kordoc 4.18.1 → 4.18.3

package/dist/{chunk-WWOTHS2E.js → chunk-BJ6RYYXV.js}

@@ -1,15 +1,88 @@
-#!/usr/bin/env node
 import {
   CELL_EDGES,
   HEADING_RATIO_H1,
   HEADING_RATIO_H2,
   HEADING_RATIO_H3,
   tidyScriptTags
-} from "./chunk-TMC4A3FM.js";
+} from "./chunk-PSJLB3SZ.js";
+import {
+  ensureModelsIn,
+  getModelsDir
+} from "./chunk-3FZGL3IL.js";
 import {
   safeMax,
   safeMin
-} from "./chunk-I7YWXECO.js";
+} from "./chunk-CAN5P6N6.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";
@@ -3404,7 +3477,9 @@ function findVerticalCutWithOutlierFilter(items, minGap) {
     const filtered = items.filter((i) => i.w >= narrowThreshold);
     if (filtered.length >= 2 && filtered.length < items.length && filtered.length >= items.length * 0.7) {
       const filteredCut = findVerticalCut(filtered);
-      if (filteredCut.gap > edgeCut.gap && filteredCut.gap >= minGap) {
+      const sameRow = (a, b) => Math.abs(a.y - b.y) <= Math.max(a.fontSize, b.fontSize) * 0.5;
+      const inLine = items.some((r) => r.w < narrowThreshold && r.x < filteredCut.position && r.x + r.w > filteredCut.position && filtered.some((o) => sameRow(o, r) && o.x + o.w <= r.x + 1) && filtered.some((o) => sameRow(o, r) && o.x >= r.x + r.w - 1));
+      if (filteredCut.gap > edgeCut.gap && filteredCut.gap >= minGap && !inLine) {
         return filteredCut;
       }
     }
@@ -4124,7 +4199,7 @@ function detectSpecialKoreanTables(blocks) {
 }
 function removeHeaderFooterBlocks(blocks, pageHeights, warnings, notes, tables = false) {
   const ZONE_RATIO = 0.12;
-  const MIN_REPEAT = 3;
+  const MIN_REPEAT2 = 3;
   const topEntries = [];
   const bottomEntries = [];
   for (let bi = 0; bi < blocks.length; bi++) {
@@ -4161,7 +4236,7 @@ function removeHeaderFooterBlocks(blocks, pageHeights, warnings, notes, tables =
       const pages = [...patternPages.get(p) ?? []];
       const span = pages.length ? Math.max(...pages) - Math.min(...pages) + 1 : 0;
       const pageNumbered = !tables && (patternNumbers.get(p)?.size ?? 0) > 1;
-      if (count >= MIN_REPEAT && pages.length >= MIN_REPEAT && (pages.length >= span * 0.4 || pageNumbered)) {
+      if (count >= MIN_REPEAT2 && pages.length >= MIN_REPEAT2 && (pages.length >= span * 0.4 || pageNumbered)) {
         repeatedPatterns.add(p);
       }
     }
@@ -4309,6 +4384,78 @@ function mergeSliverColumns(grid, colXs, maxWidth) {
   return merged;
 }
 
+// src/pdf/side-tabs.ts
+var BAND = 0.12;
+var MAX_CHARS = 10;
+var MIN_REPEAT = 3;
+var RUN_GAP = 5;
+var TAB_CELL_CHARS = 24;
+var TAB_MIN_HEIGHT = 0.3;
+var SIDE_TAB_TABLES = /* @__PURE__ */ new WeakSet();
+function isSideTabTable(box, table, others, pageWidth, pageHeight) {
+  if (table.cols !== 1 || box.x2 - box.x1 > pageWidth * BAND || box.y2 - box.y1 < pageHeight * TAB_MIN_HEIGHT) return false;
+  const side = box.x1 >= pageWidth * (1 - BAND) ? "R" : box.x2 <= pageWidth * BAND ? "L" : null;
+  if (!side) return false;
+  const texts = table.cells.flat().map((c) => c.text.replace(/\s+/g, ""));
+  if (!texts.some(Boolean) || texts.some((t) => [...t].length > TAB_CELL_CHARS)) return false;
+  return others.every((it) => it.y < box.y1 || it.y > box.y2 || (side === "R" ? it.x + it.w <= box.x1 + 1 : it.x >= box.x2 - 1));
+}
+function removeSideTabs(blocks, pageWidths) {
+  const edges = /* @__PURE__ */ new Map();
+  for (const b of blocks) {
+    if (!b.bbox || b.pageNumber === void 0 || [...b.text?.trim() ?? ""].length <= MAX_CHARS) continue;
