kordoc 4.16.1 → 4.16.2

package/dist/{chunk-3WIQSM3Z.js → chunk-UTP2WCHS.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   extractPageBlocksWithLines
-} from "./chunk-7FKPPIDZ.js";
+} from "./chunk-SFEQYKYJ.js";
 import {
   OCR_DET_MODEL,
   OCR_REC_DICT,
@@ -9,10 +9,10 @@ import {
   ensureOcrModels,
   getOcrModelsDir,
   parseCharacterDict
-} from "./chunk-BIHS4VWR.js";
+} from "./chunk-DES2LGFZ.js";
 import {
   OPTIONAL_DEP_INSTALL_HINT
-} from "./chunk-3VLVNKJV.js";
+} from "./chunk-P6ZDXTV6.js";
 
 // src/ocr/ruling-lines.ts
 var INK_LUMA_MAX = 205;
@@ -2061,4 +2061,4 @@ export {
   runPdfOcr,
   ocrItemsToBlocks
 };
-//# sourceMappingURL=chunk-3WIQSM3Z.js.map
+//# sourceMappingURL=chunk-UTP2WCHS.js.map

package/dist/{chunk-7IRHSRPU.cjs → chunk-WDQTFAXY.cjs}

@@ -2,15 +2,16 @@
 
 
 
-var _chunkDO74GYSMcjs = require('./chunk-DO74GYSM.cjs');
 
+var _chunk62W7QCEXcjs = require('./chunk-62W7QCEX.cjs');
 
 
-var _chunk2BZFKIRBcjs = require('./chunk-2BZFKIRB.cjs');
 
+var _chunkCEIATRDWcjs = require('./chunk-CEIATRDW.cjs');
 
 
-var _chunkYRRBA4HWcjs = require('./chunk-YRRBA4HW.cjs');
+
+var _chunkSTM3ORY7cjs = require('./chunk-STM3ORY7.cjs');
 
 // src/ocr/models.ts
 var _path = require('path');
@@ -38,10 +39,10 @@ var OCR_REC_DICT = {
 };
 var ALL_OCR_MODELS = [OCR_DET_MODEL, OCR_REC_MODEL, OCR_REC_DICT];
 function getOcrModelsDir() {
-  return _chunk2BZFKIRBcjs.getModelsDir.call(void 0, "ppocr");
+  return _chunkCEIATRDWcjs.getModelsDir.call(void 0, "ppocr");
 }
 async function ensureOcrModels(onProgress) {
-  return _chunk2BZFKIRBcjs.ensureModelsIn.call(void 0, getOcrModelsDir(), ALL_OCR_MODELS, onProgress);
+  return _chunkCEIATRDWcjs.ensureModelsIn.call(void 0, getOcrModelsDir(), ALL_OCR_MODELS, onProgress);
 }
 async function ocrModelsCached() {
   const dir = getOcrModelsDir();
@@ -1927,6 +1928,60 @@ function hasHorizontalLine(horizontals, y, leftX, rightX, vertexRadius) {
   return false;
 }
 
