Aislo/tools/est-test/xls-read.cjs

// 원본 엑셀형 수량산출서(TFRMMULEXCEL) 본문 — MulMainDat.Editor 에 든 옛 Excel(.xls, OLE2 복합문서 + BIFF8) 읽기.
// 읽기 전용: 시트 이름과 셀 값(문자·숫자·불·오류), 수식 셀은 저장된 계산값과 수식 여부만 꺼낸다. 서식·그림은 읽지 않는다.
const zlib = require("node:zlib");

function cfb(buffer) {
  if (buffer.readUInt32LE(0) !== 0xe011cfd0 || buffer.readUInt32LE(4) !== 0xe11ab1a1)
    throw Error("엑셀형 수량산출서 형식(OLE2)이 아닙니다.");
  const sectorSize = 1 << buffer.readUInt16LE(30),
    miniSize = 1 << buffer.readUInt16LE(32),
    fatCount = buffer.readUInt32LE(44),
    dirStart = buffer.readInt32LE(48),
    miniCutoff = buffer.readUInt32LE(56),
    miniFatStart = buffer.readInt32LE(60),
    difatStart = buffer.readInt32LE(68);
  const sector = (n) => buffer.subarray(512 + n * sectorSize, 512 + (n + 1) * sectorSize);
  const difat = [];
  for (let i = 0; i < 109; i++) difat.push(buffer.readInt32LE(76 + i * 4));
  for (let next = difatStart, guard = 0; next >= 0 && guard < 10000; guard++) {
    const s = sector(next);
    for (let i = 0; i < sectorSize / 4 - 1; i++) difat.push(s.readInt32LE(i * 4));
    next = s.readInt32LE(sectorSize - 4);
  }
  const fat = [];
  for (const n of difat.slice(0, fatCount)) {
    if (n < 0) continue;
    const s = sector(n);
    for (let i = 0; i < sectorSize / 4; i++) fat.push(s.readInt32LE(i * 4));
  }
  const chain = (start, table) => {
    const out = [];
    for (let n = start, guard = 0; n >= 0 && guard < table.length + 1; guard++) {
      out.push(n);
      n = table[n];
    }
    return out;
  };
  const read = (start) => Buffer.concat(chain(start, fat).map(sector));
  const dir = read(dirStart),
    entries = [];
  for (let at = 0; at + 128 <= dir.length; at += 128) {
    const nameLength = dir.readUInt16LE(at + 64);
    entries.push({
      name: dir.toString("utf16le", at, at + Math.max(0, nameLength - 2)),
      type: dir[at + 66],
      start: dir.readInt32LE(at + 116),
      size: dir.readUInt32LE(at + 120),
    });
  }
  const root = entries[0],
    miniStream = root ? read(root.start) : Buffer.alloc(0),
    miniFat = [];
  if (miniFatStart >= 0) {
    const m = read(miniFatStart);
    for (let i = 0; i < m.length / 4; i++) miniFat.push(m.readInt32LE(i * 4));
  }
  return (name) => {
    const entry = entries.find((e) => e.type === 2 && e.name.toLowerCase() === name.toLowerCase());
    if (!entry) return null;
    if (entry.size < miniCutoff)
      return Buffer.concat(
        chain(entry.start, miniFat).map((n) =>
          miniStream.subarray(n * miniSize, (n + 1) * miniSize),
        ),
      ).subarray(0, entry.size);
    return read(entry.start).subarray(0, entry.size);
  };
}

function rk(value) {
  const raw = value & 1 ? 100 : 1;
  const number =
    value & 2
      ? value >> 2
      : (() => {
          const b = Buffer.alloc(8);
          b.writeInt32LE(value & 0xfffffffc, 4);
          return b.readDoubleLE(0);
        })();
  return number / raw;
}

// BIFF8 유니코드 문자열(길이·옵션·리치/확장 정보) — CONTINUE 경계는 SST 에서만 따로 처리한다.
function unicode(data, at, lengthBytes = 2) {
  const length = lengthBytes === 2 ? data.readUInt16LE(at) : data[at];
  let pos = at + lengthBytes;
  const flags = data[pos++],
    wide = flags & 1;
  let runs = 0,
    ext = 0;
  if (flags & 8) ((runs = data.readUInt16LE(pos)), (pos += 2));
  if (flags & 4) ((ext = data.readUInt32LE(pos)), (pos += 4));
  const bytes = wide ? length * 2 : length;
  const text = wide
    ? data.toString("utf16le", pos, pos + bytes)
    : data.toString("latin1", pos, pos + bytes);
  return { text, end: pos + bytes + runs * 4 + ext };
}

function records(stream) {
  const out = [];
  for (let at = 0; at + 4 <= stream.length;) {
    const type = stream.readUInt16LE(at),
      size = stream.readUInt16LE(at + 2);
    out.push({ type, data: stream.subarray(at + 4, at + 4 + size) });
    at += 4 + size;
  }
  return out;
}

