kordoc 2.5.1 → 2.7.1

package/dist/{parser-BOIVVDYI.cjs → parser-SRI2TIZX.cjs}

@@ -6,7 +6,7 @@
 
 
 
-var _chunkTTSFPEDMcjs = require('./chunk-TTSFPEDM.cjs');
+var _chunkRFGEEHI4cjs = require('./chunk-RFGEEHI4.cjs');
 
 
 var _chunkMUOQXDZ4cjs = require('./chunk-MUOQXDZ4.cjs');
@@ -1179,7 +1179,7 @@ async function loadPdfWithTimeout(buffer) {
       new Promise((_, reject) => {
         timer = setTimeout(() => {
           loadingTask.destroy();
-          reject(new (0, _chunkTTSFPEDMcjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
+          reject(new (0, _chunkRFGEEHI4cjs.KordocError)("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
         }, PDF_LOAD_TIMEOUT_MS);
       })
     ]);
@@ -1188,10 +1188,11 @@ async function loadPdfWithTimeout(buffer) {
   }
 }
 async function parsePdfDocument(buffer, options) {
+  const formulaBuffer = _optionalChain([options, 'optionalAccess', _10 => _10.formulaOcr]) ? buffer.slice(0) : null;
   const doc = await loadPdfWithTimeout(buffer);
   try {
     const pageCount = doc.numPages;
-    if (pageCount === 0) throw new (0, _chunkTTSFPEDMcjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
+    if (pageCount === 0) throw new (0, _chunkRFGEEHI4cjs.KordocError)("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
     const metadata = { pageCount };
     await extractPdfMetadata(doc, metadata);
     const blocks = [];
@@ -1199,7 +1200,7 @@ async function parsePdfDocument(buffer, options) {
     let totalChars = 0;
     let totalTextBytes = 0;
     const effectivePageCount = Math.min(pageCount, MAX_PAGES);
-    const pageFilter = _optionalChain([options, 'optionalAccess', _10 => _10.pages]) ? _chunkMUOQXDZ4cjs.parsePageRange.call(void 0, options.pages, effectivePageCount) : null;
+    const pageFilter = _optionalChain([options, 'optionalAccess', _11 => _11.pages]) ? _chunkMUOQXDZ4cjs.parsePageRange.call(void 0, options.pages, effectivePageCount) : null;
     const totalTarget = pageFilter ? pageFilter.size : effectivePageCount;
     const fontSizeFreq = /* @__PURE__ */ new Map();
     const pageHeights = /* @__PURE__ */ new Map();
@@ -1228,17 +1229,17 @@ async function parsePdfDocument(buffer, options) {
           totalChars += t.replace(/\s/g, "").length;
           totalTextBytes += t.length * 2;
         }
-        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkTTSFPEDMcjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
+        if (totalTextBytes > MAX_TOTAL_TEXT) throw new (0, _chunkRFGEEHI4cjs.KordocError)("\uD14D\uC2A4\uD2B8 \uCD94\uCD9C \uD06C\uAE30 \uCD08\uACFC");
         parsedPages++;
-        _optionalChain([options, 'optionalAccess', _11 => _11.onProgress, 'optionalCall', _12 => _12(parsedPages, totalTarget)]);
+        _optionalChain([options, 'optionalAccess', _12 => _12.onProgress, 'optionalCall', _13 => _13(parsedPages, totalTarget)]);
       } catch (pageErr) {
-        if (pageErr instanceof _chunkTTSFPEDMcjs.KordocError) throw pageErr;
+        if (pageErr instanceof _chunkRFGEEHI4cjs.KordocError) throw pageErr;
         warnings.push({ page: i, message: `\uD398\uC774\uC9C0 ${i} \uD30C\uC2F1 \uC2E4\uD328: ${pageErr instanceof Error ? pageErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