+    const e = edges.get(b.pageNumber) ?? { lefts: [], rights: [] };
+    e.lefts.push(b.bbox.x);
+    e.rights.push(b.bbox.x + b.bbox.width);
+    edges.set(b.pageNumber, e);
+  }
+  const q = (xs, p) => {
+    const v = [...xs].sort((a, b) => a - b);
+    return v[Math.min(v.length - 1, Math.floor(p * v.length))];
+  };
+  const body = /* @__PURE__ */ new Map();
+  for (const [page, e] of edges) body.set(page, { left: q(e.lefts, 0.2), right: q(e.rights, 0.8) });
+  const all = [...body.values()];
+  const fallback = all.length ? { left: q(all.map((e) => e.left), 0.5), right: q(all.map((e) => e.right), 0.5) } : void 0;
+  const key = (b) => {
+    const w = b.pageNumber !== void 0 ? pageWidths.get(b.pageNumber) : void 0;
+    const text = b.text?.trim();
+    const area = (b.pageNumber !== void 0 ? body.get(b.pageNumber) : void 0) ?? fallback;
+    if (b.type === "table") {
+      if (!w || !b.bbox || !b.table || !SIDE_TAB_TABLES.has(b.table)) return null;
+      return `
${b.bbox.x >= w * (1 - BAND) ? "R" : "L"}${Math.round(b.bbox.x / 8)}`;
+    }
+    if (!w || !area || !b.bbox || !text || [...text].length > MAX_CHARS) return null;
+    const side = b.bbox.x + b.bbox.width <= Math.min(w * BAND, area.left - 1) ? "L" : b.bbox.x >= Math.max(w * (1 - BAND), area.right + 1) ? "R" : null;
+    return side && text.replace(/\s+/g, "");
+  };
+  const pages = /* @__PURE__ */ new Map();
+  for (const b of blocks) {
+    const k = key(b);
+    if (k) pages.set(k, (pages.get(k) ?? /* @__PURE__ */ new Set()).add(b.pageNumber));
+  }
+  const running = /* @__PURE__ */ new Set();
+  for (const [k, ps] of pages) {
+    const list = [...ps].sort((a, b) => a - b);
+    const steps = list.slice(1).map((p, i) => p - list[i]).sort((a, b) => a - b);
+    const gap = Math.max(RUN_GAP, (steps[steps.length >> 1] ?? 1) * 3);
+    let run = [];
+    const flush = () => {
+      if (run.length >= MIN_REPEAT && run.length >= (run[run.length - 1] - run[0] + 1) * 0.4) for (const p of run) running.add(`${p}\0${k}`);
+      run = [];
+    };
+    for (const p of list) {
+      if (run.length && p - run[run.length - 1] > gap) flush();
+      run.push(p);
+    }
+    flush();
+  }
+  return running.size ? blocks.filter((b) => {
+    const k = key(b);
+    return !k || !running.has(`${b.pageNumber}\0${k}`);
+  }) : blocks;
+}
+
 // src/pdf/table-roles.ts
 var TOC_BLOCKS = /* @__PURE__ */ new WeakSet();
 var PAGE_LABEL = /^(?:\d{1,4}|[ivxlc]{1,7})$/i;
@@ -4415,6 +4562,8 @@ var MIN_GAP_FACTOR = 2;
 var MIN_GAP_ABSOLUTE = 20;
 var MIN_COL_FILL_RATIO = 0.4;
 function detectClusterTables(items, pageNum, rejected) {
+  const tab = sideTabGlyphs(items);
+  if (tab.size) items = items.filter((i) => !tab.has(i));
   if (items.length < MIN_ROWS * MIN_COLS) return [];
   const { merged, originMap } = mergeEvenSpacedClusters(items);
   const rows = mergeOverlappingRows(groupByBaseline(merged));
@@ -4474,6 +4623,36 @@ function detectClusterTables(items, pageNum, rejected) {
     return !prose;
   });
 }
+var SIDE_TAB_MIN = 6;
+function sideTabGlyphs(items) {
+  const found = /* @__PURE__ */ new Set();
+  const singles = items.filter((i) => i.fontSize > 0 && [...i.text.trim()].length === 1);
+  if (singles.length < SIDE_TAB_MIN) return found;
+  const cols = [];
+  for (const it of [...singles].sort((a, b) => a.x - b.x)) {
+    const c = cols.find((col) => Math.abs(col[0].x - it.x) <= Math.max(1.5, it.fontSize * 0.25));
+    if (c) c.push(it);
+    else cols.push([it]);
+  }
+  for (const col of cols) {
