format.js

// Hearlark: transcript formatting (TXT, SRT, VTT, JSON). Pure functions, no DOM, so they run in Node tests too.
// © 2026 CyberMax.

/** Seconds → "HH:MM:SS,mmm" (SRT) or "HH:MM:SS.mmm" (VTT). */
export function stamp(seconds, sep = ',') {
  const ms = Math.max(0, Math.round((Number(seconds) || 0) * 1000));
  const h = Math.floor(ms / 3600000);
  const m = Math.floor((ms % 3600000) / 60000);
  const s = Math.floor((ms % 60000) / 1000);
  const r = ms % 1000;
  const p = (n, w = 2) => String(n).padStart(w, '0');
  return `${p(h)}:${p(m)}:${p(s)}${sep}${p(r, 3)}`;
}

/** Short clock for the on-page list: "m:ss" or "h:mm:ss". */
export function clock(seconds) {
  const t = Math.max(0, Math.floor(Number(seconds) || 0));
  const h = Math.floor(t / 3600);
  const m = Math.floor((t % 3600) / 60);
  const s = t % 60;
  return h ? `${h}:${String(m).padStart(2, '0')}:${String(s).padStart(2, '0')}` : `${m}:${String(s).padStart(2, '0')}`;
}

/**
 * Normalises the ASR pipeline output ({ text, chunks: [{ timestamp: [start, end], text }] }) into clean segments.
 * Fixes what Whisper output commonly has: null end times, overlapping or backwards times, empty or
 * whitespace-only chunks and runs of the same line repeated (a known Whisper hallucination on silence).
 */
export function toSegments(result, duration = null) {
  const chunks = Array.isArray(result?.chunks) ? result.chunks : [];
  const out = [];
  for (const c of chunks) {
    const text = String(c?.text ?? '').replace(/\s+/g, ' ').trim();
    if (!text) continue;
    let [start, end] = Array.isArray(c.timestamp) ? c.timestamp : [null, null];
    start = Number.isFinite(start) ? start : (out.length ? out[out.length - 1].end : 0);
    if (out.length && start < out[out.length - 1].end) start = out[out.length - 1].end;
    if (!Number.isFinite(end) || end <= start) end = Number.isFinite(duration) && duration > start ? Math.min(duration, start + 10) : start + 2;
    const prev = out[out.length - 1];
    if (prev && prev.text === text && start - prev.end < 1) { prev.end = end; continue; }
    out.push({ start: +start.toFixed(3), end: +end.toFixed(3), text });
  }
  if (!out.length && String(result?.text ?? '').trim()) {
    out.push({ start: 0, end: Number.isFinite(duration) && duration > 0 ? +duration.toFixed(3) : 0, text: String(result.text).replace(/\s+/g, ' ').trim() });
  }
  return out;
}

export function toTxt(segments, { timestamps = false } = {}) {
  if (!timestamps) return segments.map((s) => s.text).join(' ').replace(/\s+([,.!?;:])/g, '$1').trim() + '\n';
  return segments.map((s) => `[${clock(s.start)}] ${s.text}`).join('\n') + '\n';
}

export function toSrt(segments) {
  return segments.map((s, i) => `${i + 1}\n${stamp(s.start)} --> ${stamp(s.end)}\n${s.text}\n`).join('\n');
}

export function toVtt(segments) {
  return 'WEBVTT\n\n' + segments.map((s) => `${stamp(s.start, '.')} --> ${stamp(s.end, '.')}\n${s.text}\n`).join('\n');
}

export function toJson(segments, meta = {}) {
  return JSON.stringify({ ...meta, segments, text: toTxt(segments).trim() }, null, 2) + '\n';
}

/** Approximate number of 30 s windows the pipeline will decode (20 s step with a 5 s stride each side). */
export function expectedChunks(duration) {
  const d = Number(duration) || 0;
  return d <= 30 ? 1 : 1 + Math.ceil((d - 30) / 20);
}

/** Mixes decoded channels down to one Float32Array. */
export function toMono(channels) {
  if (!channels.length) return new Float32Array(0);
  if (channels.length === 1) return channels[0];
  const out = new Float32Array(channels[0].length);
  for (const ch of channels) for (let i = 0; i < out.length; i++) out[i] += ch[i] / channels.length;
  return out;
}

/** "My Talk (final).mp3" → "My Talk (final)" for download names. */
export function baseName(name) {
  const s = String(name || 'transcript').replace(/\.[^.\/\\]+$/, '').replace(/[\\/:*?"<>|]+/g, '-').trim();
  return s || 'transcript';
}