       }
     }
     const parsedPageCount = parsedPages || (pageFilter ? pageFilter.size : effectivePageCount);
     if (totalChars / Math.max(parsedPageCount, 1) < 10) {
-      if (_optionalChain([options, 'optionalAccess', _13 => _13.ocr])) {
+      if (_optionalChain([options, 'optionalAccess', _14 => _14.ocr])) {
         try {
           const { ocrPages } = await Promise.resolve().then(() => _interopRequireWildcard(require("./provider-YN2SSK4X.cjs")));
           const ocrBlocks = await ocrPages(doc, options.ocr, pageFilter, effectivePageCount);
@@ -1249,21 +1250,31 @@ async function parsePdfDocument(buffer, options) {
         } catch (e2) {
         }
       }
-      throw Object.assign(new (0, _chunkTTSFPEDMcjs.KordocError)(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
+      throw Object.assign(new (0, _chunkRFGEEHI4cjs.KordocError)(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
     }
-    if (_optionalChain([options, 'optionalAccess', _14 => _14.removeHeaderFooter]) !== false && parsedPageCount >= 3) {
+    if (_optionalChain([options, 'optionalAccess', _15 => _15.removeHeaderFooter]) !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);
       for (let ri = removed.length - 1; ri >= 0; ri--) {
         blocks.splice(removed[ri], 1);
       }
     }
+    if (_optionalChain([options, 'optionalAccess', _16 => _16.formulaOcr]) && formulaBuffer) {
+      try {
+        await applyFormulaOcr(formulaBuffer, blocks, pageFilter, effectivePageCount, warnings, options.onProgress);
+      } catch (e) {
+        warnings.push({
+          message: `\uC218\uC2DD OCR \uC2E4\uD328: ${e instanceof Error ? e.message : String(e)}`,
+          code: "PARTIAL_PARSE"
+        });
+      }
+    }
     const medianFontSize = computeMedianFontSizeFromFreq(fontSizeFreq);
     if (medianFontSize > 0) {
       detectHeadings(blocks, medianFontSize);
     }
     detectMarkerHeadings(blocks);
     const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
-    let markdown = cleanPdfText(_chunkTTSFPEDMcjs.blocksToMarkdown.call(void 0, blocks));
+    let markdown = cleanPdfText(_chunkRFGEEHI4cjs.blocksToMarkdown.call(void 0, blocks));
     return { markdown, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0 };
   } finally {
     await doc.destroy().catch(() => {
@@ -1273,7 +1284,7 @@ async function parsePdfDocument(buffer, options) {
 async function extractPdfMetadata(doc, metadata) {
   try {
     const result = await doc.getMetadata();
-    if (!_optionalChain([result, 'optionalAccess', _15 => _15.info])) return;
+    if (!_optionalChain([result, 'optionalAccess', _17 => _17.info])) return;
     const info = result.info;
     if (typeof info.Title === "string" && info.Title.trim()) metadata.title = info.Title.trim();
     if (typeof info.Author === "string" && info.Author.trim()) metadata.author = info.Author.trim();
@@ -1336,15 +1347,15 @@ function computeMedianFontSizeFromFreq(freq) {
 }
 function detectHeadings(blocks, medianFontSize) {
   for (const block of blocks) {
-    if (block.type !== "paragraph" || !block.text || !_optionalChain([block, 'access', _16 => _16.style, 'optionalAccess', _17 => _17.fontSize])) continue;
