kordoc 4.16.1 → 4.16.2

package/dist/index.d.cts

@@ -194,8 +194,14 @@ interface ParseOptions {
     onProgress?: (current: number, total: number) => void;
     /** PDF 머리글/바닥글 자동 제거 */
     removeHeaderFooter?: boolean;
+    /**
+     * 위·아래첨자를 인라인 HTML `<sup>`·`<sub>` 로 표기 — 평문으로 펴면 "10⁴ m²" 가 "104 m2", "x_i" 가 "xi" 로 값이 바뀐다.
+     * 기본: HWPX·HWP·DOCX 켬(글자 모양에 적힌 첨자), PDF 끔(글자 크기·기준선으로 추정 — 논문·수식 문서는 true 권장.
+     * 공개 벤치 ODL 정답이 첨자를 평문으로 적어 기본값을 두지 않는다). false 면 모든 형식에서 평문. OCR 로 읽은 글은 늘 평문
+     */
+    scriptTags?: boolean;
     /** 평문 Markdown — 그림 자리 표시·링크 URL·밑줄(`<u>`)·굵게(`**`) 표기를 빼고 글만 (제목·목록·표 구조는 유지).
-     *  이미지 바이트를 따로 저장하지 않는 색인·RAG 용. 기본 false. `blocks` IR 은 그대로 */
+     *  첨자 `<sup>`·`<sub>` 는 값이 남게 `10^4`·`H_2O` 로 편다. 이미지 바이트를 따로 저장하지 않는 색인·RAG 용. 기본 false. `blocks` IR 은 그대로 */
     plain?: boolean;
     /** 모든 표를 HTML 로 — 파이프 표도 HTML 표로 옮기고, 표마다 태그를 한 줄씩 들여써 낸다(BeautifulSoup prettify 모양, 첫 행 `<th>`).
      *  HTML 표만 다루는 소비자·채점기용. 기본 false(병합·중첩 없는 표는 GFM 파이프 표) */

package/dist/index.d.ts

@@ -194,8 +194,14 @@ interface ParseOptions {
     onProgress?: (current: number, total: number) => void;
     /** PDF 머리글/바닥글 자동 제거 */
     removeHeaderFooter?: boolean;
+    /**
+     * 위·아래첨자를 인라인 HTML `<sup>`·`<sub>` 로 표기 — 평문으로 펴면 "10⁴ m²" 가 "104 m2", "x_i" 가 "xi" 로 값이 바뀐다.
+     * 기본: HWPX·HWP·DOCX 켬(글자 모양에 적힌 첨자), PDF 끔(글자 크기·기준선으로 추정 — 논문·수식 문서는 true 권장.
+     * 공개 벤치 ODL 정답이 첨자를 평문으로 적어 기본값을 두지 않는다). false 면 모든 형식에서 평문. OCR 로 읽은 글은 늘 평문
+     */
+    scriptTags?: boolean;
     /** 평문 Markdown — 그림 자리 표시·링크 URL·밑줄(`<u>`)·굵게(`**`) 표기를 빼고 글만 (제목·목록·표 구조는 유지).
-     *  이미지 바이트를 따로 저장하지 않는 색인·RAG 용. 기본 false. `blocks` IR 은 그대로 */
+     *  첨자 `<sup>`·`<sub>` 는 값이 남게 `10^4`·`H_2O` 로 편다. 이미지 바이트를 따로 저장하지 않는 색인·RAG 용. 기본 false. `blocks` IR 은 그대로 */
     plain?: boolean;
     /** 모든 표를 HTML 로 — 파이프 표도 HTML 표로 옮기고, 표마다 태그를 한 줄씩 들여써 낸다(BeautifulSoup prettify 모양, 첫 행 `<th>`).
