const iconv = require('iconv-lite');

// ============================================================================
// framing.js 负责“分帧”而不是“解析”
// - 输入：连续文本/字节流（可能粘包、半包、多帧）
// - 输出：完整 XML 帧数组 + 未完成尾部余量
// 解析工作交给 index.js 的 parseAndPrintData。
// ============================================================================

const ROOT_SPECS = [
  { open: '<ORU_R31', close: '</ORU_R31>' },
  { open: '<ORF_R04', close: '</ORF_R04>' },
];

function findNextRoot(text, fromIndex) {
  // 从当前位置向后找下一个根节点起点（ORU_R31 / ORF_R04），取最早出现者。
  let best = null;
  for (const spec of ROOT_SPECS) {
    const idx = text.indexOf(spec.open, fromIndex);
    if (idx === -1) continue;
    if (!best || idx < best.idx) best = { idx, spec };
  }
  return best;
}

function maybeIncludeXmlDecl(text, frameStart, searchFloor) {
  // 若根节点前存在同一帧的 XML 声明 `<?xml ...?>`，则把声明一起纳入帧起点。
  const idxXml = text.lastIndexOf('<?xml', frameStart);
  if (idxXml < searchFloor) return frameStart;

  const endDecl = text.indexOf('?>', idxXml);
  if (endDecl === -1) return frameStart;
  if (endDecl >= frameStart) return frameStart;

  return idxXml;
}

function extractFramesFromText(text) {
  // 文本分帧：按根节点和闭合标签切片，找不到闭合则保留余量等待下次补齐。
  const frames = [];
  let cursor = 0;

  while (cursor < text.length) {
    const next = findNextRoot(text, cursor);
    if (!next) break;

    let start = next.idx;
    start = maybeIncludeXmlDecl(text, start, cursor);

    const endClose = text.indexOf(next.spec.close, next.idx);
    if (endClose === -1) {
      // Incomplete frame: keep from `start` as remainder.
      return { frames, remainderText: text.slice(start) };
    }

    const end = endClose + next.spec.close.length;
    frames.push(text.slice(start, end));
    cursor = end;
  }

  return { frames, remainderText: text.slice(cursor) };
}

function extractFramesFromUtf16leBuffer(buffer) {
  // 字节分帧：先按 UTF-16LE 解码成文本分帧，再把 remainder 映射回字节余量。
  const frames = [];

  // UTF-16LE requires even bytes; keep a tail byte if needed.
  let oddTail = Buffer.alloc(0);
  let work = buffer;
  if (work.length % 2 === 1) {
    oddTail = work.slice(work.length - 1);
    work = work.slice(0, work.length - 1);
  }

  const text = iconv.decode(work, 'utf16le');
  const { frames: found, remainderText } = extractFramesFromText(text);
  frames.push(...found);

  // Convert remainderText length (code units) back to bytes.
  // This assumes the decoded text maps 1:1 to UTF-16 code units, which is valid here.
  const consumedChars = text.length - remainderText.length;
  const consumedBytes = consumedChars * 2;

  const remainderBuffer = Buffer.concat([work.slice(consumedBytes), oddTail]);
  return { frames, remainderBuffer };
}

function guessTextEncodingFromBuffer(buffer) {
  // 编码猜测：透析机报文常见 UTF-16LE（奇数字节位置大量 0x00）。
  if (!buffer || buffer.length < 4) return 'utf16le';

  // Heuristic: UTF-16LE XML often looks like 3C 00 4F 00 ... (lots of zeros on odd bytes)
  const sampleLen = Math.min(buffer.length, 512);
  let oddPositions = 0;
  let oddZeros = 0;
  for (let i = 1; i < sampleLen; i += 2) {
    oddPositions++;
    if (buffer[i] === 0x00) oddZeros++;
  }
  const ratio = oddPositions ? oddZeros / oddPositions : 0;
  return ratio >= 0.25 ? 'utf16le' : 'utf8';
}

module.exports = {
  extractFramesFromText,
  extractFramesFromUtf16leBuffer,
  guessTextEncodingFromBuffer,
};