kordoc 4.0.6 → 4.0.8

package/dist/{parser-DMKHIV4E.js → parser-KMMUS73Q.js}

@@ -4,9 +4,10 @@ import {
   HEADING_RATIO_H3,
   KordocError,
   blocksToMarkdown,
+  encodePng,
   safeMax,
   safeMin
-} from "./chunk-HOGJ5ALY.js";
+} from "./chunk-OQPILS7B.js";
 import {
   parsePageRange
 } from "./chunk-GE43BE46.js";
@@ -1220,10 +1221,180 @@ function normalizeUndersegmentedTable(originalCells, colXs, items) {
   return rebuilt;
 }
 
+// src/pdf/image-extract.ts
+import { OPS as OPS3, ImageKind } from "pdfjs-dist/legacy/build/pdf.mjs";
+var IMAGE_RESOLVE_TIMEOUT_MS = 5e3;
+function resolveImgData(page, objId) {
+  const store = objId.startsWith("g_") ? page.commonObjs : page.objs;
+  return new Promise((resolve) => {
+    let done = false;
+    const timer = setTimeout(() => {
+      if (!done) {
+        done = true;
+        resolve(null);
+      }
+    }, IMAGE_RESOLVE_TIMEOUT_MS);
+    try {
+      store.get(objId, (data) => {
+        if (!done) {
+          done = true;
+          clearTimeout(timer);
+          resolve(data ?? null);
+        }
+      });
+    } catch {
+      if (!done) {
+        done = true;
+        clearTimeout(timer);
+        resolve(null);
+      }
+    }
+  });
+}
+var MIN_DIM = 8;
+var MAX_PIXELS = 36e6;
+var MAX_IMAGES_PER_DOC = 200;
+var MAX_TOTAL_IMAGE_BYTES = 128 * 1024 * 1024;
+function createPdfImageState() {
+  return { imageIndex: 0, totalBytes: 0, seen: /* @__PURE__ */ new Map(), capWarned: false };
+}
+function fnv1a(data) {
+  let h = 2166136261;
+  for (let i = 0; i < data.length; i++) {
+    h ^= data[i];
+    h = Math.imul(h, 16777619);
+  }
+  return h >>> 0;
+}
+function toRgba(img) {
+  const { width: w, height: h, kind, data } = img;
+  if (!data || !w || !h) return null;
+  const rgba = new Uint8Array(w * h * 4);
+  if (kind === ImageKind.RGBA_32BPP) {
+    if (data.length < w * h * 4) return null;
+    rgba.set(data.subarray(0, w * h * 4));
+    return rgba;
+  }
+  if (kind === ImageKind.RGB_24BPP) {
+    if (data.length < w * h * 3) return null;
+    for (let i = 0, s = 0, d = 0; i < w * h; i++, s += 3, d += 4) {
+      rgba[d] = data[s];
+      rgba[d + 1] = data[s + 1];
+      rgba[d + 2] = data[s + 2];
+      rgba[d + 3] = 255;
+    }
+    return rgba;
+  }
+  if (kind === ImageKind.GRAYSCALE_1BPP) {
+    const stride = w + 7 >> 3;
+    if (data.length < stride * h) return null;
+    for (let y = 0; y < h; y++) {
+      for (let x = 0; x < w; x++) {
+        const bit = data[y * stride + (x >> 3)] >> 7 - (x & 7) & 1;
+        const v = bit ? 255 : 0;
+        const d = (y * w + x) * 4;
+        rgba[d] = v;
+        rgba[d + 1] = v;
+        rgba[d + 2] = v;
+        rgba[d + 3] = 255;
+      }
+    }
+    return rgba;
+  }
+  return null;
+}
+async function extractPageImages(page, fnArray, argsArray, pageNumber, state, warnings) {
+  const blocks = [];
+  const images = [];
+  const pageSeenIds = /* @__PURE__ */ new Set();
+  for (let i = 0; i < fnArray.length; i++) {
+    const op = fnArray[i];
+    let imgData = null;
+    let dedupeId = null;