      *  HTML 표만 다루는 소비자·채점기용. 기본 false(병합·중첩 없는 표는 GFM 파이프 표) */

package/dist/index.js

@@ -16,15 +16,19 @@ import {
   mapPuaText,
   markNonLayoutTable,
   noteSuffix
-} from "./chunk-ANZTLTKU.js";
+} from "./chunk-7KNGSQU7.js";
 import {
   parsePageRange
 } from "./chunk-V4BH5SRQ.js";
 import {
   HEADING_RATIO_H1,
   HEADING_RATIO_H2,
-  HEADING_RATIO_H3
-} from "./chunk-KBAQLAEL.js";
+  HEADING_RATIO_H3,
+  plainScripts,
+  stripScriptTags,
+  tidyScriptTags,
+  wrapScript
+} from "./chunk-3IB7LBHC.js";
 import {
   KordocError,
   OPTIONAL_DEP_INSTALL_HINT,
@@ -41,17 +45,17 @@ import {
   toArrayBuffer,
   unescapeHtml,
   unzipLimitBytes
-} from "./chunk-NUPXPKIQ.js";
+} from "./chunk-MLIS253L.js";
 
 // src/plain-markdown.ts
 function toPlainMarkdown(md2) {
-  return md2.replace(/!\[[^\]\n]*\]\([^)\n]*\)/g, "").replace(/<img\b[^>]*>/gi, "").replace(/(?<!!)\[([^\]\n]*)\]\((?:https?|mailto|ftp):[^)\s]*\)/g, "$1").replace(/<\/?u>/g, "").replace(/(?<!\\)\*\*(?=\S)([^\n]*?\S)(?<!\\)\*\*/g, "$1").replace(/(<br>)+(?=<\/t[dh]>)/g, "").replace(/[ \t]+$/gm, "").replace(/\n{3,}/g, "\n\n").trim();
+  return plainScripts(md2).replace(/!\[[^\]\n]*\]\([^)\n]*\)/g, "").replace(/<img\b[^>]*>/gi, "").replace(/(?<!!)\[([^\]\n]*)\]\((?:https?|mailto|ftp):[^)\s]*\)/g, "$1").replace(/<\/?u>/g, "").replace(/(?<!\\)\*\*(?=\S)([^\n]*?\S)(?<!\\)\*\*/g, "$1").replace(/(<br>)+(?=<\/t[dh]>)/g, "").replace(/[ \t]+$/gm, "").replace(/\n{3,}/g, "\n\n").trim();
 }
 
 // src/html-tables.ts
 var TABLE_TAG = /(<\/?(?:table|tr|th|td)\b[^>]*>)/i;
 function pipeCellToHtml(cell2) {
-  return cell2.split(/(<br>|<\/?u>|<img\b[^>]*>)/).map((part, i) => i % 2 ? part : part.replace(/\\([\\`*_{}[\]()#+\-.!|~<>$])/g, "$1").replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;")).join("");
+  return cell2.split(/(<br>|<\/?(?:u|sup|sub)>|<img\b[^>]*>)/).map((part, i) => i % 2 ? part : part.replace(/\\([\\`*_{}[\]()#+\-.!|~<>$])/g, "$1").replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;")).join("");
 }
 function splitPipeRow(line) {
   const t = line.trim().replace(/^\|/, "").replace(/(?<!\\)\|$/, "");
@@ -770,7 +774,9 @@ function parseCharProperties(doc, map) {
         else if (localTag === "strikeout") {
           const shape = k.getAttribute("shape") || "";
           if (isRealStrikeShape(shape)) prop.strike = true;
-        } else if (localTag === "underline") {
+        } else if (localTag === "supscript") prop.script = "sup";
+        else if (localTag === "subscript") prop.script = "sub";
+        else if (localTag === "underline") {
           if (k.getAttribute("type") === "BOTTOM") prop.underline = true;
         }
       }
@@ -1848,6 +1854,7 @@ function extractRunSpans(para2, ctx, mode, requireMixed) {
     const span2 = { text };
     if (mode !== "foreign" && prId === KORDOC_CHAR_CODE) span2.code = true;
     else {
+      if (cp?.script) span2.text = wrapScript(text, cp.script);
       if (cp?.bold) span2.bold = true;
       if (cp?.italic) span2.italic = true;
       if (cp?.strike) span2.strike = true;
@@ -1867,6 +1874,7 @@ function extractRunSpans(para2, ctx, mode, requireMixed) {
     }
   }
   if (requireMixed && !merged.some((s) => s.strike || s.underline) && !merged.some((s) => !(s.bold || s.italic || s.code))) return null;
+  for (const s of merged) s.text = tidyScriptTags(s.text);
   return merged;
 }
 function gongmunDepthFromIndent(para2, left, ctx) {
@@ -2516,6 +2524,16 @@ function extractHyperlinkHref(fieldBegin) {
 function isInDeletedRange(ctx) {
   return (ctx?.shared.track.deleteDepth ?? 0) > 0;
