kordoc 4.16.1 → 4.16.3

package/dist/{chunk-7FKPPIDZ.js → chunk-46H2PPG7.js}

@@ -1,13 +1,87 @@
-#!/usr/bin/env node
 import {
   HEADING_RATIO_H1,
   HEADING_RATIO_H2,
-  HEADING_RATIO_H3
-} from "./chunk-T77M2QP7.js";
+  HEADING_RATIO_H3,
+  tidyScriptTags
+} from "./chunk-3IB7LBHC.js";
+import {
+  ensureModelsIn,
+  getModelsDir
+} from "./chunk-AWWFGIBA.js";
 import {
   safeMax,
   safeMin
-} from "./chunk-3VLVNKJV.js";
+} from "./chunk-7NYRZI2Z.js";
+
+// src/ocr/models.ts
+import { join } from "path";
+import { stat } from "fs/promises";
+var OCR_DET_MODEL = {
+  name: "PP-OCRv5 mobile det",
+  filename: "det.onnx",
+  url: "https://huggingface.co/PaddlePaddle/PP-OCRv5_mobile_det_onnx/resolve/main/inference.onnx",
+  sha256: "a431985659dc921974177a95adcfbb90fd9e51989a5e04d70d0b75f597b6e61d",
+  sizeMb: 5
+};
+var OCR_REC_MODEL = {
+  name: "PP-OCRv5 korean rec",
+  filename: "rec_korean.onnx",
+  url: "https://huggingface.co/PaddlePaddle/korean_PP-OCRv5_mobile_rec_onnx/resolve/main/inference.onnx",
+  sha256: "92f0b7785e64fc9090106a241cf4c1eb97472824558272751b88a2a4476d3a08",
+  sizeMb: 13
+};
+var OCR_REC_DICT = {
+  name: "PP-OCRv5 korean dict",
+  filename: "rec_korean.yml",
+  url: "https://huggingface.co/PaddlePaddle/korean_PP-OCRv5_mobile_rec_onnx/resolve/main/inference.yml",
+  sha256: "f757fa1c40e99edcf27e9cce879b93eb2a51fa46f5ef39095689b8c37dd75998",
+  sizeMb: 1
+};
+var ALL_OCR_MODELS = [OCR_DET_MODEL, OCR_REC_MODEL, OCR_REC_DICT];
+function getOcrModelsDir() {
+  return getModelsDir("ppocr");
+}
+async function ensureOcrModels(onProgress) {
+  return ensureModelsIn(getOcrModelsDir(), ALL_OCR_MODELS, onProgress);
+}
+async function ocrModelsCached() {
+  const dir = getOcrModelsDir();
+  for (const spec of ALL_OCR_MODELS) {
+    try {
+      if (!(await stat(join(dir, spec.filename))).size) return false;
+    } catch {
+      return false;
+    }
+  }
+  return true;
+}
+function parseCharacterDict(yml) {
+  const lines = yml.split("\n");
+  const chars = [];
+  let inDict = false;
+  let dictIndent = -1;
+  for (const line of lines) {
+    if (!inDict) {
+      const m2 = /^(\s*)character_dict:\s*$/.exec(line);
+      if (m2) {
+        inDict = true;
+        dictIndent = m2[1].length;
+      }
+      continue;
+    }
+    const m = /^(\s*)- (.*)$/.exec(line);
+    if (m && m[1].length >= dictIndent) {
+      let v = m[2];
+      if (v.length >= 2 && (v.startsWith("'") && v.endsWith("'") || v.startsWith('"') && v.endsWith('"'))) {
+        v = v.slice(1, -1).replace(/''/g, "'");
+      }
+      chars.push(v);
+      continue;
+    }
+    if (line.trim() !== "") break;
+  }
+  return chars;
+}
 
 // src/pdf/line-extract.ts
 import { OPS } from "pdfjs-dist/legacy/build/pdf.mjs";
@@ -1854,6 +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, (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 tidyScriptTags(out);
+}
+
 // src/pdf/line-wrap.ts
 var MARKUP = /<\/?u>|~~/g;
 var bump = (m, k) => {
@@ -2185,7 +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) => {
@@ -2198,7 +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;
@@ -2749,6 +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];
@@ -2756,8 +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 {
@@ -2955,10 +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, (faces.get(it.fontName) ?? 0) + it.text.length);
@@ -5820,17 +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;
@@ -5842,7 +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;
 }
@@ -6973,6 +7118,13 @@ function extractPageBlocksFallback(items, pageNum, fullPage = false, detectTable
 }
 
 export {
+  OCR_DET_MODEL,
+  OCR_REC_MODEL,
+  OCR_REC_DICT,
+  getOcrModelsDir,
+  ensureOcrModels,
+  ocrModelsCached,
+  parseCharacterDict,
   extractLines,
   extractImageRegions,
   filterHiddenText,
@@ -7007,4 +7159,4 @@ export {
   trimTrailingEmptyTableCols,
   extractPageBlocksWithLines
 };
-//# sourceMappingURL=chunk-7FKPPIDZ.js.map
+//# sourceMappingURL=chunk-46H2PPG7.js.map