bodyBlocks, ...doc.footerBlocks];
-  const images = cfb ? extractHwp5Images(cfb.FileIndex, blocks, warnings) : extractHwp5ImagesLenient(lenientCfb, blocks, warnings);
+  const images = cfb ? extractHwp5Images(cfb.FileIndex, blocks, warnings, !pageFilter) : extractHwp5ImagesLenient(lenientCfb, blocks, warnings, !pageFilter);
   let flatBlocks = flattenLayoutTables(blocks);
   if (options?.dedupeRunningHeaders) {
     const deduped = dedupeRunningHeaders(flatBlocks);
@@ -19297,6 +19201,102 @@ function extractRun(r) {
   }
   return { text, bold, italic };
 }
+function fieldUrlFromInstr(instr) {
+  if (!/HYPERLINK\b/i.test(instr)) return null;
+  const urlM = instr.match(/HYPERLINK\s+"([^"]*)"/i);
+  const anchorM = instr.match(/\\l\s+"([^"]*)"/i);
+  if (urlM && urlM[1]) {
+    return anchorM && anchorM[1] ? `${urlM[1]}#${anchorM[1]}` : urlM[1];
+  }
+  if (anchorM && anchorM[1]) return `#${anchorM[1]}`;
+  return null;
+}
+function makeLink(text, rawHref) {
+  if (!text) return "";
+  if (!rawHref) return text;
+  const href = sanitizeHref(rawHref);
+  return href ? `[${text}](${href})` : text;
+}
+function collectInline(p, footnotes, rels) {
+  const parts = [];
+  let bold = false;
+  let italic = false;
+  let footnoteText;
+  let fieldActive = false;
+  let fieldStage = "instr";
+  let fieldInstr = "";
+  let fieldDisplay = "";
+  const flushField = () => {
+    if (!fieldActive) return;
+    if (fieldDisplay) {
+      const url = fieldUrlFromInstr(fieldInstr);
+      parts.push(url ? makeLink(fieldDisplay, url) : fieldDisplay);
+    }
+    fieldActive = false;
+    fieldStage = "instr";
+    fieldInstr = "";
+    fieldDisplay = "";
+  };
+  for (const el of effectiveChildElements(p)) {
+    if (matchesLocal(el, "hyperlink")) {
+      const rId = getAttr(el, "id");
+      const anchor = getAttr(el, "anchor");
+      const t = findElements(el, "r").map((r) => extractRun(r).text).join("");
+      if (!t) continue;
+      let raw = null;
+      if (rId && rels.has(rId)) raw = rels.get(rId);
+      else if (anchor) raw = `#${anchor}`;
+      if (fieldActive && fieldStage === "display") fieldDisplay += t;
+      else parts.push(makeLink(t, raw));
+      continue;
+    }
+    if (matchesLocal(el, "fldSimple")) {
+      const t = findElements(el, "r").map((r) => extractRun(r).text).join("");
+      if (!t) continue;
+      const url = fieldUrlFromInstr(getAttr(el, "instr") ?? "");
+      parts.push(url ? makeLink(t, url) : t);
+      continue;
+    }
+    if (matchesLocal(el, "r")) {
+      for (const fc of getChildElements(el, "fldChar")) {
+        const ty = getAttr(fc, "fldCharType");
+        if (ty === "begin") {
+          flushField();
+          fieldActive = true;
+          fieldStage = "instr";
+        } else if (ty === "separate") {
+          if (fieldActive) fieldStage = "display";
+        } else if (ty === "end") flushField();
+      }
+      if (fieldActive && fieldStage === "instr") {
+        for (const it of getChildElements(el, "instrText")) fieldInstr += it.textContent ?? "";
+      }
+      const rr = extractRun(el);
+      if (rr.bold) bold = true;
+      if (rr.italic) italic = true;
+      const fnRefEls = getChildElements(el, "footnoteReference");
+      if (fnRefEls.length > 0) {
+        const fnId = getAttr(fnRefEls[0], "id");
+        if (fnId && footnotes.has(fnId)) footnoteText = footnotes.get(fnId);
+      }
+      if (rr.text) {
+        if (fieldActive && fieldStage === "display") fieldDisplay += rr.text;
+        else if (fieldActive && fieldStage === "instr") {
+        } else parts.push(rr.text);
+      }
+      continue;
+    }
+  }
+  flushField();
+  for (const om of collectOmmlRoots(p)) {
+    const latex = ommlElementToLatex(om);
+    if (!latex) continue;
+    if (isDisplayMath(om)) parts.push(" $$" + latex + "$$ ");
+    else parts.push(" $" + latex + "$ ");
+  }
+  const text = parts.join("").replace(/[ \t]{2,}/g, " ").trim();
+  return { text, bold, italic, footnoteText };
+}
 function parseParagraph2(p, styles, numbering, footnotes, rels) {
