kordoc 3.8.2 → 3.8.4

package/dist/{parser-NS4ZPD7B.cjs → parser-ZVKCFXR3.cjs}

@@ -6,7 +6,7 @@
 
 
 
-var _chunkYBPNKFJWcjs = require('./chunk-YBPNKFJW.cjs');
+var _chunkZ3O36LBJcjs = require('./chunk-Z3O36LBJ.cjs');
 
 
 var _chunkDCZVOIEOcjs = require('./chunk-DCZVOIEO.cjs');
@@ -1265,7 +1265,106 @@ function detectClusterTables(items, pageNum) {
       }
     }
   }
-  return results;
+  return results.filter((r) => !isTwoColumnProse(r));
+}
+function isTwoColumnProse(r) {
+  const t = r.table;
+  if (t.rows < 8) return false;
+  const dense = [];
+  for (let c = 0; c < t.cols; c++) {
+    const filled = t.cells.filter((row) => _optionalChain([row, 'access', _9 => _9[c], 'optionalAccess', _10 => _10.text, 'access', _11 => _11.trim, 'call', _12 => _12()])).length;
+    if (filled / t.rows >= 0.3) dense.push(c);
+  }
+  if (dense.length !== 2) return false;
+  for (const c of dense) {
+    const lens = t.cells.map((row) => _nullishCoalesce(_optionalChain([row, 'access', _13 => _13[c], 'optionalAccess', _14 => _14.text, 'access', _15 => _15.replace, 'call', _16 => _16(/\s+/g, ""), 'access', _17 => _17.length]), () => ( 0))).filter((n) => n > 0);
+    const avg = lens.reduce((s, v) => s + v, 0) / (lens.length || 1);
+    if (avg < 12) return false;
+  }
+  return findTwoColumnProseCutX([...r.usedItems]) !== null;
+}
+function findTwoColumnProseCutX(items) {
+  const lines = groupByBaseline(items);
+  if (lines.length < 8) return null;
+  let minX = Infinity;
+  let maxX = -Infinity;
+  for (const i of items) {
+    if (i.x < minX) minX = i.x;
+    if (i.x + i.w > maxX) maxX = i.x + i.w;
+  }
+  if (!Number.isFinite(maxX - minX)) return null;
+  if (maxX - minX < 100) return null;
+  const lo = minX + (maxX - minX) * 0.3;
+  const hi = minX + (maxX - minX) * 0.7;
+  const step = Math.max(2, (hi - lo) / 400);
+  let cutX = 0;
+  let bestCover = Infinity;
+  for (let x = lo; x <= hi; x += step) {
+    let cover = 0;
+    for (const line of lines) {
+      if (line.items.some((i) => i.x < x && i.x + i.w > x)) cover++;
+    }
+    if (cover < bestCover) {
+      bestCover = cover;
+      cutX = x;
+    }
+  }
+  if (bestCover / lines.length > 0.15) return null;
+  const left = [];
+  const right = [];
+  let twoSide = 0;
+  for (const line of lines) {
+    let hasL = false;
+    let hasR = false;
+    for (const i of line.items) {
+      if (i.x + i.w <= cutX) {
+        left.push(i);
+        hasL = true;
+      } else if (i.x >= cutX) {
+        right.push(i);
+        hasR = true;
+      }
+    }
+    if (hasL && hasR) twoSide++;
+  }
+  if (twoSide / lines.length < 0.55) return null;
+  if (left.length < 5 || right.length < 5) return null;
+  const allText = items.map((i) => i.text).join("").replace(/\s+/g, "");
+  const digits = (allText.match(/[\d,.%△—-]/g) || []).length;
+  if (allText.length === 0 || digits / allText.length > 0.15) return null;
+  const stats = [left, right].map((side) => {
+    let sMinX = Infinity;
+    let sMaxR = -Infinity;
+    let fsSum = 0;
+    const rowEnds = /* @__PURE__ */ new Map();
+    const rowLens = /* @__PURE__ */ new Map();
+    const rowFirst = /* @__PURE__ */ new Map();
+    for (const i of side) {
+      if (i.x < sMinX) sMinX = i.x;
+      const rgt = i.x + i.w;
+      if (rgt > sMaxR) sMaxR = rgt;
+      fsSum += i.fontSize;
+      const key = Math.round(i.y / 4);
+      rowEnds.set(key, Math.max(_nullishCoalesce(rowEnds.get(key), () => ( -Infinity)), rgt));
+      rowLens.set(key, (_nullishCoalesce(rowLens.get(key), () => ( 0))) + i.text.replace(/\s+/g, "").length);
+      const f = rowFirst.get(key);
+      if (!f || i.x < f.x) rowFirst.set(key, i);
+    }
+    const ends = [...rowEnds.values()].sort((a, b) => a - b);
+    const p85 = _nullishCoalesce(ends[Math.floor(ends.length * 0.85)], () => ( sMaxR));
+    const fs = fsSum / side.length;