 }
+function runHasOnlyText(run) {
+  const kids = run.childNodes;
+  for (let i = 0; i < (kids?.length ?? 0); i++) {
+    const k = kids[i];
+    if (k.nodeType !== 1) continue;
+    const t = (k.tagName || k.localName || "").replace(/^[^:]+:/, "");
+    if (t !== "t" && t !== "linesegarray") return false;
+  }
+  return true;
+}
 function extractParagraphInfo(para2, styleMap, ctx) {
   let text = "";
   let href;
@@ -2772,11 +2790,19 @@ function extractParagraphInfo(para2, styleMap, ctx) {
           }
           break;
         }
-        // run 요소에서 charPrIDRef 추출
-        case "r": {
+        // run 요소 — hp:r 은 charPrIDRef 를 문단 대표 스타일로. 첨자 글자 모양이면 이 run 이 더한 글을 <sup>·<sub> 로
+        // (개체·각주 등 컨트롤을 품은 run·필드 경계가 걸친 run 은 글 위치가 섞여 감싸지 않는다)
+        case "r":
+        case "run": {
           const runCharPr = child.getAttribute("charPrIDRef");
-          if (runCharPr && !charPrId) charPrId = runCharPr;
+          if (tag === "r" && runCharPr && !charPrId) charPrId = runCharPr;
+          const script = runCharPr ? styleMap?.charProperties.get(runCharPr)?.script : void 0;
+          const start = text.length, nLinks = linkRanges.length, nOpen = openFields.length;
           walk(child, depth + 1);
+          if (script && linkRanges.length === nLinks && openFields.length === nOpen && runHasOnlyText(child)) {
+            const added = text.slice(start);
+            if (added && !/[\x1E\x1F\n$]/.test(added)) text = text.slice(0, start) + wrapScript(added, script);
+          }
           break;
         }
         default:
@@ -2800,6 +2826,7 @@ function extractParagraphInfo(para2, styleMap, ctx) {
   }
   const leaderIdx = text.indexOf("");
   if (leaderIdx >= 0) text = text.substring(0, leaderIdx);
+  text = tidyScriptTags(text);
   const cleanParaText = (raw) => {
     let t = raw.replace(/[ \t]+/g, " ").trim();
     if (/^그림입니다\.?\s*원본\s*그림의\s*(이름|크기)/.test(t)) t = "";
@@ -3043,6 +3070,14 @@ function isExtendedOnlyCtrlChar(ch) {
 function appendParaText(state, data, resolveControl) {
   let result = "";
   let i = 0;
+  const wbase = state.wpos ?? 0;
+  const setScript = (want) => {
+    const cur = state.script ?? null;
+    if (cur === want) return;
+    if (cur) result += `</${cur}>`;
+    if (want) result += `<${want}>`;
+    state.script = want;
+  };
   const base = state.text.length;
   const resolveAt = (byteOffset, extended) => {
     const ctrlId = data.readUInt32LE(byteOffset);
@@ -3053,6 +3088,7 @@ function appendParaText(state, data, resolveControl) {
   };
   while (i + 1 < data.length) {
     const ch = data.readUInt16LE(i);
+    if (state.scriptAt) setScript(ch >= 32 ? state.scriptAt(wbase + i / 2) : null);
     i += 2;
     switch (ch) {
       // ── char 타입 (2바이트만, 확장 데이터 없음) ──
@@ -3124,6 +3160,8 @@ function appendParaText(state, data, resolveControl) {
         break;
     }
   }
+  if (state.scriptAt) setScript(null);
+  state.wpos = wbase + Math.floor(data.length / 2);
   state.text += result;
 }
 function extractEquationText(data) {
@@ -5210,6 +5248,7 @@ function parseParagraph(records, start, end, ctx) {
   const paraShapeId = header.data.length >= 10 ? header.data.readUInt16LE(8) : -1;
   const textRecords = [];
   const charShapeIds = [];
+  const charShapeRuns = [];
   const ctrls = [];
   let i = start + 1;
   while (i < end) {
@@ -5228,6 +5267,7 @@ function parseParagraph(records, start, end, ctx) {
     } else if (rec.tagId === TAG_CHAR_SHAPE && rec.level === baseLevel + 1 && rec.data.length >= 8) {
