kordoc 4.18.4 → 4.18.6

package/dist/{chunk-MXA6YW2W.js → chunk-JYX46LZT.js}

@@ -1,88 +1,15 @@
+#!/usr/bin/env node
 import {
   CELL_EDGES,
   HEADING_RATIO_H1,
   HEADING_RATIO_H2,
   HEADING_RATIO_H3,
   tidyScriptTags
-} from "./chunk-XG4SYOZL.js";
-import {
-  ensureModelsIn,
-  getModelsDir
-} from "./chunk-ELOPDQES.js";
+} from "./chunk-CWJJTCCH.js";
 import {
   safeMax,
   safeMin
-} from "./chunk-7GR2GCJF.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;
-}
+} from "./chunk-NM6WGVOR.js";
 
 // src/pdf/line-extract.ts
 import { OPS } from "pdfjs-dist/legacy/build/pdf.mjs";
@@ -4888,6 +4815,7 @@ function detectClusterTables(items, pageNum, rejected) {
   });
 }
 var SIDE_TAB_MIN = 6;
+var SIDE_TAB_CAP_GROUPS = /* @__PURE__ */ new WeakMap();
 function sideTabGlyphs(items) {
   const found = /* @__PURE__ */ new Set();
   const singles = items.filter((i) => i.fontSize > 0 && [...i.text.trim()].length === 1);
@@ -4933,11 +4861,22 @@ function sideTabGlyphs(items) {
     if (!others.length) continue;
     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);
+    const cap = (i) => i.y > top && i.y - top <= fs * 3 && i.w <= fs * 3 && i.fontSize <= fs * 3 && [...i.text.trim()].length <= 3 && Math.abs(i.x + i.w / 2 - (left + right) / 2) <= fs * 0.5;
+    const bodyRest = rest.filter((i) => !cap(i));
+    const outside = bodyRest.every((i) => i.x >= right + fs) || bodyRest.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);
+    const legacyOutside = rest.every((i) => i.x >= right + fs) || rest.every((i) => i.x + i.w <= left - fs);
+    const head = !legacyOutside ? rest.find(cap) : void 0;
+    for (const g of col) {
+      found.add(g);
+      if (head) {
+        const caps = SIDE_TAB_CAP_GROUPS.get(found) ?? /* @__PURE__ */ new Map();
+        caps.set(g, head);
+        SIDE_TAB_CAP_GROUPS.set(found, caps);
+      }
+    }
   }
   return found;
 }
@@ -6901,6 +6840,55 @@ function frameLayoutBoxes(blocks) {
   }
   return boxes;
 }
+function takeFrameSpanningText(items, cutX, pageNum, lex) {
+  let minX = Infinity, maxR = -Infinity;
+  for (const item of items) {
+    minX = Math.min(minX, item.x);
+    maxR = Math.max(maxR, item.x + item.w);
+  }
+  const span = maxR - minX;
+  const rows = groupByY([...items].sort((a, b) => b.y - a.y || a.x - b.x)).map((line) => {
+    const runs = [];
+    for (const item of [...line].sort((a, b) => a.x - b.x)) {
+      const run = runs[runs.length - 1], prev = run?.[run.length - 1];
+      if (prev && item.x - (prev.x + prev.w) <= Math.max(prev.fontSize, item.fontSize) * 0.6) run.push(item);
+      else runs.push([item]);
+    }
+    return runs;
+  });
+  const taken = /* @__PURE__ */ new Set(), blocks = [];
+  for (let r = 0; r < rows.length; r++) for (const run of rows[r]) {
+    if (taken.has(run[0]) || run[0].x >= cutX || Math.max(...run.map((i) => i.x + i.w)) <= cutX) continue;
+    const lines = [run], fs = Math.max(...run.map((i) => i.fontSize));
+    const left = run[0].x, right = Math.max(...run.map((i) => i.x + i.w));
+    if (right - left < span * 0.75) continue;
+    let prevY = run[0].y;
+    for (let next = r + 1; next < rows.length && rows[next].length === 1; next++) {
+      const tail = rows[next][0], y = tail[0].y;
+      if (taken.has(tail[0]) || prevY - y > fs * 1.8 || tail[0].x < left - 1 || tail[0].x > left + fs * 2 || tail.some((i) => Math.abs(i.fontSize - fs) > fs * 0.15) || Math.max(...tail.map((i) => i.x + i.w)) > right + 1) break;
+      lines.push(tail);
+      prevY = y;
+    }
+    for (const line of lines) for (const item of line) taken.add(item);
+    pushLineParagraphs(blocks, mergeSuperscriptLines(lines), pageNum, lex);
+  }
+  let keep = 0;
+  for (const item of items) if (!taken.has(item)) items[keep++] = item;
+  items.length = keep;
+  return blocks;
+}
+function frameColumnTextBands(items, spanning, cutX) {
+  const tops = spanning.filter((b) => b.bbox).map((b) => b.bbox.y + b.bbox.height).sort((a, b) => b - a);
+  const bands = /* @__PURE__ */ new Map();
+  for (const item of items) {
+    let band = 0;
+    while (band < tops.length && tops[band] > item.y + item.h) band++;
+    const sides = bands.get(band) ?? [[], [], []];
+    sides[item.x + item.w <= cutX ? 0 : item.x >= cutX ? 2 : 1].push(item);
+    bands.set(band, sides);
+  }
+  return [...bands].sort((a, b) => a[0] - b[0]).flatMap(([, sides]) => sides.filter((side) => side.length));
+}
 function groupFrameParagraphUnits(units, frames) {
   const source = /* @__PURE__ */ new Map();
   for (const unit of units) for (const block of unit) source.set(block, unit);
@@ -7013,6 +7001,41 @@ function extendNestedShadedHeaders(clipGrids, lineGrids, horizontals, verticals,
   return extended;
 }
 