+    if (block.type !== "paragraph" || !block.text || !_optionalChain([block, 'access', _18 => _18.style, 'optionalAccess', _19 => _19.fontSize])) continue;
     const text = block.text.trim();
     if (text.length === 0 || text.length > 200) continue;
     if (/^\d+$/.test(text)) continue;
     const ratio = block.style.fontSize / medianFontSize;
     let level = 0;
-    if (ratio >= _chunkTTSFPEDMcjs.HEADING_RATIO_H1) level = 1;
-    else if (ratio >= _chunkTTSFPEDMcjs.HEADING_RATIO_H2) level = 2;
-    else if (ratio >= _chunkTTSFPEDMcjs.HEADING_RATIO_H3) level = 3;
+    if (ratio >= _chunkRFGEEHI4cjs.HEADING_RATIO_H1) level = 1;
+    else if (ratio >= _chunkRFGEEHI4cjs.HEADING_RATIO_H2) level = 2;
+    else if (ratio >= _chunkRFGEEHI4cjs.HEADING_RATIO_H3) level = 3;
     if (level > 0) {
       block.type = "heading";
       block.level = level;
@@ -1404,7 +1415,7 @@ function detectMarkerHeadings(blocks) {
       block.level = 4;
       continue;
     }
-    if (/^[가-힣]{2,6}$/.test(text) && _optionalChain([block, 'access', _18 => _18.style, 'optionalAccess', _19 => _19.fontSize])) {
+    if (/^[가-힣]{2,6}$/.test(text) && _optionalChain([block, 'access', _20 => _20.style, 'optionalAccess', _21 => _21.fontSize])) {
       const prev = blocks[i - 1];
       const next = blocks[i + 1];
       const prevIsStructural = !prev || prev.type === "table" || prev.type === "heading" || prev.type === "separator";
@@ -1592,7 +1603,7 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
     }
     if (remaining.length > 0) {
       const allY = remaining.map((i) => i.y);
-      const pageH = _chunkTTSFPEDMcjs.safeMax.call(void 0, allY) - _chunkTTSFPEDMcjs.safeMin.call(void 0, allY);
+      const pageH = _chunkRFGEEHI4cjs.safeMax.call(void 0, allY) - _chunkRFGEEHI4cjs.safeMin.call(void 0, allY);
       const groups = xyCutOrder(remaining, Math.max(15, pageH * 0.03));
       const textBlocks = [];
       for (const group of groups) {
@@ -1680,7 +1691,7 @@ 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 = _chunkTTSFPEDMcjs.safeMax.call(void 0, allY) - _chunkTTSFPEDMcjs.safeMin.call(void 0, allY);
+      const pageHeight = _chunkRFGEEHI4cjs.safeMax.call(void 0, allY) - _chunkRFGEEHI4cjs.safeMin.call(void 0, allY);
       const gapThreshold = Math.max(15, pageHeight * 0.03);
       const orderedGroups = xyCutOrder(items, gapThreshold);
       for (const group of orderedGroups) {
@@ -1729,7 +1740,7 @@ function dominantStyle(items) {
     }
   }
   if (dominantSize === 0) return void 0;
-  const fontName = _optionalChain([items, 'access', _20 => _20.find, 'call', _21 => _21((i) => i.fontSize === dominantSize), 'optionalAccess', _22 => _22.fontName]) || void 0;
+  const fontName = _optionalChain([items, 'access', _22 => _22.find, 'call', _23 => _23((i) => i.fontSize === dominantSize), 'optionalAccess', _24 => _24.fontName]) || void 0;
   return { fontSize: dominantSize, fontName };
 }
 function normalizeItems(rawItems) {
@@ -1827,14 +1838,14 @@ 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 = _chunkTTSFPEDMcjs.safeMax.call(void 0, gaps);
+  const maxGap = _chunkRFGEEHI4cjs.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 = _chunkTTSFPEDMcjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkTTSFPEDMcjs.safeMin.call(void 0, allItems.map((i) => i.x));
