NovaSonic/backend/lib/metadata.js

const https = require('https');
const path  = require('path');
const fs    = require('fs');

function sanitize(s) {
  return (s || 'Unknown').replace(/[\\/:*?"<>|]/g, '_').trim().slice(0, 100);
}

function navidromeArtist(name) {
  if (!name) return name;
  return name
    .replace(/\s+feat\.?\s+/gi, ';')
    .replace(/\s+ft\.?\s+/gi, ';')
    .replace(/\s+featuring\s+/gi, ';')
    .replace(/\s+&\s+/g, ';')

    .replace(/,\s*/g, ';');
}

// Download hi-res cover artwork (1000x1000) to cover.jpg, return path or null
async function downloadArtwork(url, dir) {
  if (!url) return null;
  const dest  = path.join(dir, 'cover.jpg');
  const hiRes = url.replace(/\d+x\d+bb\./, '1000x1000bb.');
  return new Promise(resolve => {
    const tryUrl = (u, fallback) => {
      const file = fs.createWriteStream(dest);
      https.get(u, res => {
        if (res.statusCode !== 200) {
          res.resume();
          file.close();
          try { fs.unlinkSync(dest); } catch {}
          if (fallback) tryUrl(fallback, null);
          else resolve(null);
          return;
        }
        res.pipe(file);
        file.on('finish', () => resolve(dest));
        file.on('error',  () => resolve(null));
      }).on('error', () => {
        if (fallback) tryUrl(fallback, null);
        else resolve(null);
      });
    };
    tryUrl(hiRes, url !== hiRes ? url : null);
  });
}

const AUDIO_EXTS = ['mp3', 'flac', 'opus', 'm4a', 'ogg', 'wav'];

// Strips release-type suffixes like "- Single", "- EP", "2024" from end of title/stem.
const _RELEASE_SUFFIX_RE = /\s*[-–]\s*(?:(?:official\s+)?(?:music\s+)?(?:single|ep|lp|album|ost|soundtrack|remaster(?:ed)?|deluxe(?:\s+edition)?|bonus\s+tracks?|b-?sides?|version|edit)\b.*|(?:20|19)\d{2})\s*$/i;

// Version markers that make a track a DISTINCT recording — brackets containing these must NOT be
// stripped during duplicate detection, or a remix/live/acoustic/instrumental collapses onto the
// base track and gets falsely skipped (e.g. "Therapy (Remix)" was skipped because "Therapy" existed).
const _VERSION_KW = /\b(?:remix|live|acoustic|instrumental|version|vip|flip|bootleg|re-?work|reprise|unplugged|demo|mashup|cover|sped|slowed|nightcore|extended|edit|mix|dub)\b/i;

// Robust duplicate detection: search case-insensitively across all album folders
// under the artist folder matching firstArtist, compare normalized title ignoring
// release suffixes and track prefixes. Handles DJ OKAWARI vs DJ Okawari, and
// "recognize - Single.mp3" vs "recognize.mp3" (drift between runs).
function findExistingTrack(musicRoot, firstArtist, cleanTitle) {
  const norm = s => (s || '')
    // Akzente abtrennen und verwerfen: dieselbe Kuenstlerin kommt je nach Quelle als
    // "Cécile Corbel" oder "Cecile Corbel" — ohne das galt sie als zwei Personen und derselbe
    // Song wurde ein zweites Mal geladen. Betrifft nur den VERGLEICH, nicht Tags oder Dateinamen.
    .normalize('NFD').replace(/\p{M}+/gu, '')
    .toLowerCase()
    // Strip bracketed noise (feat. X, Official Video, Prod. by …) but KEEP version markers
    // (Remix/Live/Acoustic/…) so distinct recordings aren't collapsed onto the base track.
    .replace(/[\[(（]([^\]）)]*)[\]）)]/g, (m, inner) => _VERSION_KW.test(inner) ? m : '')
    // Filesystem-sanitized chars (/ \ : * ? " < > |) become "_" on disk — map those AND "_"
    // to a space so a title tag "Kwik Flip | YTPMV" matches the file "Kwik Flip _ YTPMV".
    .replace(/[/\\:*?"<>|_]/g, ' ')
    // Auch die typografischen Apostrophe: Anbieter schreiben "Arrietty’s Song" (U+2019), eine
    // spaetere Anfrage kommt mit geradem "'" — ohne das galten beide als verschiedene Titel.
    .replace(/[,.!'\u2018\u2019\u02bc\u00b4`]/g, '')
    .replace(/\s+/g, ' ')
    .trim();
  const normArtist = norm(firstArtist);
  const normTitle  = norm(cleanTitle);
  // Accept both the raw cleanTitle and version without release suffix (e.g. "recognize" matches "recognize - Single")
  const titleVariants = new Set([normTitle, norm(cleanTitle.replace(_RELEASE_SUFFIX_RE, ''))]);

  try {
    for (const ad of fs.readdirSync(musicRoot, { withFileTypes: true })) {
      if (!ad.isDirectory() || norm(ad.name) !== normArtist) continue;
      const artistPath = path.join(musicRoot, ad.name);
      try {
        for (const abd of fs.readdirSync(artistPath, { withFileTypes: true })) {
          if (!abd.isDirectory()) continue;
          const albumPath = path.join(artistPath, abd.name);
          try {
            for (const f of fs.readdirSync(albumPath)) {
              const ext = path.extname(f).slice(1).toLowerCase();
              if (!AUDIO_EXTS.includes(ext)) continue;
              // Strip extension + leading track prefix ("01 - ", "01-15 – " etc.)
              const stem = f.slice(0, -(ext.length + 1)).replace(/^\d[\d-]*\s*[-–—]\s*/, '');
              // Candidate titles from the filename. The old code only used the part AFTER the first
              // " - " (assuming "Artist - Title"), which mis-parses titles that themselves contain
              // " - " (e.g. "Kingdom Hearts 2 - Dearly Beloved - Main Theme") or filenames with a
              // garbage/absent artist prefix ("01-01 - …") → the duplicate went undetected and got
              // re-downloaded. Now match if ANY reading equals the query title:
              //   1) the FULL stem (title with internal dashes, or no clean artist prefix)
              //   2) the part after the first " - " (normal "Artist - Title")
              const di = stem.indexOf(' - ');
              const candidates = di >= 0 ? [stem, stem.slice(di + 3)] : [stem];
              for (const cand of candidates) {
                if (titleVariants.has(norm(cand.replace(_RELEASE_SUFFIX_RE, '')))) return path.join(albumPath, f);
              }
            }
          } catch {}
        }
      } catch {}
    }
  } catch {}
  return null;
}

// Fetch plain + synced lyrics from LRCLIB
async function fetchLyrics(artist, title, album) {
  try {
    const p = new URLSearchParams({ artist_name: artist, track_name: title });
    if (album) p.set('album_name', album);
    const r = await fetch(`https://lrclib.net/api/get?${p}`, {
      headers: { 'User-Agent': 'NovaSonic/1.0' },
      signal: AbortSignal.timeout(8000),
    });
    if (!r.ok) return null;
    const d = await r.json();
    return { plain: d.plainLyrics || null, synced: d.syncedLyrics || null };
  } catch { return null; }
}

// Track/album/artist metadata now comes from Deezer (backend/lib/deezer.js) — Apple's
// iTunes /search returns 403 from server IPs. See that module for lookupTrackMeta,
// fetchAlbumTracks, searchAlbumTracks and fetchArtistAlbums.

module.exports = {
  sanitize, navidromeArtist,
  downloadArtwork, findExistingTrack,
};