       for (let offset = 0; offset + 7 < rec.data.length; offset += 8) {
         charShapeIds.push(rec.data.readUInt32LE(offset + 4));
+        charShapeRuns.push([rec.data.readUInt32LE(offset), rec.data.readUInt32LE(offset + 4)]);
       }
     }
     i++;
@@ -5241,6 +5281,7 @@ function parseParagraph(records, start, end, ctx) {
   const state = createParaTextState();
   state.leaderMark = true;
   state.dollarMark = true;
+  state.scriptAt = scriptLookup(charShapeRuns, ctx.docInfo);
   const resolver = (idx, id) => {
     let ctrl = idx >= 0 && idx < ctrls.length ? ctrls[idx] : void 0;
     if (!ctrl || ctrl.idRaw !== id && ctrl.id !== id) {
@@ -5282,6 +5323,7 @@ function parseParagraph(records, start, end, ctx) {
   const leaderAt = text.indexOf(LEADER_TAB_MARK);
   if (leaderAt >= 0) text = text.slice(0, leaderAt);
   if (text.includes(LITERAL_DOLLAR_MARK)) text = text.replaceAll(LITERAL_DOLLAR_MARK, "\\$");
+  text = tidyScriptTags(text);
   let headingLevel = 0;
   let headMarker = null;
   const ps = ctx.docInfo && paraShapeId >= 0 && paraShapeId < ctx.docInfo.paraShapes.length ? ctx.docInfo.paraShapes[paraShapeId] : null;
@@ -5647,6 +5689,20 @@ function pictureToImageBlock(data, ctx) {
   const storageId = item && item.storageId > 0 ? item.storageId : binDataId;
   return { type: "image", text: String(storageId), pageNumber: ctx.page };
 }
+function scriptLookup(runs, docInfo) {
+  if (!docInfo || !runs.length) return void 0;
+  const kindOf = (id) => {
+    const f = docInfo.charShapes[id]?.attrFlags ?? 0;
+    return f & 1 << 15 ? "sup" : f & 1 << 16 ? "sub" : null;
+  };
+  if (!runs.some(([, id]) => kindOf(id))) return void 0;
+  let k = -1;
+  return (pos) => {
+    if (k >= 0 && runs[k][0] > pos) k = -1;
+    while (k + 1 < runs.length && runs[k + 1][0] <= pos) k++;
+    return k >= 0 ? kindOf(runs[k][1]) : null;
+  };
+}
 function resolveCharStyle(charShapeIds, docInfo) {
   if (charShapeIds.length === 0 || docInfo.charShapes.length === 0) return void 0;
   const freq = /* @__PURE__ */ new Map();
@@ -20734,6 +20790,9 @@ function extractRun(r) {
   if (rPrEls.length > 0) {
     bold = getChildElements(rPrEls[0], "b").length > 0;
     italic = getChildElements(rPrEls[0], "i").length > 0;
+    const va = getChildElements(rPrEls[0], "vertAlign")[0];
+    const v = va ? getAttr(va, "val") : null;
+    if (v === "superscript" || v === "subscript") text = wrapScript(text, v === "superscript" ? "sup" : "sub");
   }
   return { text, bold, italic };
 }
@@ -20831,7 +20890,7 @@ function collectInline(p, footnotes, rels) {
     if (isDisplayMath(om)) parts.push(" $$" + latex + "$$ ");
     else parts.push(" $" + latex + "$ ");
   }
-  const text = parts.join("").replace(/[ \t]{2,}/g, " ").trim();
+  const text = tidyScriptTags(parts.join("")).replace(/[ \t]{2,}/g, " ").trim();
   return { text, bold, italic, footnoteText };
 }
 var COUNTERS = /* @__PURE__ */ new WeakMap();
@@ -21704,7 +21763,7 @@ var ValueCursor = class {
   }
 };
 function normalizeLabel(label) {
-  return label.trim().replace(/[:：\s()（）·]/g, "");
+  return label.trim().replace(/<\/?su[bp]>/g, "").replace(/[:：\s()（）·]/g, "");
 }
 function findMatchingKey(cellLabel, values) {
   if (values.has(cellLabel)) return cellLabel;
@@ -26780,6 +26839,7 @@ function parseInlineMarkdown(text) {
   text = text.replace(/!\[([^\]]*)\]\([^)]*\)/g, "$1");
   text = text.replace(/~~([^~]+)~~/g, "$1");
   text = text.replace(/<u>([^<]*)<\/u>/g, "$1");
+  text = text.replace(/<(sup|sub)>([^<]*)<\/\1>/g, "$2");
   let spans;
   if (inlineDoc) {