+    if (op === OPS3.paintImageXObject || op === OPS3.paintImageXObjectRepeat) {
+      const objId = argsArray[i]?.[0];
+      if (typeof objId !== "string" || pageSeenIds.has(objId)) continue;
+      pageSeenIds.add(objId);
+      dedupeId = objId;
+      imgData = await resolveImgData(page, objId);
+    } else if (op === OPS3.paintInlineImageXObject) {
+      imgData = argsArray[i]?.[0];
+    } else {
+      continue;
+    }
+    if (!imgData?.data || !imgData.width || !imgData.height) continue;
+    const { width: w, height: h } = imgData;
+    if (w < MIN_DIM || h < MIN_DIM) continue;
+    if (w * h > MAX_PIXELS) {
+      warnings.push({ page: pageNumber, message: `\uC774\uBBF8\uC9C0 \uD06C\uAE30 \uCD08\uACFC\uB85C \uCD94\uCD9C \uC0DD\uB7B5 (${w}\xD7${h})`, code: "SKIPPED_IMAGE" });
+      continue;
+    }
+    const hash = `${w}x${h}k${imgData.kind ?? "?"}h${fnv1a(imgData.data)}${dedupeId ? "" : "inline"}`;
+    if (state.seen.has(hash)) continue;
+    if (state.imageIndex >= MAX_IMAGES_PER_DOC || state.totalBytes >= MAX_TOTAL_IMAGE_BYTES) {
+      if (!state.capWarned) {
+        state.capWarned = true;
+        warnings.push({ page: pageNumber, message: `\uC774\uBBF8\uC9C0 \uCD94\uCD9C \uC0C1\uD55C \uB3C4\uB2EC (${MAX_IMAGES_PER_DOC}\uAC1C/128MB) \u2014 \uC774\uD6C4 \uC774\uBBF8\uC9C0 \uC0DD\uB7B5`, code: "SKIPPED_IMAGE" });
+      }
+      continue;
+    }
+    const rgba = toRgba(imgData);
+    if (!rgba) continue;
+    const png = encodePng(w, h, rgba);
+    state.imageIndex++;
+    state.totalBytes += png.length;
+    const filename = `image_${String(state.imageIndex).padStart(3, "0")}.png`;
+    state.seen.set(hash, filename);
+    images.push({ filename, data: png, mimeType: "image/png" });
+    blocks.push({ type: "image", text: filename, pageNumber });
+  }
+  return { blocks, images };
+}
+function injectPageImageBlocks(blocks, pageImages) {
+  if (pageImages.size === 0) return;
+  const lastIndexForPage = /* @__PURE__ */ new Map();
+  for (let i = 0; i < blocks.length; i++) {
+    const p = blocks[i].pageNumber;
+    if (p !== void 0) lastIndexForPage.set(p, i);
+  }
+  const pageAtIndex = /* @__PURE__ */ new Map();
+  for (const [p, last] of lastIndexForPage) pageAtIndex.set(last, p);
+  const result = [];
+  const injected = /* @__PURE__ */ new Set();
+  for (let i = 0; i < blocks.length; i++) {
+    result.push(blocks[i]);
+    const p = pageAtIndex.get(i);
+    if (p !== void 0 && pageImages.has(p)) {
+      result.push(...pageImages.get(p));
+      injected.add(p);
+    }
+  }
+  for (const p of [...pageImages.keys()].sort((a, b) => a - b)) {
+    if (injected.has(p)) continue;
+    let at = result.length;
+    for (let i = 0; i < result.length; i++) {
+      const bp = result[i].pageNumber;
+      if (bp !== void 0 && bp > p) {
+        at = i;
+        break;
+      }
+    }
+    result.splice(at, 0, ...pageImages.get(p));
+  }
+  blocks.length = 0;
+  blocks.push(...result);
+}
+
 // src/pdf/quality.ts
+var RARE_JONG = /* @__PURE__ */ new Set([3, 5, 6, 9, 10, 11, 12, 13, 14, 15, 18, 24, 25, 26]);
