kordoc 2.7.2 → 2.9.0

package/dist/mcp.js

@@ -8,7 +8,7 @@ import {
   fillHwpx,
   markdownToHwpx,
   parse
-} from "./chunk-4NWDJGAU.js";
+} from "./chunk-M24KMDAR.js";
 import {
   detectFormat,
   detectZipFormat
@@ -19,7 +19,7 @@ import {
   blocksToMarkdown,
   sanitizeError,
   toArrayBuffer
-} from "./chunk-4SK2PDMQ.js";
+} from "./chunk-SJ5TPMBT.js";
 import "./chunk-MOL7MDBG.js";
 
 // src/mcp.ts
@@ -191,7 +191,7 @@ server.tool(
           break;
         case "pdf":
           try {
-            const { extractPdfMetadataOnly } = await import("./parser-QMMQ7Y7R.js");
+            const { extractPdfMetadataOnly } = await import("./parser-XBYGROQB.js");
             metadata = await extractPdfMetadataOnly(buffer);
           } catch {
             metadata = void 0;

package/dist/{parser-7OFQ67QL.cjs → parser-EL5YETUA.cjs}

@@ -6,7 +6,7 @@
 
 
 
-var _chunkY476BOHIcjs = require('./chunk-Y476BOHI.cjs');
+var _chunkQB7CS534cjs = require('./chunk-QB7CS534.cjs');
 
 
 var _chunkMUOQXDZ4cjs = require('./chunk-MUOQXDZ4.cjs');
@@ -1142,6 +1142,120 @@ function buildClusterTable(rows, columns, pageNum) {
   };
 }
 
