kordoc 4.17.0 → 4.17.2

package/dist/{parser-CTRB7PKD.js → parser-VH3F4EQU.js}

@@ -4,7 +4,7 @@ import {
   blocksToPages,
   encodePng,
   escapeLiteralDollar
-} from "./chunk-YZ5HHS6A.js";
+} from "./chunk-SDRKHEVT.js";
 import {
   hasRequestedPagesAfter,
   parsePageRange
@@ -50,18 +50,18 @@ import {
   startsNewItem,
   trimTrailingEmptyTableCols,
   wrapJoiner
-} from "./chunk-4XVNQAK5.js";
+} from "./chunk-3N272C5L.js";
 import {
   CELL_EDGES,
   CONTENT_CELLS,
   stripScriptTags,
   unframeLayoutTables
-} from "./chunk-MJBCXW5R.js";
-import "./chunk-EFO6JZY6.js";
+} from "./chunk-UX3HHAWH.js";
+import "./chunk-5JZEY5IL.js";
 import {
   KordocError,
   sanitizeHref
-} from "./chunk-ML4NHUR7.js";
+} from "./chunk-YGKC6E6A.js";
 
 // src/pdf/ocr-region-merge.ts
 function mergeOcrImageRegions(blocks, page, regions, ocrBlocks) {
@@ -681,6 +681,7 @@ function splitContactTables(blocks) {
 }
 