+    const justified = ends.filter((e) => Math.abs(e - p85) <= fs).length / ends.length;
+    const lens = [...rowLens.values()];
+    const avgLen = lens.reduce((s, v) => s + v, 0) / (lens.length || 1);
+    const markers = [...rowFirst.values()].filter((i) => /^[•▪◦‣∙·\-–—□■◇◆▶※*]/.test(i.text)).length;
+    return { width: sMaxR - sMinX, justified, avgLen, markerRatio: markers / (rowFirst.size || 1) };
+  });
+  if (stats.some((s) => s.avgLen < 12)) return null;
+  if (stats.some((s) => s.markerRatio > 0.1)) return null;
+  const widthSym = Math.min(stats[0].width, stats[1].width) / Math.max(stats[0].width, stats[1].width);
+  if (widthSym < 0.6) return null;
+  if (Math.min(stats[0].justified, stats[1].justified) < 0.55) return null;
+  return cutX;
 }
 function mergeEvenSpacedClusters(items) {
   const originMap = /* @__PURE__ */ new Map();
@@ -1655,7 +1754,7 @@ function buildClusterTable(rows, columns, pageNum) {
     const nonEmptyCols = cells[r].filter((c) => c.text.trim()).length;
     if (nonEmptyCols !== 1) continue;
     if (cells[r][0].text.trim() !== "") continue;
-    const contentText = _optionalChain([cells, 'access', _9 => _9[r], 'access', _10 => _10.find, 'call', _11 => _11((c) => c.text.trim()), 'optionalAccess', _12 => _12.text, 'access', _13 => _13.trim, 'call', _14 => _14()]) || "";
+    const contentText = _optionalChain([cells, 'access', _18 => _18[r], 'access', _19 => _19.find, 'call', _20 => _20((c) => c.text.trim()), 'optionalAccess', _21 => _21.text, 'access', _22 => _22.trim, 'call', _23 => _23()]) || "";
     if (/^[○●▶\-·]/.test(contentText)) continue;
     for (let pr = r - 1; pr >= 0; pr--) {
       if (cells[pr].some((c) => c.text.trim())) {
@@ -1877,14 +1976,14 @@ 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 = _chunkYBPNKFJWcjs.safeMax.call(void 0, gaps);
+  const maxGap = _chunkZ3O36LBJcjs.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 = _chunkYBPNKFJWcjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkYBPNKFJWcjs.safeMin.call(void 0, allItems.map((i) => i.x));
+  const pageWidth = _chunkZ3O36LBJcjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkZ3O36LBJcjs.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++) {
@@ -2100,15 +2199,15 @@ function computeMedianFontSizeFromFreq(freq) {
 }
 function detectHeadings(blocks, medianFontSize) {
   for (const block of blocks) {
-    if (block.type !== "paragraph" || !block.text || !_optionalChain([block, 'access', _15 => _15.style, 'optionalAccess', _16 => _16.fontSize])) continue;
+    if (block.type !== "paragraph" || !block.text || !_optionalChain([block, 'access', _24 => _24.style, 'optionalAccess', _25 => _25.fontSize])) continue;
     const text = block.text.trim();
     if (text.length === 0 || text.length > 200) continue;
     if (/^\d+$/.test(text)) continue;
     const ratio = block.style.fontSize / medianFontSize;
     let level = 0;
-    if (ratio >= _chunkYBPNKFJWcjs.HEADING_RATIO_H1) level = 1;
-    else if (ratio >= _chunkYBPNKFJWcjs.HEADING_RATIO_H2) level = 2;
-    else if (ratio >= _chunkYBPNKFJWcjs.HEADING_RATIO_H3) level = 3;
+    if (ratio >= _chunkZ3O36LBJcjs.HEADING_RATIO_H1) level = 1;
+    else if (ratio >= _chunkZ3O36LBJcjs.HEADING_RATIO_H2) level = 2;
+    else if (ratio >= _chunkZ3O36LBJcjs.HEADING_RATIO_H3) level = 3;
     if (level > 0) {
       block.type = "heading";
       block.level = level;
@@ -2157,7 +2256,7 @@ function detectMarkerHeadings(blocks) {
       block.level = 4;
       continue;
     }
-    if (/^[가-힣]{2,6}$/.test(text) && _optionalChain([block, 'access', _17 => _17.style, 'optionalAccess', _18 => _18.fontSize])) {
+    if (/^[가-힣]{2,6}$/.test(text) && _optionalChain([block, 'access', _26 => _26.style, 'optionalAccess', _27 => _27.fontSize])) {
       const prev = blocks[i - 1];
       const next = blocks[i + 1];