+// src/pdf/script-items.ts
+function scriptKindsOfLine(sorted) {
+  const kinds = sorted.map(() => null);
+  if (sorted.length < 2) return kinds;
+  const weight = /* @__PURE__ */ new Map();
+  for (const it of sorted) {
+    const k = Math.round(it.fontSize * 2) / 2;
+    weight.set(k, (_nullishCoalesce(weight.get(k), () => ( 0))) + it.text.trim().length);
+  }
+  let bodyFs = 0, best = -1;
+  for (const [fs, n] of weight) if (n > best || n === best && fs > bodyFs) {
+    best = n;
+    bodyFs = fs;
+  }
+  if (bodyFs <= 0) return kinds;
+  const ys = sorted.filter((it) => Math.abs(it.fontSize - bodyFs) <= bodyFs * 0.12).map((it) => it.y).sort((a, b) => a - b);
+  const bodyY = ys[ys.length >> 1];
+  const isBody = (it) => it.fontSize > bodyFs * 0.85;
+  for (let i = 1; i < sorted.length; i++) {
+    const it = sorted[i], prev = sorted[i - 1];
+    const t = it.text.trim();
+    if (!t || t.length > 16 || isBody(it)) continue;
+    if (!isBody(prev) && !kinds[i - 1]) continue;
+    const gap = it.x - (prev.x + prev.w);
+    if (gap > bodyFs * 0.35 || gap < -bodyFs * 0.5) continue;
+    const dy = it.y - bodyY;
+    if (dy >= bodyFs * 0.15) kinds[i] = "sup";
+    else if (dy <= -bodyFs * 0.08) kinds[i] = "sub";
+  }
+  return kinds;
+}
+function tagScripts(text, lines) {
+  const marks = [];
+  let cursor = 0;
+  for (const line of lines) {
+    const kinds = scriptKindsOfLine(line);
+    for (let k = 0; k < line.length; k++) {
+      const t = line[k].text;
+      if (!t) continue;
+      const at = text.indexOf(t, cursor);
+      if (at < 0 || /\S/.test(text.slice(cursor, at))) return text;
+      if (kinds[k]) marks.push([at, at + t.length, kinds[k]]);
+      cursor = at + t.length;
+    }
+  }
+  if (!marks.length) return text;
+  let out = text;
+  for (let m = marks.length - 1; m >= 0; m--) {
+    const [s, e, kind] = marks[m];
+    out = out.slice(0, s) + `<${kind}>` + out.slice(s, e) + `</${kind}>` + out.slice(e);
+  }
+  return _chunk62W7QCEXcjs.tidyScriptTags.call(void 0, out);
+}
+
 // src/pdf/line-wrap.ts
 var MARKUP = /<\/?u>|~~/g;
 var bump = (m, k) => {
@@ -2258,7 +2313,8 @@ function cellTextToString(items, wrap) {
     }
     return result;
   });
-  if (!wrap) return mergeCellTextLines(textLines);
+  const scripted = (s) => tagScripts(s, merged.map((l) => sortLineByX([...l])));
+  if (!wrap) return scripted(mergeCellTextLines(textLines));
   let contentLeft = Infinity;
   for (const it of items) if (it.x < contentLeft) contentLeft = it.x;
   const lineEnds = merged.map((line) => {
@@ -2271,7 +2327,7 @@ function cellTextToString(items, wrap) {
   let cellRight = -Infinity;
   for (const e of lineEnds) if (e.right > cellRight) cellRight = e.right;
   const wraps = lineEnds.slice(0, -1).map((a, i) => cellLineWraps(wrap.box, contentLeft, a.right, a.fontSize, lineEnds[i + 1].firstCharW) || lineEnds.length >= 3 && !textLines[i + 1].startsWith("(") && cellLineFills(wrap.box, contentLeft, cellRight, a.right, a.fontSize));
-  return mergeCellTextLines(textLines, { wraps, lex: wrap.lex });
+  return scripted(mergeCellTextLines(textLines, { wraps, lex: wrap.lex }));
 }
 function mergeSuperscriptRows(lines) {
   if (lines.length <= 1) return lines;
@@ -2822,6 +2878,13 @@ function mergeSuperscriptLines(lines) {
     for (const i of line) total += i.text.trim().length;
     return total > 0 && total <= 10;
   };
+  const isMarkers = (line, host) => line.length > 1 && line.length <= 16 && line.every((i) => {
+    if (i.text.trim().length > 3 || !i.text.trim()) return false;
+    return host.some((h) => {
+      const g = i.x - (h.x + h.w);
+      return g <= h.fontSize * 0.35 && g >= -h.fontSize * 0.1;
+    });
+  });
   const result = [lines[0]];
   for (let i = 1; i < lines.length; i++) {
     const prev = result[result.length - 1];
@@ -2829,8 +2892,8 @@ function mergeSuperscriptLines(lines) {
     const a = band(prev);
     const b = band(curr);
     const overlap2 = Math.min(a.top, b.top) - Math.max(a.bottom, b.bottom);
-    const prevIsFrag = isFrag(prev) && a.height <= b.height * 0.8 && overlap2 >= a.height * 0.5;