+// src/pdf/quality.ts
+function computePageQuality(page, text) {
+  let total = 0;
+  let hangul = 0;
+  let control = 0;
+  let replacement = 0;
+  let pua = 0;
+  for (let i = 0; i < text.length; i++) {
+    const code = text.charCodeAt(i);
+    if (code === 32 || code === 9 || code === 10 || code === 13) continue;
+    total++;
+    if (code < 32 || code === 127 || code >= 128 && code <= 159) {
+      control++;
+      continue;
+    }
+    if (code === 65533) {
+      replacement++;
+      continue;
+    }
+    if (code >= 44032 && code <= 55203) {
+      hangul++;
+      continue;
+    }
+    if (code >= 57344 && code <= 63743 || code >= 56192 && code <= 56319) {
+      pua++;
+      continue;
+    }
+  }
+  const denom = total || 1;
+  const puaRatio = pua / denom;
+  const controlCharRatio = control / denom;
+  const replacementCharRatio = replacement / denom;
+  let needsOcr = false;
+  let ocrReason;
+  if (total < LOW_TEXT_THRESHOLD) {
+    needsOcr = true;
+    ocrReason = "low_text";
+  } else if (puaRatio >= HIGH_PUA_THRESHOLD) {
+    needsOcr = true;
+    ocrReason = "high_pua";
+  } else if (controlCharRatio >= HIGH_CONTROL_THRESHOLD) {
+    needsOcr = true;
+    ocrReason = "high_control";
+  } else if (replacementCharRatio >= HIGH_REPLACEMENT_THRESHOLD) {
+    needsOcr = true;
+    ocrReason = "high_replacement";
+  }
+  return {
+    page,
+    textChars: total,
+    hangulRatio: hangul / denom,
+    controlCharRatio,
+    replacementCharRatio,
+    puaRatio,
+    needsOcr,
+    ocrReason
+  };
+}
+var LOW_TEXT_THRESHOLD = 20;
+var HIGH_PUA_THRESHOLD = 0.2;
+var HIGH_CONTROL_THRESHOLD = 0.05;
+var HIGH_REPLACEMENT_THRESHOLD = 0.05;
+var DOC_NEEDS_OCR_PAGE_RATIO = 0.3;
+function stripControlChars(text) {
+  return text.replace(/[\x00-\x08\x0B\x0C\x0E-\x1F\x7F\x80-\x9F]/g, "");
+}
+function summarizeDocumentQuality(pages) {
+  if (pages.length === 0) {
+    return {
+      totalPages: 0,
+      totalTextChars: 0,
+      avgHangulRatio: 0,
+      avgControlCharRatio: 0,
+      avgReplacementCharRatio: 0,
+      avgPuaRatio: 0,
+      lowTextPageCount: 0,
+      highPuaPageCount: 0,
+      needsOcr: false,
+      ocrCandidatePages: []
+    };
+  }
+  let textChars = 0;
+  let hangul = 0;
+  let control = 0;
+  let replacement = 0;
+  let pua = 0;
+  let lowText = 0;
+  let highPua = 0;
+  const ocrCandidatePages = [];
+  for (const p of pages) {
+    textChars += p.textChars;
+    hangul += p.hangulRatio;
+    control += p.controlCharRatio;
+    replacement += p.replacementCharRatio;
+    pua += p.puaRatio;
+    if (p.textChars < LOW_TEXT_THRESHOLD) lowText++;
+    if (p.puaRatio >= HIGH_PUA_THRESHOLD) highPua++;
+    if (p.needsOcr) ocrCandidatePages.push(p.page);
+  }
+  const n = pages.length;
+  return {
+    totalPages: n,
+    totalTextChars: textChars,
+    avgHangulRatio: hangul / n,
+    avgControlCharRatio: control / n,
+    avgReplacementCharRatio: replacement / n,
+    avgPuaRatio: pua / n,
+    lowTextPageCount: lowText,
+    highPuaPageCount: highPua,
+    needsOcr: ocrCandidatePages.length / n >= DOC_NEEDS_OCR_PAGE_RATIO,
+    ocrCandidatePages
+  };
+}
+
 // src/pdf/polyfill.ts
 var _pdfworkermjs = require('pdfjs-dist/legacy/build/pdf.worker.mjs'); var pdfjsWorker = _interopRequireWildcard(_pdfworkermjs);
 var g = globalThis;
@@ -1179,7 +1293,7 @@ async function loadPdfWithTimeout(buffer) {
       new Promise((_, reject) => {
         timer = setTimeout(() => {
           loadingTask.destroy();
-          reject(new (0, _chunkY476BOHIcjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
+          reject(new (0, _chunkQB7CS534cjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
         }, PDF_LOAD_TIMEOUT_MS);
       })
     ]);
@@ -1192,11 +1306,12 @@ async function parsePdfDocument(buffer, options) {
   const doc = await loadPdfWithTimeout(buffer);
   try {
     const pageCount = doc.numPages;
-    if (pageCount === 0) throw new (0, _chunkY476BOHIcjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
