/**
 * Vietnamese number/date/text normalization for DISPLAY only.
 *
 * IMPORTANT: This only modifies the SUBTITLES and CHAT display text.
 * It does NOT affect the TTS audio output path — audio comes directly
 * from the S2S backend as PCM16, unaffected by this text processing.
 *
 * "Mất âm thanh TTS" root cause analysis:
 * The problem was NOT this file — the Vietnamese formatting examples in
 * app.js instructions were too verbose, causing session.update WebSocket
 * messages to be rejected or delayed. Fixed by simplifying instructions.
 *
 * Normalize cautiously: only run on complete (non-partial) transcripts,
 * never on every delta event. Heavy regex on 50+ deltas/second lags the
 * event loop and can starve the audio buffer processing.
 */

// ── Vietnamese number words ───────────────────────────────────────────────
const DIGITS = [
  "không", "một", "hai", "ba", "bốn", "năm", "sáu", "bảy", "tám", "chín",
];

function numberToVietnamese(n) {
  if (n === 0) return "không";
  if (n < 0) return "âm " + numberToVietnamese(-n);

  const units = ["", "nghìn", "triệu", "tỷ"];
  const groups = [];
  let temp = n;
  while (temp > 0) {
    groups.push(temp % 1000);
    temp = Math.floor(temp / 1000);
  }
  if (groups.length === 0) groups.push(0);

  const readGroup = (g) => {
    if (g === 0) return "";
    const h = Math.floor(g / 100);
    const r = g % 100;
    let s = "";
    if (h > 0) s += DIGITS[h] + " trăm ";
    else if (groups.length > 1) s += "không trăm ";
    if (r === 0) return s.trim();
    if (r < 10) {
      s += (h > 0 && r === 5) ? "lẻ năm" : "lẻ " + DIGITS[r];
    } else if (r < 20) {
      s += "mười" + (r === 10 ? "" : (r === 15 ? " lăm" : " " + DIGITS[r % 10]));
    } else {
      const t = Math.floor(r / 10);
      const o = r % 10;
      s += ["", "", "hai mươi", "ba mươi", "bốn mươi", "năm mươi",
            "sáu mươi", "bảy mươi", "tám mươi", "chín mươi"][t];
      if (o === 1) s += " mốt";
      else if (o === 5) s += " lăm";
      else if (o > 0) s += " " + DIGITS[o];
    }
    return s.replace(/\s+/g, " ").trim();
  };

  let result = "";
  for (let i = groups.length - 1; i >= 0; i--) {
    const g = groups[i];
    if (g === 0) continue;
    result += readGroup(g);
    if (i > 0) result += " " + units[i] + " ";
  }
  return result.replace(/\s+/g, " ").trim();
}

/**
 * Normalize Vietnamese text for display: convert numbers/dates to words.
 * Only handles patterns that are likely to occur in Vietnamese assistant
 * responses. Runs ONLY on final transcripts, not partial deltas.
 */
export function normalizeVietnameseText(text) {
  if (!text || typeof text !== "string") return text;

  let result = text;

  // 1. Dates: dd/mm/yyyy
  result = result.replace(
    /\b(\d{1,2})\/(\d{1,2})\/(\d{4})\b/g,
    (_, d, m, y) => `ngày ${parseInt(d)} tháng ${parseInt(m)} năm ${numberToVietnamese(parseInt(y))}`,
  );

  // 2. Percentages
  result = result.replace(/\b(\d+(?:[.,]\d+)?)%\b/g, (_, num) => {
    const val = parseFloat(num.replace(",", "."));
    return Number.isInteger(val)
      ? `${numberToVietnamese(val)} phần trăm`
      : `${numberToVietnamese(Math.floor(val))} phẩy ${numberToVietnamese(Math.round((val % 1) * 100))} phần trăm`;
  });

  // 3. Currency VND: 1.000.000₫ or 500k
  result = result.replace(/\b(\d{1,3}(?:\.\d{3})+)\s*(₫|đ|vnd)\b/gi, (_, num) =>
    `${numberToVietnamese(parseInt(num.replace(/\./g, "")))} đồng`);
  result = result.replace(/\b(\d+)k\b/gi, (_, num) =>
    `${numberToVietnamese(parseInt(num) * 1000)} đồng`);

  // 4. Decimals (not years — 1900-2100 are already handled by the model)
  result = result.replace(/\b(\d+)\.(\d{1,3})\b/g, (_, intPart, decPart) => {
    if (intPart.length > 4) return `${intPart}.${decPart}`; // not a decimal
    const decVal = parseInt(decPart);
    return `${numberToVietnamese(parseInt(intPart))} phẩy ${numberToVietnamese(decVal)}`;
  });

  // 5. Large standalone numbers (3+ digits = years, prices, quantities)
  // Only replace if surrounded by Vietnamese text context
  result = result.replace(/\b(\d{3,})\b/g, (match) => {
    const n = parseInt(match);
    if (n >= 1900 && n <= 2100) return `năm ${numberToVietnamese(n)}`;
    return numberToVietnamese(n);
  });

  // Clean spaces
  result = result.replace(/\s+/g, " ").trim();
  return result;
}

/**
 * Check if text contains Vietnamese characters.
 */
export function containsVietnamese(text) {
  return /[àáạảãâầấậẩẫăằắặẳẵèéẹẻẽêềếệểễìíịỉĩòóọỏõôồốộổỗơờớợởỡùúụủũưừứựửữỳýỵỷỹđ]/i.test(text);
}

/**
 * Smart normalizer: only run on Vietnamese text, only on final transcripts.
 * Pass `partial=false` from the transcript event to skip partial deltas.
 */
export function smartNormalize(text, partial = false) {
  if (!text || partial) return text;
  if (containsVietnamese(text)) {
    return normalizeVietnameseText(text);
  }
  return text;
}