-    const currIsFrag = isFrag(curr) && b.height <= a.height * 0.8 && overlap2 >= b.height * 0.5;
+    const prevIsFrag = (isFrag(prev) || isMarkers(prev, curr)) && a.height <= b.height * 0.8 && overlap2 >= a.height * 0.5;
+    const currIsFrag = (isFrag(curr) || isMarkers(curr, prev)) && b.height <= a.height * 0.8 && overlap2 >= b.height * 0.5;
     if (prevIsFrag || currIsFrag) {
       result[result.length - 1] = [...prev, ...curr];
     } else {
@@ -3028,10 +3091,13 @@ function pushLineParagraphs(out, yLines, pageNum, lex) {
   for (let i = 0; i < lines.length; ) {
     let text = lines[i].text;
     const items = [...lines[i].items];
+    const srcLines = [lines[i].items];
     for (; i + 1 < lines.length && joins[i] !== "\n"; i++) {
       text += joins[i] + lines[i + 1].text;
       items.push(...lines[i + 1].items);
+      srcLines.push(lines[i + 1].items);
     }
+    text = tagScripts(text, srcLines.map((l) => sortLineByX([...l])));
     const block = { type: "paragraph", text, pageNumber: pageNum, bbox: computeBBox(items, pageNum), style: dominantStyle(items) };
     const faces = /* @__PURE__ */ new Map();
     for (const it of items) faces.set(it.fontName, (_nullishCoalesce(faces.get(it.fontName), () => ( 0))) + it.text.length);
@@ -3065,9 +3131,9 @@ function detectHeadings(blocks, medianFontSize) {
     if (/^\d+$/.test(text) && !isChapterNumber(block, blocks[bi + 1], medianFontSize)) continue;
     const ratio = block.style.fontSize / medianFontSize;
     let level = 0;
-    if (ratio >= _chunkDO74GYSMcjs.HEADING_RATIO_H1) level = 1;
-    else if (ratio >= _chunkDO74GYSMcjs.HEADING_RATIO_H2) level = 2;
-    else if (ratio >= _chunkDO74GYSMcjs.HEADING_RATIO_H3) level = 3;
+    if (ratio >= _chunk62W7QCEXcjs.HEADING_RATIO_H1) level = 1;
+    else if (ratio >= _chunk62W7QCEXcjs.HEADING_RATIO_H2) level = 2;
+    else if (ratio >= _chunk62W7QCEXcjs.HEADING_RATIO_H3) level = 3;
     if (level > 0) {
       block.type = "heading";
       block.level = level;
@@ -3080,7 +3146,7 @@ function isChapterNumber(block, next, medianFontSize) {
   const nextSize = _nullishCoalesce(_optionalChain([next, 'optionalAccess', _76 => _76.style, 'optionalAccess', _77 => _77.fontSize]), () => ( 0));
   const text = _nullishCoalesce(_optionalChain([next, 'optionalAccess', _78 => _78.text, 'optionalAccess', _79 => _79.trim, 'call', _80 => _80()]), () => ( ""));
   const a = block.bbox, b = _optionalChain([next, 'optionalAccess', _81 => _81.bbox]);
-  return /^\d{1,2}$/.test(_nullishCoalesce(_optionalChain([block, 'access', _82 => _82.text, 'optionalAccess', _83 => _83.trim, 'call', _84 => _84()]), () => ( ""))) && size >= medianFontSize * 2.5 && !!next && (next.type === "paragraph" || next.type === "heading") && next.pageNumber === block.pageNumber && nextSize >= medianFontSize * _chunkDO74GYSMcjs.HEADING_RATIO_H1 && text.length > 0 && text.length <= 80 && !/^[\d\s.]+$/.test(text) && !!a && !!b && a.y > b.y && a.y - (b.y + b.height) <= size * 2;
+  return /^\d{1,2}$/.test(_nullishCoalesce(_optionalChain([block, 'access', _82 => _82.text, 'optionalAccess', _83 => _83.trim, 'call', _84 => _84()]), () => ( ""))) && size >= medianFontSize * 2.5 && !!next && (next.type === "paragraph" || next.type === "heading") && next.pageNumber === block.pageNumber && nextSize >= medianFontSize * _chunk62W7QCEXcjs.HEADING_RATIO_H1 && text.length > 0 && text.length <= 80 && !/^[\d\s.]+$/.test(text) && !!a && !!b && a.y > b.y && a.y - (b.y + b.height) <= size * 2;
 }
 function mergeStackedHeadingLines(blocks, medianFontSize) {
   for (let i = 0; i < blocks.length - 1; ) {
@@ -5161,16 +5227,16 @@ 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 = _chunkYRRBA4HWcjs.safeMax.call(void 0, gaps);
+  const maxGap = _chunkSTM3ORY7cjs.safeMax.call(void 0, gaps);
   const avgLen = items.reduce((s, i) => s + i.text.length, 0) / items.length;
   if (items.length >= 4 && maxGap < 40 && avgLen < 5) return true;
-  const minGap = _chunkYRRBA4HWcjs.safeMin.call(void 0, gaps);