       const prevIsStructural = !prev || prev.type === "table" || prev.type === "heading" || prev.type === "separator";
@@ -2410,7 +2509,7 @@ function removeHeaderFooterBlocks(blocks, pageHeights, warnings) {
   const bottomEntries = [];
   for (let bi = 0; bi < blocks.length; bi++) {
     const b = blocks[bi];
-    if (!b.bbox || !b.pageNumber || !_optionalChain([b, 'access', _19 => _19.text, 'optionalAccess', _20 => _20.trim, 'call', _21 => _21()])) continue;
+    if (!b.bbox || !b.pageNumber || !_optionalChain([b, 'access', _28 => _28.text, 'optionalAccess', _29 => _29.trim, 'call', _30 => _30()])) continue;
     const ph = pageHeights.get(b.bbox.page) || pageHeights.get(b.pageNumber);
     if (!ph) continue;
     const blockTop = ph - (b.bbox.y + b.bbox.height);
@@ -2433,7 +2532,7 @@ function removeHeaderFooterBlocks(blocks, pageHeights, warnings) {
     }
     const repeatedPatterns = /* @__PURE__ */ new Set();
     for (const [p, count] of patternCount) {
-      if (count >= MIN_REPEAT && (_nullishCoalesce(_optionalChain([patternPages, 'access', _22 => _22.get, 'call', _23 => _23(p), 'optionalAccess', _24 => _24.size]), () => ( 0))) >= MIN_REPEAT) {
+      if (count >= MIN_REPEAT && (_nullishCoalesce(_optionalChain([patternPages, 'access', _31 => _31.get, 'call', _32 => _32(p), 'optionalAccess', _33 => _33.size]), () => ( 0))) >= MIN_REPEAT) {
         repeatedPatterns.add(p);
       }
     }
@@ -2462,7 +2561,7 @@ function extractPageBlocksWithLines(items, pageNum, opList, pageWidth, pageHeigh
   if (grids.length > 0) {
     return extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals);
   }
-  return extractPageBlocksFallback(items, pageNum);
+  return extractPageBlocksFallback(items, pageNum, true);
 }
 var STRIKE_MAX_THICKNESS = 2;
 var STRIKE_MAX_THICKNESS_RATIO = 0.25;
@@ -2627,7 +2726,7 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
     }
     if (remaining.length > 0) {
       const allY = remaining.map((i) => i.y);
-      const pageH = _chunkYBPNKFJWcjs.safeMax.call(void 0, allY) - _chunkYBPNKFJWcjs.safeMin.call(void 0, allY);
+      const pageH = _chunkZ3O36LBJcjs.safeMax.call(void 0, allY) - _chunkZ3O36LBJcjs.safeMin.call(void 0, allY);
       const groups = xyCutOrder(remaining, Math.max(15, pageH * 0.03));
       const textBlocks = [];
       for (const group of groups) {
@@ -2707,7 +2806,53 @@ function mergeAdjacentTableBlocks(blocks) {
   }
   return result;
 }
-function extractPageBlocksFallback(items, pageNum) {
+function splitTwoColumnProse(items, cutX) {
+  const left = [];
+  const right = [];
+  const cross = [];
+  for (const i of items) {
+    if (i.x + i.w <= cutX) left.push(i);
+    else if (i.x >= cutX) right.push(i);
+    else cross.push(i);
+  }
+  if (cross.length === 0) {
+    return [left, right].filter((g2) => g2.length > 0);
+  }
+  cross.sort((a, b) => b.y - a.y);
+  const crossLines = [];
+  for (const c of cross) {
+    const last = crossLines[crossLines.length - 1];
+    if (last && Math.abs(last[0].y - c.y) <= 3) last.push(c);
+    else crossLines.push([c]);
+  }
+  const bandItem = (arr) => arr.filter((i) => {
+    for (const cl of crossLines) {
+      if (Math.abs(cl[0].y - i.y) <= 3) {
+        cl.push(i);
+        return false;
+      }
+    }
+    return true;
+  });
+  const leftRest = bandItem(left);
+  const rightRest = bandItem(right);
+  const boundYs = crossLines.map((cl) => cl[0].y);
+  const bandOf = (y) => {
+    let k = 0;
+    while (k < boundYs.length && y < boundYs[k]) k++;
+    return k;
+  };
+  const groups = [];
+  for (let k = 0; k <= crossLines.length; k++) {
+    const L = leftRest.filter((i) => bandOf(i.y) === k);
+    const R = rightRest.filter((i) => bandOf(i.y) === k);
+    if (L.length > 0) groups.push(L);
+    if (R.length > 0) groups.push(R);
+    if (k < crossLines.length) groups.push(crossLines[k]);
+  }
+  return groups;
+}
+function extractPageBlocksFallback(items, pageNum, fullPage = false) {
