kordoc 4.17.2 → 4.17.4

package/dist/{chunk-3N272C5L.js → chunk-YBIWPCUX.js}

@@ -4,15 +4,15 @@ import {
   HEADING_RATIO_H2,
   HEADING_RATIO_H3,
   tidyScriptTags
-} from "./chunk-UX3HHAWH.js";
+} from "./chunk-NNNRN6BY.js";
 import {
   ensureModelsIn,
   getModelsDir
-} from "./chunk-5JZEY5IL.js";
+} from "./chunk-VJP7O6BL.js";
 import {
   safeMax,
   safeMin
-} from "./chunk-YGKC6E6A.js";
+} from "./chunk-PTMYCK7R.js";
 
 // src/ocr/models.ts
 import { join } from "path";
@@ -2170,6 +2170,7 @@ var hasBatchim = (c) => {
   return k >= 0 && k < 11172 && k % 28 !== 0;
 };
 var CJK = /[\u3000-\u303F\u3040-\u30FF\u3400-\u4DBF\u4E00-\u9FFF\uF900-\uFAFF\uFF00-\uFFEF]/;
+var unspacedCjkBoundary = (prev, next) => CJK.test(prev.trimEnd().slice(-1)) && CJK.test(next.trimStart().charAt(0)) && !/[가-힣ㄱ-ㅎㅏ-ㅣ]/.test(prev + next);
 var CLOSE_TAIL = /[’”」』)\]〉》>]+$/;
 var TAIL = "[.,)\u300D\u300F\u2019\u201D]*";
 var PARTICLE = new RegExp(`^(?:\uB2E4\\.|\uC758|\uC5D0|\uC5D0\uC11C|\uC5D0\uB294|\uC5D0\uAC8C|\uC5D0\uB3C4|\uC5D0\uC11C\uB294|\uAE4C\uC9C0|\uBD80\uD130|\uCC98\uB7FC|\uB9C8\uB2E4|\uB9CC|\uB3C4|\uB4E4|\uB4E4\uC774|\uB4E4\uC758|\uB4E4\uC740|\uB4E4\uC744|\uB4E4\uC5D0\uAC8C|\uC11C\uB294|\uC11C\uB3C4|\uB77C\uACE0|\uB77C\uBA70|\uC740|\uB294|\uC744|\uB97C|\uC73C\uB85C|\uB85C|\uC73C\uB85C\uC11C|\uB85C\uC11C|\uC73C\uB85C\uC368|\uB85C\uC368)${TAIL}$`);
@@ -2219,7 +2220,7 @@ function wrapJoiner(prevText, nextText, lex) {
   const prev = prevText.replace(MARKUP, "").trimEnd(), next = nextText.replace(MARKUP, "").trimStart();
   const a = prev[prev.length - 1], b = next[0];
   if (!a || !b || /[,;:!?]/.test(a) || DATE_DAY_END.test(prev) && b !== "(") return " ";
-  if (CJK.test(a) && CJK.test(b) && !/[가-힣ㄱ-ㅎㅏ-ㅣ]/.test(prev + next)) return "";
+  if (unspacedCjkBoundary(prev, next)) return "";
   let s = prev.length;
   while (s > 0 && !/\s/.test(prev[s - 1])) s--;
   const left = prev.slice(s), right = next.match(/^\S+/)[0];
@@ -2262,17 +2263,32 @@ function joinPageBreakWraps(blocks, lex) {
   }
 }
 var HANGING = /[，。、：；！？）」』】〕》〉．]\s*$/;
+var MULTI_LEVEL_NUMBER = /^\d{1,2}(?:\.\d{1,3})+\.?\s/;
 var BODY_FULL_TOL = 0.25;
 var BODY_MIN_WIDTH_EM = 12;
 var BODY_MAX_PITCH_EM = 2;
 var BODY_PITCH_REL = 1.05;
 var BODY_MAX_PITCH_ABS_EM = 3.5;
 function bodyLineJoins(lines, lex) {
+  lines = lines.map((l) => {
+    const text = l.text.replace(MARKUP, "");
+    return text === l.text ? l : { ...l, text };
+  });
   let right = -Infinity;
   for (const l of lines) if (l.right > right) right = l.right;
   let inner = -Infinity;
   for (const l of lines) if (!HANGING.test(l.text) && l.right > inner) inner = l.right;
   const full = (l) => right - l.right < BODY_FULL_TOL * l.fontSize || right - inner <= 1.2 * l.fontSize && inner - l.right < BODY_FULL_TOL * l.fontSize;
+  const nextWordNoRoom = (a, b) => {
+    const w = b.text.trim().match(/^\S+/)?.[0];
+    const n = [...b.text.trim()].length;
+    if (!w || n === 0) return false;
+    const em = (b.right - b.left) / n;
+    const advance = unspacedCjkBoundary(a.text, b.text) ? 1 : [...w].length + 1;
+    return right - a.right < advance * em;
+  };
+  const candidates = lines.filter((l, i) => full(l) || i + 1 < lines.length && nextWordNoRoom(l, lines[i + 1]));
+  const ragged = candidates.filter(full).length * 2 < candidates.length;
   const pitch = (k) => lines[k].y - lines[k + 1].y;
   const sameSize = (k) => Math.abs(lines[k + 1].fontSize - lines[k].fontSize) <= 0.15 * lines[k].fontSize;
   let p1 = Infinity, p2 = Infinity, i1 = -1;
@@ -2291,8 +2307,9 @@ function bodyLineJoins(lines, lex) {
     const fs = a.fontSize;
     const others = i === i1 ? p2 : p1;
     const maxPitch = Number.isFinite(others) ? Math.min(BODY_MAX_PITCH_ABS_EM * fs, Math.max(BODY_MAX_PITCH_EM * fs, others * BODY_PITCH_REL)) : BODY_MAX_PITCH_EM * fs;
-    const wraps = fs > 0 && full(a) && a.right - a.left >= BODY_MIN_WIDTH_EM * fs && a.y - b.y > 0 && a.y - b.y < maxPitch && Math.abs(b.fontSize - fs) <= 0.15 * fs && !startsNewItem(a.text, b.text);
-    out.push(wraps ? wrapJoiner(a.text, b.text, lex) : "\n");
+    const wordWrap = ragged && !full(a) && nextWordNoRoom(a, b) && !MULTI_LEVEL_NUMBER.test(b.text.trimStart());