+  const minGap = _chunkSTM3ORY7cjs.safeMin.call(void 0, gaps);
   return minGap > 0 && maxGap - minGap <= Math.max(2, maxGap * 0.15) && items.filter((i) => /^\P{L}*\p{Ll}/u.test(i.text)).length * 2 >= items.length;
 }
 function detectColumns(yLines) {
   const allItems = yLines.flat();
   if (allItems.length === 0) return null;
-  const pageWidth = _chunkYRRBA4HWcjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkYRRBA4HWcjs.safeMin.call(void 0, allItems.map((i) => i.x));
+  const pageWidth = _chunkSTM3ORY7cjs.safeMax.call(void 0, allItems.map((i) => i.x + i.w)) - _chunkSTM3ORY7cjs.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++) {
@@ -5893,17 +5959,23 @@ function splitTwoColumnProse(items, cutX) {
     if (last && Math.abs(last[0].y - c.y) <= 3) last.push(c);
     else crossLines.push([c]);
   }
-  const bandItem = (arr) => arr.filter((i) => {
+  const scriptOf = (i, cl) => {
+    const fs = cl[0].fontSize, dy = i.y - cl[0].y;
+    if (i.fontSize > fs * 0.85 || dy > fs * 0.6 || dy < -fs * 0.35) return false;
+    return cl.some((c) => c !== i && c.fontSize > fs * 0.85 && i.x - (c.x + c.w) <= fs * 0.35 && i.x - (c.x + c.w) >= -fs * 0.1);
+  };
+  const bandItem = (arr, fits) => arr.filter((i) => {
     for (const cl of crossLines) {
-      if (Math.abs(cl[0].y - i.y) <= 3) {
+      if (fits(i, cl)) {
         cl.push(i);
         return false;
       }
     }
     return true;
   });
-  const leftRest = bandItem(left);
-  const rightRest = bandItem(right);
+  const sameY = (i, cl) => Math.abs(cl[0].y - i.y) <= 3;
+  const leftRest = bandItem(bandItem(left, sameY), scriptOf);
+  const rightRest = bandItem(bandItem(right, sameY), scriptOf);
   const boundYs = crossLines.map((cl) => cl[0].y);
   const bandOf = (y) => {
     let k = 0;
@@ -5915,7 +5987,7 @@ function splitTwoColumnProse(items, cutX) {
     const L = leftRest.filter((i) => bandOf(i.y) === k);
     const R = rightRest.filter((i) => bandOf(i.y) === k);
     groups.push(...columnPair(L, R));
-    if (k < crossLines.length) groups.push(crossLines[k]);
+    if (k < crossLines.length) groups.push(crossLines[k].sort((a, b) => b.y - a.y || a.x - b.x));
   }
   return groups;
 }
@@ -6691,7 +6763,7 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
     if (gutterX !== null) {
       const gx = gutterX;
       const allY = remaining.map((i) => i.y);
-      const pageH = _chunkYRRBA4HWcjs.safeMax.call(void 0, allY) - _chunkYRRBA4HWcjs.safeMin.call(void 0, allY);
+      const pageH = _chunkSTM3ORY7cjs.safeMax.call(void 0, allY) - _chunkSTM3ORY7cjs.safeMin.call(void 0, allY);
       const gapThreshold = Math.max(15, pageH * 0.03);
       const sides = [
         remaining.filter((i) => i.x + i.w <= gx),
@@ -6711,7 +6783,7 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
       finalTextBlocks = detectListBlocks(textBlocks);
     } else {
       const allY = remaining.map((i) => i.y);
-      const pageH = _chunkYRRBA4HWcjs.safeMax.call(void 0, allY) - _chunkYRRBA4HWcjs.safeMin.call(void 0, allY);
+      const pageH = _chunkSTM3ORY7cjs.safeMax.call(void 0, allY) - _chunkSTM3ORY7cjs.safeMin.call(void 0, allY);
       const groups = xyCutOrder(remaining, Math.max(15, pageH * 0.03));
       const textBlocks = [];
       for (const group of groups) {
@@ -7025,7 +7097,7 @@ function extractPageBlocksFallback(items, pageNum, fullPage = false, detectTable
       blocks.push(...columnTextToBlocks(tableText, pageNum, bbox, dominantStyle(items)));
     } else {
       const allY = items.map((i) => i.y);
-      const pageHeight = _chunkYRRBA4HWcjs.safeMax.call(void 0, allY) - _chunkYRRBA4HWcjs.safeMin.call(void 0, allY);
+      const pageHeight = _chunkSTM3ORY7cjs.safeMax.call(void 0, allY) - _chunkSTM3ORY7cjs.safeMin.call(void 0, allY);
       const gapThreshold = Math.max(15, pageHeight * 0.03);
       const orderedGroups = proseCutX !== null ? splitTwoColumnProse(items, proseCutX) : xyCutOrder(items, gapThreshold);
       for (const group of orderedGroups) {
@@ -7087,4 +7159,4 @@ function extractPageBlocksFallback(items, pageNum, fullPage = false, detectTable
 