   const pPrEls = getChildElements(p, "pPr");
   let styleId = "";
@@ -19313,54 +19313,7 @@ function parseParagraph2(p, styles, numbering, footnotes, rels) {
       ilvl = ilvlEls.length > 0 ? parseInt(getAttr(ilvlEls[0], "val") ?? "0", 10) : 0;
     }
   }
-  const parts = [];
-  let hasBold = false;
-  let hasItalic = false;
-  let href;
-  let footnoteText;
-  const hyperlinks = getChildElements(p, "hyperlink");
-  const hyperlinkTexts = /* @__PURE__ */ new Set();
-  for (const hl of hyperlinks) {
-    const rId = getAttr(hl, "id");
-    const hlText = [];
-    const runs2 = findElements(hl, "r");
-    for (const r of runs2) {
-      const result = extractRun(r);
-      hlText.push(result.text);
-    }
-    const text2 = hlText.join("");
-    if (text2) {
-      hyperlinkTexts.add(text2);
-      if (rId && rels.has(rId)) {
-        href = rels.get(rId);
-        parts.push(text2);
-      } else {
-        parts.push(text2);
-      }
-    }
-  }
-  const runs = getChildElements(p, "r");
-  for (const r of runs) {
-    if (r.parentNode && r.parentNode.localName === "hyperlink") continue;
-    const result = extractRun(r);
-    if (result.bold) hasBold = true;
-    if (result.italic) hasItalic = true;
-    const fnRefEls = getChildElements(r, "footnoteReference");
-    if (fnRefEls.length > 0) {
-      const fnId = getAttr(fnRefEls[0], "id");
-      if (fnId && footnotes.has(fnId)) {
-        footnoteText = footnotes.get(fnId);
-      }
-    }
-    if (result.text) parts.push(result.text);
-  }
-  for (const om of collectOmmlRoots(p)) {
-    const latex = ommlElementToLatex(om);
-    if (!latex) continue;
-    if (isDisplayMath(om)) parts.push(" $$" + latex + "$$ ");
-    else parts.push(" $" + latex + "$ ");
-  }
-  const text = parts.join("").replace(/[ \t]{2,}/g, " ").trim();
+  const { text, bold: hasBold, italic: hasItalic, footnoteText } = collectInline(p, footnotes, rels);
   if (!text) return null;
   const style = styles.get(styleId);
   if (style?.outlineLevel !== void 0 && style.outlineLevel >= 0 && style.outlineLevel <= 5) {
@@ -19380,7 +19333,6 @@ function parseParagraph2(p, styles, numbering, footnotes, rels) {
   if (hasBold || hasItalic) {
     block.style = { bold: hasBold || void 0, italic: hasItalic || void 0 };
   }
-  if (href) block.href = href;
   if (footnoteText) block.footnoteText = footnoteText;
   return block;
 }
@@ -19481,46 +19433,81 @@ function collectCellText(tc2, styles, numbering, footnotes, rels, depth) {
   }
   return parts;
 }
-async function extractImages(zip, rels, doc, warnings) {
-  const blocks = [];
+async function buildImageMap(zip, rels, doc, warnings) {
+  const imageMap = /* @__PURE__ */ new Map();
   const images = [];
-  const drawingElements = findElements(doc.documentElement, "drawing");
   let imgIdx = 0;
-  for (const drawing of drawingElements) {
-    const blips = findElements(drawing, "blip");
-    for (const blip of blips) {
-      const embedId = getAttr(blip, "embed");
-      if (!embedId) continue;
-      const target = rels.get(embedId);
-      if (!target) continue;
-      const imgPath = target.startsWith("/") ? target.slice(1) : target.startsWith("word/") ? target : `word/${target}`;
-      const imgFile = zip.file(imgPath);
-      if (!imgFile) continue;
-      try {
-        const data = await imgFile.async("uint8array");
-        imgIdx++;
-        const ext = imgPath.split(".").pop()?.toLowerCase() ?? "png";
-        const mimeMap = {
-          png: "image/png",
-          jpg: "image/jpeg",
-          jpeg: "image/jpeg",
-          gif: "image/gif",
-          bmp: "image/bmp",
-          wmf: "image/wmf",
-          emf: "image/emf"
-        };
-        const filename = `image_${String(imgIdx).padStart(3, "0")}.${ext}`;
-        images.push({ filename, data, mimeType: mimeMap[ext] ?? "image/png" });
-        blocks.push({ type: "image", text: filename });
-      } catch (err) {
-        warnings.push({
-          code: "SKIPPED_IMAGE",