 // src/pdf/table-parts.ts
+var flowText = (text) => text.replace(/<\/?u>|~~/g, "");
 var PART_COL_TOL = 1;
 var PART_COL_SNAP = 0.3;
 function anchorsOf(table) {
@@ -728,7 +729,7 @@ var indexOf = (coords, x) => {
   }
   return bestD <= PART_COL_TOL ? best : -1;
 };
-var rowText = (anchors, r) => anchors.filter((a) => a.r === r).map((a) => a.cell.text.replace(/\s+/g, "")).join("|");
+var rowText = (anchors, r) => anchors.filter((a) => a.r === r).map((a) => flowText(a.cell.text).replace(/\s+/g, "")).join("|");
 var hasText = (t) => t.replace(/\|/g, "") !== "";
 function joinSplitParts(prev, pcx, curr, ccx, dx = 0, prevBottom, lex) {
   if (pcx.length !== prev.cols + 1 || ccx.length !== curr.cols + 1) return null;
@@ -944,8 +945,8 @@ function continuesAcross(u, d, x1, x2) {
 var OUTLINE_HEAD = /^[□■]\s*\S/;
 var OUTLINE_ITEM = /^[ㅇ○◦\-‐–·․‧※*]\s*\S/;
 function outlineContinues(u, d) {
-  const U = u.text.split("\n").map((l) => l.trim()).filter(Boolean);
-  const D = d.text.split("\n").map((l) => l.trim()).filter(Boolean);
+  const U = flowText(u.text).split("\n").map((l) => l.trim()).filter(Boolean);
+  const D = flowText(d.text).split("\n").map((l) => l.trim()).filter(Boolean);
   return U.length >= 2 && D.length >= 1 && U.some((l) => OUTLINE_HEAD.test(l)) && OUTLINE_ITEM.test(D[0]);
 }
 function lineEnded(c, x1, x2, next) {
@@ -957,7 +958,7 @@ function lineEnded(c, x1, x2, next) {
   return x2 - l.r - (l.l - x1) >= Math.max(LABEL_ROOM, firstWordUnits(next?.text ?? "")) * h;
 }
 function firstWordUnits(text) {
-  const word = text.trim().replace(/^([(\[「『<〈])\s+/, "$1").split(/\s+/)[0] ?? "";
+  const word = flowText(text).trim().replace(/^([(\[「『<〈])\s+/, "$1").split(/\s+/)[0] ?? "";
   let units = 0;
   for (const ch of word) units += /[가-힣\u3400-\u9fff]/.test(ch) ? 1 : 0.55;
   return units;
@@ -977,7 +978,7 @@ var batchim = (ch) => {
   return k >= 0 && k < 11172 && k % 28 !== 0;
 };
 function clauseContinues(u, d) {
-  const tail = u.text.trim().split(/\s+/).pop() ?? "", head = d.text.trim();
+  const tail = flowText(u.text).trim().split(/\s+/).pop() ?? "", head = flowText(d.text).trim();
   if (!/^[가-힣]{2,}$/.test(tail) || !/^[가-힣]/.test(head) || startsNewItem(u.text, head)) return false;
   const last = tail[tail.length - 1], before = tail[tail.length - 2];
   return last === "\uC744" && batchim(before) || last === "\uB97C" && !batchim(before) || CLAUSE_OPEN_ENDING.test(tail);
@@ -990,6 +991,7 @@ function wordOverflows(u, d, x1, x2) {
 }
 var KO_ORDER = "\uAC00\uB098\uB2E4\uB77C\uB9C8\uBC14\uC0AC\uC544\uC790\uCC28\uCE74\uD0C0\uD30C\uD558";
 function itemMark(line) {
+  line = flowText(line).trim();
   let m = /^(\d{1,2})([.)])(?!\d)/.exec(line);
   if (m) return { style: "1" + m[2], n: +m[1] };
   if (m = /^\((\d{1,2})\)/.exec(line)) return { style: "(1)", n: +m[1] };
@@ -1041,7 +1043,7 @@ function mergeSplitRow(table, owner, first, colXs, textOnly = false, cut = NO_CU
   const contentPairs = pairs.filter(([u, d]) => hasContent(u.cell) && hasContent(d.cell));
   const carriedSplit = !textOnly && carried && !newCell && (contentPairs.length <= 1 || contentPairs.every(([, d]) => d.rs > 1) || !cut.ruled && contentPairs.some(([u, d]) => wordOverflows(cell(u), cell(d), colXs[u.c], colXs[u.c + u.cs])));
   if (!carriedSplit) {
-    const norm = (c) => c.text.replace(/\s+/g, "");
+    const norm = (c) => flowText(c.text).replace(/\s+/g, "");
     if (pairs.some(([u, d]) => hasContent(d.cell) && norm(cell(d)) !== norm(cell(u)) && lineEnded(cell(u), colXs[u.c], colXs[u.c + u.cs], cell(d)))) return false;
     const flows = ([u, d]) => continuesAcross(cell(u), cell(d), colXs[u.c], colXs[u.c + u.cs]) || clauseContinues(cell(u), cell(d));
     if (!(cut.open && !newCell)) {
@@ -1138,6 +1140,20 @@ function mergeCrossPageTables(blocks, pageHeights, lex) {
       blocks.splice(j, 1);
       continue;
     }
+    if (px && cx && !CLIP_TABLES.has(prev.table) && !CLIP_TABLES.has(curr.table) && currCells.length < curr.table.cells.length && curr.table.cells[0].filter((c) => c.text.trim()).length >= 2 && prev.table.cells.at(-1).concat(currCells[0]).every((c) => c.rowSpan === 1 && c.colSpan === 1 && CELL_LINES.has(c))) {
+      const dx = px[0] - cx[0];
+      const part = joinSplitParts(prev.table, px, curr.table, cx.map((x) => x + dx), dx, void 0, lex);
+      if (part?.split && part.table.rows === prev.table.rows + currCells.length - 1) {
+        const last = prev.table.rows - 1;
+        for (let c = 0; c < part.table.cols; c++) {
+          const U = CELL_LINES.get(prev.table.cells[last][c]), D = CELL_LINES.get(currCells[0][c]);
+          CELL_LINES.set(part.table.cells[last][c], U.concat(D.map((l) => ({ ...l, l: l.l + dx, r: l.r + dx }))));
+        }
+        TABLE_COLXS.set(part.table, part.colXs);
+        placeJoined(blocks, i, j, part.table);
+        continue;
+      }
+    }
     const merged = {
       rows: prev.table.rows + currCells.length,
       cols: prev.table.cols,
@@ -1222,7 +1238,7 @@ function prevLacksLeftCols(prev, px, xs, pageHeights) {
   return xs[0] < px[0] && xs.slice(1, -1).some((x) => Math.abs(x - px[0]) <= CONTINUATION_COL_TOL);
 }
 function headingRow(row) {
-  const heads = row.map((c) => c.text.replace(/\s+/g, "")).filter(Boolean);
+  const heads = row.map((c) => flowText(c.text).replace(/\s+/g, "")).filter(Boolean);
   if (heads.length === 1) return heads[0].length <= 8 || /^[①-⑳❶-❿]/.test(heads[0]);
   return heads.length === 2 && /^(?:\d{1,2}|[ⅠⅡⅢⅣⅤⅥⅦⅧⅨⅩ]+|[①-⑳❶-❿])$/.test(heads[0]);
 }
@@ -1233,7 +1249,7 @@ function looksContinued(prev, curr, pageHeights) {
   const firstRow = curr.table.cells[0] ?? [];
   const firstText = firstRow.find((c) => c.text.trim())?.text ?? "";
   if (!ANNEX_HEAD_RE.test(firstText)) return true;
-  const filled = firstRow.map((c) => c.text.replace(/\s+/g, "")).filter(Boolean);
+  const filled = firstRow.map((c) => flowText(c.text).replace(/\s+/g, "")).filter(Boolean);
   return filled.length >= 2 && filled.every((t) => t === filled[0]);
 }
 function pushedLead(prev, curr) {
@@ -1262,7 +1278,7 @@ function tailOf(b) {
 function firstRowSig(t) {
   const x = TABLE_COLXS.get(t);
   if (!x || x.length !== t.cols + 1) return null;
-  return anchorsOf(t).filter((a) => a.r === 0).map((a) => ({ x1: x[a.c], x2: x[a.c + a.cs], t: a.cell.text.replace(/\s+/g, "") }));
+  return anchorsOf(t).filter((a) => a.r === 0).map((a) => ({ x1: x[a.c], x2: x[a.c + a.cs], t: flowText(a.cell.text).replace(/\s+/g, "") }));
 }
 function restartsTable(blocks, i, curr, pageHeights) {
   const cs = firstRowSig(curr);
@@ -1287,7 +1303,7 @@ function restartsTitledForm(head, curr) {
   if (head.rows < 2 || curr.rows < 2 || head.cols !== curr.cols) return false;
   const full = (t) => t.cells[0][0]?.colSpan === t.cols;
   if (!full(head) || !full(curr)) return false;
-  const norm = (c) => (c?.text ?? "").replace(/\s+/g, "");
+  const norm = (c) => flowText(c?.text ?? "").replace(/\s+/g, "");
   const ht = norm(head.cells[0][0]), ct = norm(curr.cells[0][0]);
   if (!ht || !ct || ht === ct) return false;
   const titleLike = (t) => t.length <= TITLE_ROW_MAX_CHARS && !/^[ㅇ○◦•·\-–※*□■▪☞]/.test(t);
@@ -1323,7 +1339,7 @@ function chainHead(blocks, i, pageHeights) {
 }
 function rowTextsEqual(a, b) {
   if (a.length !== b.length) return false;
-  const norm = (t) => t.replace(/\s+/g, "");
+  const norm = (t) => flowText(t).replace(/\s+/g, "");
   for (let i = 0; i < a.length; i++) {
     if (norm(a[i].text) !== norm(b[i].text)) return false;
   }
@@ -2688,7 +2704,7 @@ function mergeLinkRuns(markdown) {
 