+var MOJIBAKE_MIN_HANGUL = 30;
+var MOJIBAKE_MAX_NOBATCHIM = 0.15;
+var MOJIBAKE_MIN_RAREBATCHIM = 0.25;
 function computePageQuality(page, text) {
   let total = 0;
   let hangul = 0;
+  let hangulNoBatchim = 0;
+  let hangulRareBatchim = 0;
   let control = 0;
   let replacement = 0;
   let pua = 0;
@@ -1241,6 +1412,9 @@ function computePageQuality(page, text) {
     }
     if (code >= 44032 && code <= 55203) {
       hangul++;
+      const jong = (code - 44032) % 28;
+      if (jong === 0) hangulNoBatchim++;
+      else if (RARE_JONG.has(jong)) hangulRareBatchim++;
       continue;
     }
     if (code >= 57344 && code <= 63743 || code >= 56192 && code <= 56319) {
@@ -1252,6 +1426,9 @@ function computePageQuality(page, text) {
   const puaRatio = pua / denom;
   const controlCharRatio = control / denom;
   const replacementCharRatio = replacement / denom;
+  const hangulNoBatchimRatio = hangul > 0 ? hangulNoBatchim / hangul : 0;
+  const hangulRareBatchimRatio = hangul > 0 ? hangulRareBatchim / hangul : 0;
+  const garbledHangul = hangul >= MOJIBAKE_MIN_HANGUL && hangulNoBatchimRatio < MOJIBAKE_MAX_NOBATCHIM && hangulRareBatchimRatio >= MOJIBAKE_MIN_RAREBATCHIM;
   let needsOcr = false;
   let ocrReason;
   if (total < LOW_TEXT_THRESHOLD) {
@@ -1266,6 +1443,9 @@ function computePageQuality(page, text) {
   } else if (replacementCharRatio >= HIGH_REPLACEMENT_THRESHOLD) {
     needsOcr = true;
     ocrReason = "high_replacement";
+  } else if (garbledHangul) {
+    needsOcr = true;
+    ocrReason = "garbled_hangul";
   }
   return {
     page,
@@ -1274,6 +1454,8 @@ function computePageQuality(page, text) {
     controlCharRatio,
     replacementCharRatio,
     puaRatio,
+    hangulNoBatchimRatio,
+    hangulRareBatchimRatio,
     needsOcr,
     ocrReason
   };
@@ -3565,6 +3747,9 @@ async function parsePdfDocument(buffer, options) {
     const pageHeights = /* @__PURE__ */ new Map();
     const pagesWithLargeImage = /* @__PURE__ */ new Set();
     const skippedImagePages = /* @__PURE__ */ new Map();
+    const imageState = createPdfImageState();
+    const extractedImages = [];
+    const pageImageBlocks = /* @__PURE__ */ new Map();
     let parsedPages = 0;
     for (let i = 1; i <= effectivePageCount; i++) {
       if (pageFilter && !pageFilter.has(i)) continue;
@@ -3602,6 +3787,12 @@ async function parsePdfDocument(buffer, options) {
         }
         const pageBlocks = extractPageBlocksWithLines(visible, i, opList, viewport.width, viewport.height);
         for (const b of pageBlocks) blocks.push(b);
+        try {
+          const { blocks: imgBlocks, images: pageImages } = await extractPageImages(page, opList.fnArray, opList.argsArray, i, imageState, warnings);
+          if (imgBlocks.length > 0) pageImageBlocks.set(i, imgBlocks);
+          extractedImages.push(...pageImages);
+        } catch {
+        }
         let pageText = "";
         for (const b of pageBlocks) {
           const t = b.text || "";
@@ -3623,11 +3814,11 @@ async function parsePdfDocument(buffer, options) {