-          message: `DOCX \uC774\uBBF8\uC9C0 \uCD94\uCD9C \uC2E4\uD328 (${imgPath}): ${err instanceof Error ? err.message : String(err)}`
-        });
-      }
+  const imageRefs = [];
+  for (const blip of findElements(doc.documentElement, "blip")) {
+    const embedId = getAttr(blip, "embed");
+    if (embedId) imageRefs.push(embedId);
+  }
+  for (const imagedata of collectNonFallbackImagedata(doc.documentElement)) {
+    const embedId = getAttr(imagedata, "id");
+    if (embedId) imageRefs.push(embedId);
+  }
+  for (const embedId of imageRefs) {
+    if (imageMap.has(embedId)) continue;
+    const target = rels.get(embedId);
+    if (!target) continue;
+    const imgPath = target.startsWith("/") ? target.slice(1) : target.startsWith("word/") ? target : `word/${target}`;
+    const imgFile = zip.file(imgPath);
+    if (!imgFile) continue;
+    try {
+      const data = await imgFile.async("uint8array");
+      imgIdx++;
+      const ext = imgPath.split(".").pop()?.toLowerCase() ?? "png";
+      const mimeMap = {
+        png: "image/png",
+        jpg: "image/jpeg",
+        jpeg: "image/jpeg",
+        gif: "image/gif",
+        bmp: "image/bmp",
+        wmf: "image/wmf",
+        emf: "image/emf"
+      };
+      const filename = `image_${String(imgIdx).padStart(3, "0")}.${ext}`;
+      images.push({ filename, data, mimeType: mimeMap[ext] ?? "image/png" });
+      imageMap.set(embedId, filename);
+    } catch (err) {
+      warnings.push({
+        code: "SKIPPED_IMAGE",
+        message: `DOCX \uC774\uBBF8\uC9C0 \uCD94\uCD9C \uC2E4\uD328 (${imgPath}): ${err instanceof Error ? err.message : String(err)}`
+      });
     }
   }
-  return { blocks, images };
+  return { imageMap, images };
+}
+function collectParagraphBlips(node, out = [], depth = 0) {
+  if (depth > 40) return out;
+  for (const el of effectiveChildElements(node)) {
+    if (matchesLocal(el, "txbxContent") || matchesLocal(el, "Fallback")) continue;
+    if (matchesLocal(el, "blip") || matchesLocal(el, "imagedata")) out.push(el);
+    else collectParagraphBlips(el, out, depth + 1);
+  }
+  return out;
+}
+function collectNonFallbackImagedata(node, out = [], depth = 0) {
+  if (depth > 60) return out;
+  const children = node.childNodes;
+  for (let i = 0; i < children.length; i++) {
+    if (children[i].nodeType !== 1) continue;
+    const el = children[i];
+    if (matchesLocal(el, "Fallback")) continue;
+    if (matchesLocal(el, "imagedata")) out.push(el);
+    else collectNonFallbackImagedata(el, out, depth + 1);
+  }
+  return out;
+}
+function emitParagraphImages(p, imageMap, linked, out) {
+  for (const blip of collectParagraphBlips(p)) {
+    const embedId = getAttr(blip, "embed") ?? getAttr(blip, "id");
+    if (!embedId) continue;
+    const filename = imageMap.get(embedId);
+    if (!filename) continue;
+    linked.add(embedId);
+    out.push({ type: "image", text: filename });
+  }
 }
 async function parseDocxDocument(buffer, options) {
   precheckZipSize(buffer, MAX_DECOMPRESS_SIZE4);
@@ -19577,6 +19564,8 @@ 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 linkedImages = /* @__PURE__ */ new Set();
   const blocks = [];
   const bodyEl = body[0];
   const topLevel = effectiveChildElements(bodyEl);
@@ -19585,16 +19574,20 @@ async function parseDocxDocument(buffer, options) {
     if (localName2 === "p") {
       const block = parseParagraph2(el, styles, numbering, footnotes, rels);
       if (block) blocks.push(block);
+      if (imageMap.size > 0) emitParagraphImages(el, imageMap, linkedImages, blocks);
       for (const tp of collectTextboxParagraphs(el)) {
         const tb = parseParagraph2(tp, styles, numbering, footnotes, rels);
         if (tb) blocks.push(tb);
+        if (imageMap.size > 0) emitParagraphImages(tp, imageMap, linkedImages, blocks);
       }
     } else if (localName2 === "tbl") {
       const block = parseTable(el, styles, numbering, footnotes, rels);
       if (block) blocks.push(block);
     }
   }
-  const { blocks: imgBlocks, images } = await extractImages(zip, rels, doc, warnings);