+    if (col.length < SIDE_TAB_MIN) continue;
+    if (col.filter((g) => /[\p{L}\p{N}]/u.test(g.text)).length < col.length * 0.8) continue;
+    col.sort((a, b) => b.y - a.y);
+    const fs = [...col.map((i) => i.fontSize)].sort((a, b) => a - b)[col.length >> 1];
+    if (col.slice(1).some((it, k) => col[k].y - it.y > fs * 3)) continue;
+    const members = new Set(col);
+    const top = col[0].y + fs, bottom = col[col.length - 1].y - fs;
+    const others = items.filter((i) => !members.has(i) && i.y <= top && i.y >= bottom);
+    const left = Math.min(...col.map((i) => i.x)), right = Math.max(...col.map((i) => i.x + i.w));
+    const rest = items.filter((i) => !members.has(i) && i.text.trim());
+    const outside = rest.every((i) => i.x >= right + fs) || rest.every((i) => i.x + i.w <= left - fs);
+    if (!outside) continue;
+    const alone = col.filter((g) => !others.some((i) => Math.abs(i.y - g.y) <= Y_TOL)).length;
+    if (alone < col.length * 0.4) continue;
+    for (const g of col) found.add(g);
+  }
+  return found;
+}
 function detectCompactStatisticalTables(rows, pageNum) {
   const found = [];
   const parts = /* @__PURE__ */ new Map();
@@ -6661,6 +6840,11 @@ function extractPageBlocksWithLines(items, pageNum, opList, pageWidth, pageHeigh
     if (carry) carry.clip = void 0;
     return [];
   }
+  const tab = sideTabGlyphs(items);
+  if (tab.size) {
+    const glyphs = [...tab].sort((a, b) => b.y - a.y).map((g) => ({ type: "paragraph", text: g.text.trim(), pageNumber: pageNum, bbox: computeBBox([g], pageNum), style: dominantStyle([g]) }));
+    return [...glyphs, ...extractPageBlocksWithLines(items.filter((i) => !tab.has(i)), pageNum, opList, pageWidth, pageHeight, extraLines, detectTables, carry, lexicon)];
+  }
   const lex = lexicon ?? new WrapLexicon();
   for (const line of streamLines(items)) lex.addLine(mergeLineSimple(line));
   const extracted = extractLines(opList.fnArray, opList.argsArray);
@@ -7022,6 +7206,12 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
     const emptyPart = !hasContent && !!grid.cells && !grid.clipParent && !nestedAttached;
     if (!hasContent && !emptyPart) continue;
     if (emptyPart) EMPTY_PARTS.add(irTable);
+    const ownItems = numCols === 1 ? new Set(tableItems) : null;
+    if (ownItems && isSideTabTable(grid.bbox, irTable, items.filter((it) => !ownItems.has(it)), pageWidth, pageHeight)) {
+      SIDE_TAB_TABLES.add(irTable);
+      blocks.push({ type: "table", table: irTable, pageNumber: pageNum, bbox: { page: pageNum, x: grid.bbox.x1, y: grid.bbox.y1, width: gridW, height: grid.bbox.y2 - grid.bbox.y1 } });
+      continue;
+    }
     if (grid.clipParent) {
       const nb = { page: pageNum, x: grid.bbox.x1, y: grid.bbox.y1, width: grid.bbox.x2 - grid.bbox.x1, height: grid.bbox.y2 - grid.bbox.y1 };
       pendingNested.push({ parent: grid.clipParent, block: { type: "table", table: irTable, pageNumber: pageNum, bbox: nb } });
@@ -7491,6 +7681,13 @@ function extractPageBlocksFallback(items, pageNum, fullPage = false, detectTable
 }
 
 export {
+  OCR_DET_MODEL,
+  OCR_REC_MODEL,
+  OCR_REC_DICT,
+  getOcrModelsDir,
+  ensureOcrModels,
+  ocrModelsCached,
+  parseCharacterDict,
   extractLines,
   extractImageRegions,
   filterHiddenText,
@@ -7529,6 +7726,7 @@ export {
   trimTrailingEmptyTableCols,
   NO_EDGES,
   joinCellEdges,
+  removeSideTabs,
   extractPageBlocksWithLines
 };
-//# sourceMappingURL=chunk-WWOTHS2E.js.map
+//# sourceMappingURL=chunk-BJ6RYYXV.js.map