     if (totalChars / Math.max(parsedPageCount, 1) < 10) {
       if (options?.ocr) {
         try {
-          const { ocrPages } = await import("./provider-H4WWSPOV.js");
+          const { ocrPages } = await import("./provider-4ZJKV3DC.js");
           const ocrBlocks = await ocrPages(doc, options.ocr, pageFilter, effectivePageCount);
           if (ocrBlocks.length > 0) {
             const ocrMarkdown = ocrBlocks.map((b) => b.text || "").filter(Boolean).join("\n\n");
-            return { markdown: ocrMarkdown, blocks: ocrBlocks, metadata, warnings, isImageBased: true, pageQuality, qualitySummary: summarizeDocumentQuality(pageQuality) };
+            return { markdown: ocrMarkdown, blocks: ocrBlocks, metadata, warnings, isImageBased: true, pageQuality, qualitySummary: summarizeDocumentQuality(pageQuality), images: extractedImages.length > 0 ? extractedImages : void 0 };
           }
         } catch {
         }
@@ -3643,7 +3834,8 @@ async function parsePdfDocument(buffer, options) {
         low_text: "\uD14D\uC2A4\uD2B8\uAC00 \uAC70\uC758 \uC5C6\uB294 \uD398\uC774\uC9C0 (\uC2A4\uCE94/\uC774\uBBF8\uC9C0 \uCD94\uC815)",
         high_pua: "\uAE00\uAF34 \uB9E4\uD551 \uC2E4\uD328 (PUA \uBE44\uC728 \uB192\uC74C) \u2014 \uCD94\uCD9C \uD14D\uC2A4\uD2B8 \uC2E0\uB8B0 \uBD88\uAC00",
         high_control: "\uC81C\uC5B4\uBB38\uC790 \uBE44\uC728 \uB192\uC74C \u2014 \uCD94\uCD9C \uD14D\uC2A4\uD2B8 \uC2E0\uB8B0 \uBD88\uAC00",
-        high_replacement: "\uB300\uCCB4\uBB38\uC790(U+FFFD) \uBE44\uC728 \uB192\uC74C \u2014 \uCD94\uCD9C \uD14D\uC2A4\uD2B8 \uC2E0\uB8B0 \uBD88\uAC00"
+        high_replacement: "\uB300\uCCB4\uBB38\uC790(U+FFFD) \uBE44\uC728 \uB192\uC74C \u2014 \uCD94\uCD9C \uD14D\uC2A4\uD2B8 \uC2E0\uB8B0 \uBD88\uAC00",
+        garbled_hangul: "\uAE00\uAF34 \uB9E4\uD551 \uC2E4\uD328 (\uD55C\uAE00 \uC790\uC18C \uBD84\uD3EC \uC774\uC0C1) \u2014 \uCD94\uCD9C \uD14D\uC2A4\uD2B8\uAC00 \uAE68\uC84C\uC744 \uC218 \uC788\uC74C"
       };
       for (const pq of pageQuality) {
         if (!pq.needsOcr || !pq.ocrReason) continue;
@@ -3663,6 +3855,7 @@ async function parsePdfDocument(buffer, options) {
       }
     }
     mergeCrossPageTables(blocks);
+    injectPageImageBlocks(blocks, pageImageBlocks);
     if (options?.formulaOcr && formulaBuffer) {
       try {
         await applyFormulaOcr(formulaBuffer, blocks, pageFilter, effectivePageCount, warnings, options.onProgress);
@@ -3691,7 +3884,8 @@ async function parsePdfDocument(buffer, options) {
       warnings: warnings.length > 0 ? warnings : void 0,
       isImageBased: isImageBased || void 0,
       pageQuality,
-      qualitySummary: summarizeDocumentQuality(pageQuality)
+      qualitySummary: summarizeDocumentQuality(pageQuality),
+      images: extractedImages.length > 0 ? extractedImages : void 0
     };
   } finally {
     await doc.destroy().catch(() => {
@@ -3741,4 +3935,4 @@ export {
   parsePdfDocument,
   removeHeaderFooterBlocks
 };
-//# sourceMappingURL=parser-DMKHIV4E.js.map
+//# sourceMappingURL=parser-KMMUS73Q.js.map