+  for (const [embedId, filename] of imageMap) {
+    if (!linkedImages.has(embedId)) blocks.push({ type: "image", text: filename });
+  }
   const metadata = {};
   const coreFile = zip.file("docProps/core.xml");
   if (coreFile) {
@@ -21201,7 +21194,7 @@ function parseCentralDirectory(buf) {
   }
   return { entries, cdOffset, cdSize, eocdOffset };
 }
-var CRC_TABLE2 = (() => {
+var CRC_TABLE = (() => {
   const table2 = new Uint32Array(256);
   for (let n = 0; n < 256; n++) {
     let c = n;
@@ -21210,10 +21203,10 @@ var CRC_TABLE2 = (() => {
   }
   return table2;
 })();
-function crc322(data) {
+function crc32(data) {
   let crc = 4294967295;
   for (let i = 0; i < data.length; i++) {
-    crc = CRC_TABLE2[(crc ^ data[i]) & 255] ^ crc >>> 8;
+    crc = CRC_TABLE[(crc ^ data[i]) & 255] ^ crc >>> 8;
   }
   return (crc ^ 4294967295) >>> 0;
 }
@@ -21252,7 +21245,7 @@ function patchZipEntries(original, replacements, additions) {
     const hview = new DataView(header.buffer, header.byteOffset, header.byteLength);
     const method = e.method;
     const compData = method === 0 ? newData : new Uint8Array(deflateRawSync(newData));
-    const crc = crc322(newData);
+    const crc = crc32(newData);
     const flags = e.flags & ~8;
     hview.setUint16(6, flags, true);
     hview.setUint32(14, crc, true);
@@ -21270,7 +21263,7 @@ function patchZipEntries(original, replacements, additions) {
       const deflated = new Uint8Array(deflateRawSync(data));
       const method = deflated.length < data.length ? 8 : 0;
       const compData = method === 8 ? deflated : data;
-      const crc = crc322(data);
+      const crc = crc32(data);
       const header = new Uint8Array(30 + nameBytes.length);
       const hv = new DataView(header.buffer);
       hv.setUint32(0, LOCAL_SIG, true);
@@ -26812,16 +26805,16 @@ function buildOrigUnits(blocks) {
   for (let i = 0; i < blocks.length; i++) {
     const block = blocks[i];
     let consume = 1;
-    let chunk2;
+    let chunk;
     if (block.type === "paragraph" && block.text && /^\[별표\s*\d+/.test(sanitizeText(block.text))) {
       const next = blocks[i + 1];
       if (next?.type === "paragraph" && next.text && /관련\)?$/.test(next.text)) consume = 2;
-      chunk2 = blocksToMarkdown(blocks.slice(i, i + consume));
+      chunk = blocksToMarkdown(blocks.slice(i, i + consume));
     } else {
-      chunk2 = blocksToMarkdown([block]);
+      chunk = blocksToMarkdown([block]);
     }
-    if (chunk2) {
-      const subUnits = splitMarkdownUnits(chunk2);
+    if (chunk) {
+      const subUnits = splitMarkdownUnits(chunk);
       const isFragment = consume === 2 || (block.type === "paragraph" || block.type === "heading") && subUnits.length > 1;
       for (let s = 0; s < subUnits.length; s++) {
         const u = { ...subUnits[s], blockIdx: i, fragment: isFragment || void 0 };
@@ -29814,7 +29807,7 @@ async function parseHwp(buffer, options) {
 async function parsePdf(buffer, options) {
   let parsePdfDocument;
   try {
-    const mod = await import("./parser-DMKHIV4E.js");
+    const mod = await import("./parser-KMMUS73Q.js");
     parsePdfDocument = mod.parsePdfDocument;
   } catch {
     return {
@@ -29825,8 +29818,8 @@ async function parsePdf(buffer, options) {
     };
   }
   try {
-    const { markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary } = await parsePdfDocument(buffer, options);
-    return { success: true, fileType: "pdf", markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary };
+    const { markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary, images } = await parsePdfDocument(buffer, options);
+    return { success: true, fileType: "pdf", markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary, images };
   } catch (err) {
     const isImageBased = err instanceof Error && "isImageBased" in err ? true : void 0;
     return { success: false, fileType: "pdf", error: err instanceof Error ? err.message : "PDF \uD30C\uC2F1 \uC2E4\uD328", code: classifyError(err), isImageBased };