+  const pageWidth = _chunkRFGEEHI4cjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkRFGEEHI4cjs.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++) {
@@ -2034,7 +2045,7 @@ function buildGridTable(lines, columns) {
   return md.join("\n");
 }
 function mergeLineSimple(items) {
-  if (items.length <= 1) return _optionalChain([items, 'access', _23 => _23[0], 'optionalAccess', _24 => _24.text]) || "";
+  if (items.length <= 1) return _optionalChain([items, 'access', _25 => _25[0], 'optionalAccess', _26 => _26.text]) || "";
   const sorted = [...items].sort((a, b) => a.x - b.x);
   const isEvenSpaced = detectEvenSpacedItems(sorted);
   let result = sorted[0].text;
@@ -2071,7 +2082,10 @@ function mergeLineSimple(items) {
 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, "")
-  ).replace(/^(?!\| ---).*$/gm, (line) => collapseEvenSpacing(line)).replace(/([□■◆○●▶ㅇ])\s+([가-힣])\s+([가-힣])/g, "$1 $2$3").replace(/\n{3,}/g, "\n\n").trim();
+  ).replace(/^(?!\| ---).*$/gm, (line) => {
+    if (/^\s*\${1,2}.+\${1,2}\s*$/.test(line)) return line;
+    return collapseEvenSpacing(line);
+  }).replace(/([□■◆○●▶ㅇ])\s+([가-힣])\s+([가-힣])/g, "$1 $2$3").replace(/\n{3,}/g, "\n\n").trim();
 }
 function startsWithMarker(line) {
   const t = line.trimStart();
@@ -2194,7 +2208,7 @@ 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', _25 => _25.text, 'optionalAccess', _26 => _26.trim, 'call', _27 => _27()])) continue;
+    if (!b.bbox || !b.pageNumber || !_optionalChain([b, 'access', _27 => _27.text, 'optionalAccess', _28 => _28.trim, 'call', _29 => _29()])) continue;
     const ph = pageHeights.get(b.bbox.page) || pageHeights.get(b.pageNumber);
     if (!ph) continue;
     const blockTop = ph - (b.bbox.y + b.bbox.height);
@@ -2217,7 +2231,7 @@ function removeHeaderFooterBlocks(blocks, pageHeights, warnings) {
     }
     const repeatedPatterns = /* @__PURE__ */ new Set();
     for (const [p, count] of patternCount) {
-      if (count >= MIN_REPEAT && (_nullishCoalesce(_optionalChain([patternPages, 'access', _28 => _28.get, 'call', _29 => _29(p), 'optionalAccess', _30 => _30.size]), () => ( 0))) >= MIN_REPEAT) {
+      if (count >= MIN_REPEAT && (_nullishCoalesce(_optionalChain([patternPages, 'access', _30 => _30.get, 'call', _31 => _31(p), 'optionalAccess', _32 => _32.size]), () => ( 0))) >= MIN_REPEAT) {
         repeatedPatterns.add(p);
       }
     }
@@ -2274,9 +2288,127 @@ function mergeKoreanLines(text) {
   }
   return result.join("\n");
 }
+async function applyFormulaOcr(buffer, blocks, pageFilter, effectivePageCount, warnings, _onProgress) {
+  const formulaMod = await Promise.resolve().then(() => _interopRequireWildcard(require("./formula-XGG6ZP42.cjs")));
+  const { FormulaPipeline, ensureFormulaModels } = formulaMod;
+  await ensureFormulaModels((p) => {
+    if (p.phase === "download" && p.total) {
+      const pct = Math.floor(p.downloaded / p.total * 100);
+      process.stderr.write(`\r[kordoc-formula] ${p.spec.name} ${pct}% (${formatMb(p.downloaded)}/${formatMb(p.total)})`);
+      if (p.downloaded >= p.total) process.stderr.write("\n");
+    } else if (p.phase === "verify") {
+      process.stderr.write(`[kordoc-formula] ${p.spec.name} SHA-256 \uAC80\uC99D \uC911...