 // src/pdf/formula-ocr.ts
 async function applyFormulaOcr(buffer, blocks, pageFilter, effectivePageCount, warnings, _onProgress) {
-  const formulaMod = await import("./formula-SHBXRXOJ.js");
+  const formulaMod = await import("./formula-QZZDW5GW.js");
   const { FormulaPipeline, ensureFormulaModels } = formulaMod;
   await ensureFormulaModels((p) => {
     if (p.phase === "download" && p.total) {
@@ -3071,7 +3087,7 @@ async function parsePdfDocument(buffer, options) {
       }
       if (targets.size > 0) {
         try {
-          const { runPdfOcr } = await import("./pdf-ocr-462ZFSHX.js");
+          const { runPdfOcr } = await import("./pdf-ocr-XC33L43D.js");
           const mode = typeof options?.ocr === "function" ? options.ocr : "builtin";
           const regionPages = new Map([...ocrRegions].filter(([p]) => options?.ocr !== "force" && !isImageBased && !pageQuality.find((q) => q.page === p)?.needsOcr));
           const ocrPageBlocks = await runPdfOcr(ocrBuffer, targets, mode, warnings, options?.onProgress, options?.tables !== false, vectorPageOps, regionPages);
@@ -3196,7 +3212,8 @@ async function parsePdfDocument(buffer, options) {
       return b.bbox.y > h * 0.1 && b.bbox.y + b.bbox.height < h * 0.9;
     });
     joinLatinCellWraps(outBlocks);
-    if (options?.layoutTables !== "keep" && await isHancomPdf(doc)) {
+    if (options?.layoutTables !== "keep") {
+      const hancom = await isHancomPdf(doc);
       const shown = /* @__PURE__ */ new Map();
       const markContent = (bs) => {
         for (const b of bs ?? []) if (b.table) for (const row of b.table.cells) for (const c of row) {
@@ -3206,6 +3223,7 @@ async function parsePdfDocument(buffer, options) {
       };
       const visual = (bs) => bs.flatMap((b) => {
         if (b.type !== "table") return [b];
+        if (!hancom && b.table && CLIP_TABLES.has(b.table)) return [b];
         let v = shown.get(b);
         if (!v) {
           markContent([b]);
@@ -3301,4 +3319,4 @@ export {
   parsePdfDocument,
   removeHeaderFooterBlocks
 };
-//# sourceMappingURL=parser-CTRB7PKD.js.map
+//# sourceMappingURL=parser-VH3F4EQU.js.map