   if (items.length === 0) return [];
   const blocks = [];
   const clusterItems = items.map((i) => ({
@@ -2748,17 +2893,18 @@ function extractPageBlocksFallback(items, pageNum) {
       return by - ay;
     });
   } else {
+    const proseCutX = fullPage ? findTwoColumnProseCutX(items) : null;
     const allYLines = mergeSuperscriptLines(groupByY(items));
-    const columns = detectColumns(allYLines);
+    const columns = proseCutX !== null ? null : detectColumns(allYLines);
     if (columns && columns.length >= 3) {
       const tableText = extractWithColumns(allYLines, columns);
       const bbox = computeBBox(items, pageNum);
       blocks.push({ type: "paragraph", text: tableText, pageNumber: pageNum, bbox, style: dominantStyle(items) });
     } else {
       const allY = items.map((i) => i.y);
-      const pageHeight = _chunkYBPNKFJWcjs.safeMax.call(void 0, allY) - _chunkYBPNKFJWcjs.safeMin.call(void 0, allY);
+      const pageHeight = _chunkZ3O36LBJcjs.safeMax.call(void 0, allY) - _chunkZ3O36LBJcjs.safeMin.call(void 0, allY);
       const gapThreshold = Math.max(15, pageHeight * 0.03);
-      const orderedGroups = xyCutOrder(items, gapThreshold);
+      const orderedGroups = proseCutX !== null ? splitTwoColumnProse(items, proseCutX) : xyCutOrder(items, gapThreshold);
       for (const group of orderedGroups) {
         if (group.length === 0) continue;
         const yLines = mergeSuperscriptLines(groupByY(group));
@@ -3009,7 +3155,7 @@ async function loadPdfWithTimeout(buffer) {
       new Promise((_, reject) => {
         timer = setTimeout(() => {
           loadingTask.destroy();
-          reject(new (0, _chunkYBPNKFJWcjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
+          reject(new (0, _chunkZ3O36LBJcjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
         }, PDF_LOAD_TIMEOUT_MS);
       })
     ]);
@@ -3018,11 +3164,11 @@ async function loadPdfWithTimeout(buffer) {
   }
 }
 async function parsePdfDocument(buffer, options) {
-  const formulaBuffer = _optionalChain([options, 'optionalAccess', _25 => _25.formulaOcr]) ? buffer.slice(0) : null;
+  const formulaBuffer = _optionalChain([options, 'optionalAccess', _34 => _34.formulaOcr]) ? buffer.slice(0) : null;
   const doc = await loadPdfWithTimeout(buffer);
   try {
     const pageCount = doc.numPages;
-    if (pageCount === 0) throw new (0, _chunkYBPNKFJWcjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
+    if (pageCount === 0) throw new (0, _chunkZ3O36LBJcjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
     const metadata = { pageCount };
     await extractPdfMetadata(doc, metadata);
     const blocks = [];
@@ -3031,7 +3177,7 @@ async function parsePdfDocument(buffer, options) {
     let totalChars = 0;
     let totalTextBytes = 0;
     const effectivePageCount = Math.min(pageCount, MAX_PAGES);
-    const pageFilter = _optionalChain([options, 'optionalAccess', _26 => _26.pages]) ? _chunkDCZVOIEOcjs.parsePageRange.call(void 0, options.pages, effectivePageCount) : null;
+    const pageFilter = _optionalChain([options, 'optionalAccess', _35 => _35.pages]) ? _chunkDCZVOIEOcjs.parsePageRange.call(void 0, options.pages, effectivePageCount) : null;
     const totalTarget = pageFilter ? pageFilter.size : effectivePageCount;
     const fontSizeFreq = /* @__PURE__ */ new Map();
     const pageHeights = /* @__PURE__ */ new Map();
@@ -3082,18 +3228,18 @@ async function parsePdfDocument(buffer, options) {
           pageText += pageText ? "\n" + t : t;
         }
         pageQuality.push(computePageQuality(i, pageText));
-        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkYBPNKFJWcjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
+        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkZ3O36LBJcjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
         parsedPages++;
