kordoc 2.5.1 → 2.7.1

package/dist/{parser-NZFDRZLS.js → parser-6L6DZCOB.js}

@@ -6,7 +6,7 @@ import {
   blocksToMarkdown,
   safeMax,
   safeMin
-} from "./chunk-OCVWJSG7.js";
+} from "./chunk-GNN6MHH4.js";
 import {
   parsePageRange
 } from "./chunk-SBVRCJFH.js";
@@ -1188,6 +1188,7 @@ async function loadPdfWithTimeout(buffer) {
   }
 }
 async function parsePdfDocument(buffer, options) {
+  const formulaBuffer = options?.formulaOcr ? buffer.slice(0) : null;
   const doc = await loadPdfWithTimeout(buffer);
   try {
     const pageCount = doc.numPages;
@@ -1257,6 +1258,16 @@ async function parsePdfDocument(buffer, options) {
         blocks.splice(removed[ri], 1);
       }
     }
+    if (options?.formulaOcr && formulaBuffer) {
+      try {
+        await applyFormulaOcr(formulaBuffer, blocks, pageFilter, effectivePageCount, warnings, options.onProgress);
+      } catch (e) {
+        warnings.push({
+          message: `\uC218\uC2DD OCR \uC2E4\uD328: ${e instanceof Error ? e.message : String(e)}`,
+          code: "PARTIAL_PARSE"
+        });
+      }
+    }
     const medianFontSize = computeMedianFontSizeFromFreq(fontSizeFreq);
     if (medianFontSize > 0) {
       detectHeadings(blocks, medianFontSize);
@@ -2071,7 +2082,10 @@ function mergeLineSimple(items) {
 function cleanPdfText(text) {
   return mergeKoreanLines(
     text.replace(/^\d{1,4}\n/, "").replace(/^[\s]*[-–—]\s*[-–—]?\d+[-–—]?[\s]*[-–—]?[\s]*$/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();
@@ -2274,9 +2288,127 @@ function mergeKoreanLines(text) {
   }
   return result.join("\n");
 }
+async function applyFormulaOcr(buffer, blocks, pageFilter, effectivePageCount, warnings, _onProgress) {
+  const formulaMod = await import("./formula-3AQUUIRF.js");
+  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`;
+}
 export {
   cleanPdfText,
   extractPdfMetadataOnly,
   parsePdfDocument
 };
-//# sourceMappingURL=parser-NZFDRZLS.js.map
+//# sourceMappingURL=parser-6L6DZCOB.js.map