+    if (pageCount === 0) throw new (0, _chunkQB7CS534cjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
     const metadata = { pageCount };
     await extractPdfMetadata(doc, metadata);
     const blocks = [];
     const warnings = [];
+    const pageQuality = [];
     let totalChars = 0;
     let totalTextBytes = 0;
     const effectivePageCount = Math.min(pageCount, MAX_PAGES);
@@ -1224,16 +1339,19 @@ async function parsePdfDocument(buffer, options) {
         const opList = await page.getOperatorList();
         const pageBlocks = extractPageBlocksWithLines(visible, i, opList, viewport.width, viewport.height);
         for (const b of pageBlocks) blocks.push(b);
+        let pageText = "";
         for (const b of pageBlocks) {
           const t = b.text || "";
           totalChars += t.replace(/\s/g, "").length;
           totalTextBytes += t.length * 2;
+          pageText += pageText ? "\n" + t : t;
         }
-        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkY476BOHIcjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
+        pageQuality.push(computePageQuality(i, pageText));
+        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkQB7CS534cjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
         parsedPages++;
         _optionalChain([options, 'optionalAccess', _12 => _12.onProgress, 'optionalCall', _13 => _13(parsedPages, totalTarget)]);
       } catch (pageErr) {
-        if (pageErr instanceof _chunkY476BOHIcjs.KordocError) throw pageErr;
+        if (pageErr instanceof _chunkQB7CS534cjs.KordocError) throw pageErr;
         warnings.push({ page: i, message: `\uD398\uC774\uC9C0 ${i} \uD30C\uC2F1 \uC2E4\uD328: ${pageErr instanceof Error ? pageErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
       }
     }
@@ -1245,12 +1363,12 @@ async function parsePdfDocument(buffer, options) {
           const ocrBlocks = await ocrPages(doc, options.ocr, pageFilter, effectivePageCount);
           if (ocrBlocks.length > 0) {
             const ocrMarkdown = ocrBlocks.map((b) => b.text || "").filter(Boolean).join("\n\n");
-            return { markdown: ocrMarkdown, blocks: ocrBlocks, metadata, warnings, isImageBased: true };
+            return { markdown: ocrMarkdown, blocks: ocrBlocks, metadata, warnings, isImageBased: true, pageQuality, qualitySummary: summarizeDocumentQuality(pageQuality) };
           }
         } catch (e2) {
         }
       }
-      throw Object.assign(new (0, _chunkY476BOHIcjs.KordocError)(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
+      throw Object.assign(new (0, _chunkQB7CS534cjs.KordocError)(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
     }
     if (_optionalChain([options, 'optionalAccess', _15 => _15.removeHeaderFooter]) !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);
@@ -1274,8 +1392,17 @@ async function parsePdfDocument(buffer, options) {
     }
     detectMarkerHeadings(blocks);
     const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
-    let markdown = cleanPdfText(_chunkY476BOHIcjs.blocksToMarkdown.call(void 0, blocks));
-    return { markdown, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0 };
+    sanitizeBlockControlChars(blocks);
+    let markdown = cleanPdfText(_chunkQB7CS534cjs.blocksToMarkdown.call(void 0, blocks));
+    return {
+      markdown,
+      blocks,
+      metadata,
+      outline: outline.length > 0 ? outline : void 0,
+      warnings: warnings.length > 0 ? warnings : void 0,
+      pageQuality,
+      qualitySummary: summarizeDocumentQuality(pageQuality)
+    };
   } finally {
     await doc.destroy().catch(() => {
     });
@@ -1353,9 +1480,9 @@ function detectHeadings(blocks, medianFontSize) {
     if (/^\d+$/.test(text)) continue;
     const ratio = block.style.fontSize / medianFontSize;
     let level = 0;
-    if (ratio >= _chunkY476BOHIcjs.HEADING_RATIO_H1) level = 1;