+// src/pdf/side-tab-blocks.ts
+function attachSideTabBlocks(tab, body, pageNum) {
+  const ordered = [...tab].sort((a, b) => b.y - a.y);
+  const caps = SIDE_TAB_CAP_GROUPS.get(tab);
+  const block = (items) => ({
+    type: "paragraph",
+    text: mergeLineSimple(items).trim(),
+    pageNumber: pageNum,
+    bbox: computeBBox(items, pageNum),
+    style: dominantStyle(items)
+  });
+  const prefix = ordered.filter((g) => !caps?.has(g)).map((g) => ({
+    type: "paragraph",
+    text: g.text.trim(),
+    pageNumber: pageNum,
+    bbox: computeBBox([g], pageNum),
+    style: dominantStyle([g])
+  }));
+  const groups = /* @__PURE__ */ new Map();
+  for (const g of ordered) {
+    const cap = caps?.get(g);
+    if (cap) groups.set(cap, [...groups.get(cap) ?? [], g]);
+  }
+  for (const [cap, glyphs] of groups) {
+    const blocks = groupByY(glyphs).map(block);
+    const anchor = body.reduce((best, b, i) => {
+      const box = b.bbox;
+      if (b.type === "image" || b.type === "separator" || !box || !b.text && !b.table || box.x > cap.x + 1 || box.x + box.width < cap.x + cap.w - 1 || box.y > cap.y + 1 || box.y + box.height < cap.y + cap.fontSize - 1) return best;
+      return best < 0 || box.width * box.height < body[best].bbox.width * body[best].bbox.height ? i : best;
+    }, -1);
+    body.splice(anchor >= 0 ? anchor + 1 : body.length, 0, ...blocks);
+  }
+  return [...prefix, ...body];
+}
+
 // src/pdf/page-regions.ts
 function splitTwoColumnProse(items, cutX) {
   const left = [];
@@ -7347,8 +7370,8 @@ function extractPageBlocksWithLines(items, pageNum, opList, pageWidth, pageHeigh
   }
   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, verifiedRightArrows)];
+    const body = extractPageBlocksWithLines(items.filter((i) => !tab.has(i)), pageNum, opList, pageWidth, pageHeight, extraLines, detectTables, carry, lexicon, verifiedRightArrows);
+    return attachSideTabBlocks(tab, body, pageNum);
   }
   const lex = lexicon ?? new WrapLexicon();
   for (const line of streamLines(items)) lex.addLine(mergeLineSimple(line));
@@ -7811,14 +7834,15 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
   if (remaining.length > 0) {
     if (gutterX !== null) {
       const gx = gutterX;
+      const spanning = detectListBlocks(takeFrameSpanningText(remaining, gx, pageNum, lex));
+      for (const block of spanning) {
+        finalTextBlocks.push(block);
+        groupSizes.push(1);
+      }
       const allY = remaining.map((i) => i.y);
       const pageH = safeMax(allY) - safeMin(allY);
       const gapThreshold = Math.max(15, pageH * 0.03);
-      const sides = [
-        remaining.filter((i) => i.x + i.w <= gx),
-        remaining.filter((i) => i.x < gx && i.x + i.w > gx),
-        remaining.filter((i) => i.x >= gx)
-      ];
+      const sides = frameColumnTextBands(remaining, spanning, gx);
       const textBlocks = [];
       for (const side of sides) {
         if (side.length === 0) continue;
@@ -7829,7 +7853,7 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
           groupSizes.push(groupBlocks.length);
         }
       }
-      finalTextBlocks = detectListBlocks(textBlocks);
+      finalTextBlocks.push(...detectListBlocks(textBlocks));
     } else {
       const allY = remaining.map((i) => i.y);
       const pageH = safeMax(allY) - safeMin(allY);
@@ -8178,13 +8202,6 @@ function extractPageBlocksFallback(items, pageNum, fullPage = false, detectTable
 }
 
 export {
-  OCR_DET_MODEL,
-  OCR_REC_MODEL,
-  OCR_REC_DICT,
-  getOcrModelsDir,
-  ensureOcrModels,
-  ocrModelsCached,
-  parseCharacterDict,
   extractLines,
   extractImageRegions,
   filterHiddenText,
@@ -8235,4 +8252,4 @@ export {
   removeSideTabs,
   extractPageBlocksWithLines
 };
-//# sourceMappingURL=chunk-MXA6YW2W.js.map
+//# sourceMappingURL=chunk-JYX46LZT.js.map