-        _optionalChain([options, 'optionalAccess', _27 => _27.onProgress, 'optionalCall', _28 => _28(parsedPages, totalTarget)]);
+        _optionalChain([options, 'optionalAccess', _36 => _36.onProgress, 'optionalCall', _37 => _37(parsedPages, totalTarget)]);
       } catch (pageErr) {
-        if (pageErr instanceof _chunkYBPNKFJWcjs.KordocError) throw pageErr;
+        if (pageErr instanceof _chunkZ3O36LBJcjs.KordocError) throw pageErr;
         warnings.push({ page: i, message: `\uD398\uC774\uC9C0 ${i} \uD30C\uC2F1 \uC2E4\uD328: ${pageErr instanceof Error ? pageErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
       }
     }
     const parsedPageCount = parsedPages || (pageFilter ? pageFilter.size : effectivePageCount);
     let isImageBased = false;
     if (totalChars / Math.max(parsedPageCount, 1) < 10) {
-      if (_optionalChain([options, 'optionalAccess', _29 => _29.ocr])) {
+      if (_optionalChain([options, 'optionalAccess', _38 => _38.ocr])) {
         try {
           const { ocrPages } = await Promise.resolve().then(() => _interopRequireWildcard(require("./provider-G4C2V2PD.cjs")));
           const ocrBlocks = await ocrPages(doc, options.ocr, pageFilter, effectivePageCount);
@@ -3128,14 +3274,14 @@ async function parsePdfDocument(buffer, options) {
         warnings.push({ page, message: `${count}\uAC1C \uC774\uBBF8\uC9C0 \uC601\uC5ED\uC5D0 \uCD94\uCD9C \uAC00\uB2A5\uD55C \uD14D\uC2A4\uD2B8 \uC5C6\uC74C (\uADF8\uB9BC/\uCC28\uD2B8/\uB3C4\uC7A5 \uB0B4\uC6A9 \uB204\uB77D \uAC00\uB2A5)`, code: "SKIPPED_IMAGE" });
       }
     }
-    if (_optionalChain([options, 'optionalAccess', _30 => _30.removeHeaderFooter]) !== false && parsedPageCount >= 3) {
+    if (_optionalChain([options, 'optionalAccess', _39 => _39.removeHeaderFooter]) !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);
       for (let ri = removed.length - 1; ri >= 0; ri--) {
         blocks.splice(removed[ri], 1);
       }
     }
     mergeCrossPageTables(blocks);
-    if (_optionalChain([options, 'optionalAccess', _31 => _31.formulaOcr]) && formulaBuffer) {
+    if (_optionalChain([options, 'optionalAccess', _40 => _40.formulaOcr]) && formulaBuffer) {
       try {
         await applyFormulaOcr(formulaBuffer, blocks, pageFilter, effectivePageCount, warnings, options.onProgress);
       } catch (e) {
@@ -3154,7 +3300,7 @@ async function parsePdfDocument(buffer, options) {
     detectKoreanListBlocks(blocks);
     const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
     sanitizeBlockControlChars(blocks);
-    let markdown = cleanPdfText(_chunkYBPNKFJWcjs.blocksToMarkdown.call(void 0, blocks));
+    let markdown = cleanPdfText(_chunkZ3O36LBJcjs.blocksToMarkdown.call(void 0, blocks));
     return {
       markdown,
       blocks,
@@ -3173,7 +3319,7 @@ async function parsePdfDocument(buffer, options) {
 async function extractPdfMetadata(doc, metadata) {
   try {
     const result = await doc.getMetadata();
-    if (!_optionalChain([result, 'optionalAccess', _32 => _32.info])) return;
+    if (!_optionalChain([result, 'optionalAccess', _41 => _41.info])) return;
     const info = result.info;
     if (typeof info.Title === "string" && info.Title.trim()) metadata.title = info.Title.trim();
     if (typeof info.Author === "string" && info.Author.trim()) metadata.author = info.Author.trim();
@@ -3213,4 +3359,4 @@ async function extractPdfMetadataOnly(buffer) {
 
 
 exports.cleanPdfText = cleanPdfText; exports.detectKoreanListBlocks = detectKoreanListBlocks; exports.detectTableCaptions = detectTableCaptions; exports.extractPdfMetadataOnly = extractPdfMetadataOnly; exports.mergeCrossPageTables = mergeCrossPageTables; exports.parsePdfDocument = parsePdfDocument; exports.removeHeaderFooterBlocks = removeHeaderFooterBlocks;
-//# sourceMappingURL=parser-NS4ZPD7B.cjs.map
+//# sourceMappingURL=parser-ZVKCFXR3.cjs.map