     spans = [];
@@ -40540,10 +40600,14 @@ async function dispatch(format2, buffer, opts) {
       return { success: false, fileType: "unknown", error: "\uC9C0\uC6D0\uD558\uC9C0 \uC54A\uB294 \uD30C\uC77C \uD615\uC2DD\uC785\uB2C8\uB2E4.", code: "UNSUPPORTED_FORMAT" };
   }
 }
+function scriptsOff(r, off) {
+  return off ? stripScriptTags(r) : r;
+}
 async function parseImage(buffer, options) {
   try {
-    const { parseImageDocument } = await import("./image-ocr-GTDG4SAJ.js");
+    const { parseImageDocument } = await import("./image-ocr-5BUJQ4TK.js");
     const { blocks, warnings } = await parseImageDocument(buffer, options);
+    stripScriptTags({ blocks });
     return {
       success: true,
       fileType: "image",
@@ -40567,7 +40631,7 @@ async function parseHwp3(buffer, options) {
 async function parseHwpx(buffer, options) {
   try {
     const { markdown, blocks, metadata, outline, warnings, images } = await parseHwpxDocument(buffer, options);
-    return { success: true, fileType: "hwpx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount };
+    return scriptsOff({ success: true, fileType: "hwpx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
   } catch (err) {
     return { success: false, fileType: "hwpx", error: sanitizeError(err), code: classifyError(err) };
   }
@@ -40593,7 +40657,7 @@ async function parseHwp(buffer, options) {
       } catch {
       }
     }
-    return { success: true, fileType: "hwp", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount };
+    return scriptsOff({ success: true, fileType: "hwp", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
   } catch (err) {
     return { success: false, fileType: "hwp", error: sanitizeError(err), code: classifyError(err) };
   }
@@ -40601,7 +40665,7 @@ async function parseHwp(buffer, options) {
 async function parsePdf(buffer, options) {
   let parsePdfDocument;
   try {
-    const mod = await import("./parser-UEROAPYZ.js");
+    const mod = await import("./parser-ZQM3664Z.js");
     parsePdfDocument = mod.parsePdfDocument;
   } catch {
     return {
@@ -40613,7 +40677,7 @@ async function parsePdf(buffer, options) {
   }
   try {
     const { markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary, images, pages } = await parsePdfDocument(buffer, options);
-    return { success: true, fileType: "pdf", markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary, images, pages, pageCount: metadata?.pageCount };
+    return scriptsOff({ success: true, fileType: "pdf", markdown, blocks, metadata, outline, warnings, isImageBased, pageQuality, qualitySummary, images, pages, pageCount: metadata?.pageCount }, options?.scriptTags !== true);
   } catch (err) {
     const isImageBased = err instanceof Error && "isImageBased" in err ? true : void 0;
     return { success: false, fileType: "pdf", error: sanitizeError(err), code: classifyError(err), isImageBased };
@@ -40638,7 +40702,7 @@ async function parseXls(buffer, options) {
 async function parseDocx(buffer, options) {
   try {
     const { markdown, blocks, metadata, outline, warnings, images } = await parseDocxDocument(buffer, options);
-    return { success: true, fileType: "docx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount };
+    return scriptsOff({ success: true, fileType: "docx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0, pageCount: metadata?.pageCount }, options?.scriptTags === false);
   } catch (err) {
     return { success: false, fileType: "docx", error: sanitizeError(err), code: classifyError(err) };
   }