+`);
+    } else if (p.phase === "done") {
+      process.stderr.write(`[kordoc-formula] ${p.spec.name} \uC900\uBE44 \uC644\uB8CC
+`);
+    } else if (p.phase === "skip") {
+    }
+  });
+  const pipeline = await FormulaPipeline.create();
+  try {
+    const pagesResult = await pipeline.runOnBuffer(buffer, pageFilter);
+    if (pagesResult.length === 0) return;
+    let insertedCount = 0;
+    let removedDupCount = 0;
+    for (const page of pagesResult) {
+      const pageNumber = page.pageNumber;
+      const pdfHeight = page.pdfHeight;
+      const scaleX = page.renderedWidth > 0 ? page.pdfWidth / page.renderedWidth : 0.5;
+      const scaleY = page.renderedHeight > 0 ? page.pdfHeight / page.renderedHeight : 0.5;
+      const candidates = [];
+      for (const r of page.regions) {
+        if (!r.latex || !r.latex.trim()) continue;
+        const wrapped = r.kind === "display" ? `$$${r.latex}$$` : `$${r.latex}$`;
+        const x1 = r.bbox.x1 * scaleX;
+        const x2 = r.bbox.x2 * scaleX;
+        const yTop = pdfHeight - r.bbox.y1 * scaleY;
+        const yBottom = pdfHeight - r.bbox.y2 * scaleY;
+        const centerY = (yTop + yBottom) / 2;
+        const width = x2 - x1;
+        const height = yTop - yBottom;
+        candidates.push({
+          block: {
+            type: "paragraph",
+            text: wrapped,
+            pageNumber,
+            bbox: { page: pageNumber, x: x1, y: yBottom, width, height }
+          },
+          pdfBbox: { x1, x2, yTop, yBottom },
+          centerY
+        });
+      }
+      if (candidates.length === 0) continue;
+      const OVERLAP_THRESHOLD = 0.6;
+      const indicesToRemove = /* @__PURE__ */ new Set();
+      for (let i = 0; i < blocks.length; i++) {
+        const b = blocks[i];
+        if (b.pageNumber !== pageNumber) continue;
+        if (b.type === "table") continue;
+        if (!b.bbox || b.bbox.width <= 0 || b.bbox.height <= 0) continue;
+        const blockArea = b.bbox.width * b.bbox.height;
+        if (blockArea <= 0) continue;
+        for (const c of candidates) {
+          const ox1 = Math.max(b.bbox.x, c.pdfBbox.x1);
+          const ox2 = Math.min(b.bbox.x + b.bbox.width, c.pdfBbox.x2);
+          const oy1 = Math.max(b.bbox.y, c.pdfBbox.yBottom);
+          const oy2 = Math.min(b.bbox.y + b.bbox.height, c.pdfBbox.yTop);
+          const interArea = Math.max(0, ox2 - ox1) * Math.max(0, oy2 - oy1);
+          if (interArea / blockArea >= OVERLAP_THRESHOLD) {
+            indicesToRemove.add(i);
+            break;
+          }
+        }
+      }
+      if (indicesToRemove.size > 0) {
+        const sorted = [...indicesToRemove].sort((a, b) => b - a);
+        for (const idx of sorted) blocks.splice(idx, 1);
+        removedDupCount += indicesToRemove.size;
+      }
+      candidates.sort((a, b) => b.centerY - a.centerY);
+      for (const c of candidates) {
+        let insertIdx = -1;
+        let pageFirstIdx = -1;
+        let pageLastIdx = -1;
+        for (let i = 0; i < blocks.length; i++) {
+          const b = blocks[i];
+          if (b.pageNumber !== pageNumber) continue;
+          if (pageFirstIdx === -1) pageFirstIdx = i;
+          pageLastIdx = i;
+          if (!b.bbox) continue;
+          const blockCenter = b.bbox.y + b.bbox.height / 2;
+          if (blockCenter < c.centerY) {
+            insertIdx = i;
+            break;
+          }
+        }
+        if (insertIdx !== -1) {
+          blocks.splice(insertIdx, 0, c.block);
+        } else if (pageLastIdx !== -1) {
+          blocks.splice(pageLastIdx + 1, 0, c.block);
+        } else {
+          blocks.push(c.block);
+        }
+        insertedCount++;
+      }
+    }
+    if (insertedCount > 0 || removedDupCount > 0) {
+      process.stderr.write(
+        `[kordoc-formula] ${insertedCount}\uAC1C \uC218\uC2DD \uC0BD\uC785, ${removedDupCount}\uAC1C \uC911\uBCF5 block \uC81C\uAC70 (${pagesResult.length}\uAC1C \uD398\uC774\uC9C0)
+`
+      );
+    }
+  } finally {
+    await pipeline.destroy().catch(() => {
+    });
+  }
+}
+function formatMb(bytes) {
+  return `${(bytes / 1024 / 1024).toFixed(1)}MB`;
+}
 
 
 
 
 exports.cleanPdfText = cleanPdfText; exports.extractPdfMetadataOnly = extractPdfMetadataOnly; exports.parsePdfDocument = parsePdfDocument;
-//# sourceMappingURL=parser-BOIVVDYI.cjs.map
+//# sourceMappingURL=parser-SRI2TIZX.cjs.map