 
 exports.OCR_DET_MODEL = OCR_DET_MODEL; exports.OCR_REC_MODEL = OCR_REC_MODEL; exports.OCR_REC_DICT = OCR_REC_DICT; exports.getOcrModelsDir = getOcrModelsDir; exports.ensureOcrModels = ensureOcrModels; exports.ocrModelsCached = ocrModelsCached; exports.parseCharacterDict = parseCharacterDict; exports.extractLines = extractLines; exports.extractImageRegions = extractImageRegions; exports.filterHiddenText = filterHiddenText; exports.collapseEvenSpacing = collapseEvenSpacing; exports.normalizeItems = normalizeItems; exports.WrapLexicon = WrapLexicon; exports.startsNewItem = startsNewItem; exports.wrapJoiner = wrapJoiner; exports.joinPageBreakWraps = joinPageBreakWraps; exports.latinSoftWrap = latinSoftWrap; exports.TOC_BLOCKS = TOC_BLOCKS; exports.computeMedianFontSizeFromFreq = computeMedianFontSizeFromFreq; exports.detectHeadings = detectHeadings; exports.mergeStackedHeadingLines = mergeStackedHeadingLines; exports.detectTypographyHeadings = detectTypographyHeadings; exports.detectDocumentStyleHeadings = detectDocumentStyleHeadings; exports.detectSiblingStyleHeadings = detectSiblingStyleHeadings; exports.detectRepeatedPageLabels = detectRepeatedPageLabels; exports.detectPageLeadHeadings = detectPageLeadHeadings; exports.refineDocumentStyleHeadings = refineDocumentStyleHeadings; exports.detectMarkerHeadings = detectMarkerHeadings; exports.detectTableCaptions = detectTableCaptions; exports.detectKoreanListBlocks = detectKoreanListBlocks; exports.removeHeaderFooterBlocks = removeHeaderFooterBlocks; exports.CLIP_TABLES = CLIP_TABLES; exports.TABLE_COLXS = TABLE_COLXS; exports.EMPTY_PARTS = EMPTY_PARTS; exports.IMAGE_CELLS = IMAGE_CELLS; exports.FILLER_CELLS = FILLER_CELLS; exports.CONT_PARTS = CONT_PARTS; exports.CELL_LINES = CELL_LINES; exports.trimTrailingEmptyTableCols = trimTrailingEmptyTableCols; exports.extractPageBlocksWithLines = extractPageBlocksWithLines;
-//# sourceMappingURL=chunk-7IRHSRPU.cjs.map
+//# sourceMappingURL=chunk-WDQTFAXY.cjs.map