+    const wraps = fs > 0 && (full(a) || wordWrap) && a.right - a.left >= BODY_MIN_WIDTH_EM * fs && a.y - b.y > 0 && a.y - b.y < maxPitch && Math.abs(b.fontSize - fs) <= 0.15 * fs && !startsNewItem(a.text, b.text);
+    out.push(!wraps ? "\n" : wordWrap ? unspacedCjkBoundary(a.text, b.text) ? "" : " " : wrapJoiner(a.text, b.text, lex));
   }
   return out;
 }
@@ -3341,15 +3358,18 @@ function detectTypographyHeadings(blocks) {
     const faceWeight = /* @__PURE__ */ new Map();
     for (const block of page) {
       if (block.type !== "paragraph" || !block.text || !block.style?.fontName) continue;
-      const face = block.style.fontName;
-      faceWeight.set(face, (faceWeight.get(face) ?? 0) + block.text.length);
+      const faces = FACE_CHARS.get(block);
+      if (faces) for (const [face, n] of faces) faceWeight.set(face, (faceWeight.get(face) ?? 0) + n);
+      else faceWeight.set(block.style.fontName, (faceWeight.get(block.style.fontName) ?? 0) + block.text.length);
     }
     const bodyFace = [...faceWeight].sort((a, b) => b[1] - a[1])[0]?.[0];
     if (!bodyFace || (faceWeight.get(bodyFace) ?? 0) < 250) continue;
     const sizeWeight = /* @__PURE__ */ new Map();
     for (const block of page) {
-      if (block.type !== "paragraph" || block.style?.fontName !== bodyFace || !block.style.fontSize) continue;
-      sizeWeight.set(block.style.fontSize, (sizeWeight.get(block.style.fontSize) ?? 0) + block.text.length);
+      if (block.type !== "paragraph" || !block.style?.fontSize) continue;
+      const faces = FACE_CHARS.get(block);
+      const chars = faces ? faces.get(bodyFace) ?? 0 : block.style.fontName === bodyFace ? block.text?.length ?? 0 : 0;
+      if (chars) sizeWeight.set(block.style.fontSize, (sizeWeight.get(block.style.fontSize) ?? 0) + chars);
     }
     const bodySize = [...sizeWeight].sort((a, b) => b[1] - a[1])[0]?.[0] ?? 0;
     for (let i = 0; i < page.length; i++) {
@@ -3458,7 +3478,9 @@ function detectRepeatedPageLabels(blocks) {
   for (const page of byPage.values()) {
     const bodyChars = /* @__PURE__ */ new Map();
     for (const b of page) if (b.type === "paragraph" && b.text && b.style?.fontName) {
-      bodyChars.set(b.style.fontName, (bodyChars.get(b.style.fontName) ?? 0) + b.text.length);
+      const faces = FACE_CHARS.get(b);
+      if (faces) for (const [face, n] of faces) bodyChars.set(face, (bodyChars.get(face) ?? 0) + n);
+      else bodyChars.set(b.style.fontName, (bodyChars.get(b.style.fontName) ?? 0) + b.text.length);
     }
     const bodyFace = [...bodyChars].sort((a, b) => b[1] - a[1])[0]?.[0];
     const candidates = page.filter((b) => b.type === "paragraph" && b.text && b.bbox && b.style?.fontName && b.style.fontName !== bodyFace && b.bbox.height <= (b.style.fontSize ?? 0) * 1.6 && b.text.trim().length >= 5 && b.text.trim().length <= 80);
@@ -7186,9 +7208,8 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
       continue;
     }
     let lastLeft = -1;
-    const beside = (b) => !!b.bbox && b.bbox.x + b.bbox.width <= box.x - 3 && b.bbox.y + b.bbox.height >= box.y && b.bbox.y <= box.y + box.height;
     units.forEach((unit, index) => {
-      if (unit.some(beside) && unit.every((b) => beside(b) || !!b.bbox && b.bbox.y + b.bbox.height < box.y)) lastLeft = index;
+      if (isProseBesidePanel(unit, box)) lastLeft = index;
     });
     const current = units.findIndex((unit) => unit.includes(sidebar));
     if (lastLeft > current) {
@@ -7204,6 +7225,10 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
   }
   return mergeAdjacentTableBlocks(ordered);
 }
+function isProseBesidePanel(unit, box) {
+  const beside = (b) => !!b.bbox && b.bbox.x + b.bbox.width <= box.x - 3 && b.bbox.y + b.bbox.height >= box.y && b.bbox.y <= box.y + box.height;
+  return unit.some(beside) && unit.every((b) => beside(b) || !!b.bbox && b.bbox.y < box.y);
+}
 var SIDEBAR_ITEMS = /* @__PURE__ */ new WeakMap();
 function clusterTableBlock(cr, source, pageNum, horizontals = []) {
   const b = cr.bbox;
@@ -7484,4 +7509,4 @@ export {
   joinCellEdges,
   extractPageBlocksWithLines
 };
-//# sourceMappingURL=chunk-3N272C5L.js.map
+//# sourceMappingURL=chunk-YBIWPCUX.js.map