function sst(list, index) {
  // SST 와 뒤따르는 CONTINUE 를 문자열 단위로 이어 읽는다.
  const parts = [list[index].data];
  for (let i = index + 1; i < list.length && list[i].type === 0x3c; i++) parts.push(list[i].data);
  const strings = [],
    count = parts[0].readUInt32LE(4);
  let part = 0,
    data = parts[0],
    pos = 8;
  const next = () => {
    part++;
    data = parts[part];
    pos = 0;
  };
  while (strings.length < count && data) {
    if (pos >= data.length) {
      next();
      continue;
    }
    const length = data.readUInt16LE(pos);
    let flags = data[pos + 2],
      p = pos + 3,
      runs = 0,
      ext = 0;
    if (flags & 8) ((runs = data.readUInt16LE(p)), (p += 2));
    if (flags & 4) ((ext = data.readUInt32LE(p)), (p += 4));
    let text = "",
      left = length;
    pos = p;
    while (left > 0) {
      if (pos >= data.length) {
        next();
        if (!data) break;
        flags = data[pos++];
      }
      const wide = flags & 1,
        room = Math.floor((data.length - pos) / (wide ? 2 : 1)),
        take = Math.min(left, room);
      text += wide
        ? data.toString("utf16le", pos, pos + take * 2)
        : data.toString("latin1", pos, pos + take);
      pos += take * (wide ? 2 : 1);
      left -= take;
    }
    let skip = runs * 4 + ext;
    while (skip > 0 && data) {
      const step = Math.min(skip, data.length - pos);
      pos += step;
      skip -= step;
      if (skip > 0) next();
    }
    strings.push(text);
  }
  return strings;
}

function readXls(input) {
  let buffer = Buffer.from(input);
  if (buffer[0] === 0x78) buffer = zlib.inflateSync(buffer);
  const open = cfb(buffer),
    stream = open("Workbook") || open("Book");
  if (!stream) throw Error("엑셀형 수량산출서에 Workbook 이 없습니다.");
  const list = records(stream);
  let strings = [];
  const sheets = [];
  list.forEach((record, i) => {
    if (record.type === 0xfc) strings = sst(list, i);
    if (record.type === 0x85)
      sheets.push({
        offset: record.data.readUInt32LE(0),
        name: unicode(record.data, 6, 1).text,
        cells: {},
      });
  });
  const byOffset = new Map();
  let at = 0;
  for (const record of list) {
    byOffset.set(at, record);
    at += 4 + record.data.length;
  }
  const offsets = [...byOffset.keys()];
  for (const sheet of sheets) {
    let index = offsets.indexOf(sheet.offset);
    if (index < 0) continue;
    let pendingFormula = null;
    for (index++; index < offsets.length; index++) {
      const { type, data } = byOffset.get(offsets[index]);
      if (type === 0x0a) break;
      const put = (row, col, value, formula = false) =>
        (sheet.cells[`${row},${col}`] = formula ? { value, formula: true } : { value });
      if (type === 0xfd)
        put(data.readUInt16LE(0), data.readUInt16LE(2), strings[data.readUInt32LE(6)] ?? "");
      else if (type === 0x203)
        put(data.readUInt16LE(0), data.readUInt16LE(2), data.readDoubleLE(6));
      else if (type === 0x27e)
        put(data.readUInt16LE(0), data.readUInt16LE(2), rk(data.readInt32LE(6)));
      else if (type === 0xbd) {
        const row = data.readUInt16LE(0),
          first = data.readUInt16LE(2),
          last = data.readUInt16LE(data.length - 2);
        for (let c = first; c <= last; c++)
          put(row, c, rk(data.readInt32LE(4 + (c - first) * 6 + 2)));
      } else if (type === 0x204)
        put(data.readUInt16LE(0), data.readUInt16LE(2), unicode(data, 6).text);
      else if (type === 0x205) {
        const v = data[6];
        put(data.readUInt16LE(0), data.readUInt16LE(2), data[7] ? `#ERR${v}` : Boolean(v));
      } else if (type === 0x06) {
        const row = data.readUInt16LE(0),
          col = data.readUInt16LE(2);
        if (data.readUInt16LE(12) === 0xffff) {
          const kind = data[6];
          if (kind === 0) pendingFormula = { row, col };
          else
            put(row, col, kind === 1 ? Boolean(data[8]) : kind === 2 ? `#ERR${data[8]}` : "", true);
        } else put(row, col, data.readDoubleLE(6), true);
      } else if (type === 0x207 && pendingFormula) {
        put(pendingFormula.row, pendingFormula.col, unicode(data, 0).text, true);
        pendingFormula = null;
      }
    }
  }
  return sheets.map(({ name, cells }) => ({ name, cells }));
}

module.exports = { readXls, cfb };