-    else if (ratio >= _chunkY476BOHIcjs.HEADING_RATIO_H2) level = 2;
-    else if (ratio >= _chunkY476BOHIcjs.HEADING_RATIO_H3) level = 3;
+    if (ratio >= _chunkQB7CS534cjs.HEADING_RATIO_H1) level = 1;
+    else if (ratio >= _chunkQB7CS534cjs.HEADING_RATIO_H2) level = 2;
+    else if (ratio >= _chunkQB7CS534cjs.HEADING_RATIO_H3) level = 3;
     if (level > 0) {
       block.type = "heading";
       block.level = level;
@@ -1603,7 +1730,7 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
     }
     if (remaining.length > 0) {
       const allY = remaining.map((i) => i.y);
-      const pageH = _chunkY476BOHIcjs.safeMax.call(void 0, allY) - _chunkY476BOHIcjs.safeMin.call(void 0, allY);
+      const pageH = _chunkQB7CS534cjs.safeMax.call(void 0, allY) - _chunkQB7CS534cjs.safeMin.call(void 0, allY);
       const groups = xyCutOrder(remaining, Math.max(15, pageH * 0.03));
       const textBlocks = [];
       for (const group of groups) {
@@ -1691,7 +1818,7 @@ function extractPageBlocksFallback(items, pageNum) {
       blocks.push({ type: "paragraph", text: tableText, pageNumber: pageNum, bbox, style: dominantStyle(items) });
     } else {
       const allY = items.map((i) => i.y);
-      const pageHeight = _chunkY476BOHIcjs.safeMax.call(void 0, allY) - _chunkY476BOHIcjs.safeMin.call(void 0, allY);
+      const pageHeight = _chunkQB7CS534cjs.safeMax.call(void 0, allY) - _chunkQB7CS534cjs.safeMin.call(void 0, allY);
       const gapThreshold = Math.max(15, pageHeight * 0.03);
       const orderedGroups = xyCutOrder(items, gapThreshold);
       for (const group of orderedGroups) {
@@ -1838,14 +1965,14 @@ function isProseSpread(items) {
   for (let i = 1; i < sorted.length; i++) {
     gaps.push(sorted[i].x - (sorted[i - 1].x + sorted[i - 1].w));
   }
-  const maxGap = _chunkY476BOHIcjs.safeMax.call(void 0, gaps);
+  const maxGap = _chunkQB7CS534cjs.safeMax.call(void 0, gaps);
   const avgLen = items.reduce((s, i) => s + i.text.length, 0) / items.length;
   return maxGap < 40 && avgLen < 5;
 }
 function detectColumns(yLines) {
   const allItems = yLines.flat();
   if (allItems.length === 0) return null;
-  const pageWidth = _chunkY476BOHIcjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkY476BOHIcjs.safeMin.call(void 0, allItems.map((i) => i.x));
+  const pageWidth = _chunkQB7CS534cjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkQB7CS534cjs.safeMin.call(void 0, allItems.map((i) => i.x));
   if (pageWidth < 100) return null;
   let bigoLineIdx = -1;
   for (let i = 0; i < yLines.length; i++) {
@@ -2079,9 +2206,22 @@ function mergeLineSimple(items) {
   }
   return result;
 }
+function sanitizeBlockControlChars(blocks) {
+  for (const b of blocks) {
+    if (b.text) b.text = stripControlChars(b.text);
+    if (b.table) {
+      for (const row of b.table.cells) {
+        for (const cell of row) {
+          if (cell.text) cell.text = stripControlChars(cell.text);
+        }
+      }
+    }
+    if (b.children) sanitizeBlockControlChars(b.children);
+  }
+}
 function cleanPdfText(text) {
   return mergeKoreanLines(
-    text.replace(/^\d{1,4}\n/, "").replace(/^[\s]*[-–—]\s*[-–—]?\d+[-–—]?[\s]*[-–—]?[\s]*$/gm, "").replace(/^\s*\d+\s*\/\s*\d+\s*$/gm, "").replace(/\n\d{1,4}\n/g, "\n").replace(/\n\d{1,4}$/, "").replace(/^#{1,6}\s*\d{1,4}\s*$/gm, "")
+    stripControlChars(text).replace(/^\d{1,4}\n/, "").replace(/^[\s]*[-–—]\s*[-–—]?\d+[-–—]?[\s]*[-–—]?[\s]*$/gm, "").replace(/^\s*\d+\s*\/\s*\d+\s*$/gm, "").replace(/\n\d{1,4}\n/g, "\n").replace(/\n\d{1,4}$/, "").replace(/^#{1,6}\s*\d{1,4}\s*$/gm, "")
   ).replace(/^(?!\| ---).*$/gm, (line) => {
     if (/^\s*\${1,2}.+\${1,2}\s*$/.test(line)) return line;
     return collapseEvenSpacing(line);
@@ -2411,4 +2551,4 @@ function formatMb(bytes) {
 
 
 exports.cleanPdfText = cleanPdfText; exports.extractPdfMetadataOnly = extractPdfMetadataOnly; exports.parsePdfDocument = parsePdfDocument;
-//# sourceMappingURL=parser-7OFQ67QL.cjs.map
+//# sourceMappingURL=parser-EL5YETUA.cjs.map