// The canonical video id derived from a video URL — the name every video's
// data/<id>/ dir carries, on every platform. Deliberately a leaf module whose
// one import (archiveOrgId.ts) is itself a leaf with no imports: it lives here
// rather than in ytdlp/runYtdlp.ts (its original home, which still re-exports it) so that low-level stores like
// controller/rosterStore.ts can canonicalize a URL without pulling in execa,
// the settings loader and the whole download pipeline.
//
// Canonical is NOT the same as yt-dlp's native extractor id: the two coincide
// on YouTube and diverge everywhere else. See archiveIdForUrl in runYtdlp.ts
// for the native-id resolution that reads metadata.info.json.

import { isArchiveOrgItemHost, parseArchiveOrgUrl, archiveOrgVideoId } from "./archiveOrgId";
import { isJwPlayerHost, isWaybackHost, jwPlayerMediaId, parseWaybackUrl } from "./wayback";

export function extractVideoId(url: string): string | null {
  return extractVideoIdAt(url, 0);
}

function extractVideoIdAt(url: string, depth: number): string | null {
  try {
    const u = new URL(url);
    const host = u.hostname.toLowerCase();
    if (isWaybackHost(host)) {
      // A Wayback capture is named by what it is a capture OF
      // (lib/wayback.ts): an archived YouTube page by its YouTube id, a JW
      // Player file by its media id. Its own path's last segment is the
      // original's (`watch`, `<id>-<rendition>.mp4`) — a name two captures
      // share. A capture of a capture is not unwrapped twice.
      const ref = depth === 0 ? parseWaybackUrl(url) : null;
      return ref ? extractVideoIdAt(ref.originalUrl, depth + 1) : null;
    }
    if (isJwPlayerHost(host)) {
      // One media id across every rendition and host; a JW URL that names
      // none falls through to the last segment below.
      const jw = jwPlayerMediaId(u);
      if (jw) return jw;
    }
    if (isArchiveOrgItemHost(host)) {
      // A whole item → its identifier; one file inside an item → a stable
      // `<identifier>__<slug>-<hash>` (lib/archiveOrgId.ts). A URL that names
      // no item (a search page, a collection listing) has no id.
      const ref = parseArchiveOrgUrl(url);
      return ref ? archiveOrgVideoId(ref) : null;
    }
    if (host.endsWith("youtube.com") || host === "youtu.be") {
      const v = u.searchParams.get("v");
      if (v) return v;
      const seg = u.pathname.split("/").filter(Boolean).pop();
      return seg ?? null;
    }
    if (host.endsWith("rumble.com")) {
      const seg = u.pathname.split("/").filter(Boolean).pop();
      // Rumble paths often look like /v123abc-some-title.html
      if (seg) {
        const trimmed = seg.replace(/\.html?$/i, "");
        const dashIdx = trimmed.indexOf("-");
        return dashIdx > 0 ? trimmed.slice(0, dashIdx) : trimmed;
      }
      return null;
    }
    if (host.endsWith("odysee.com")) {
      const decoded = decodeURIComponent(u.pathname);
      const lastColon = decoded.lastIndexOf(":");
      if (lastColon > 0) return decoded.slice(lastColon + 1);
      return null;
    }
    if (host === "bitchute.com" || host.endsWith(".bitchute.com")) {
      // A video page or its embed: /video/<id>/, /embed/<id>/ (www. or old.).
      // A channel, playlist or profile page names no video.
      const segs = u.pathname.split("/").filter(Boolean);
      return (segs[0] === "video" || segs[0] === "embed") && segs[1] ? segs[1] : null;
    }
    if (host.endsWith("twitch.tv")) {
      // VODs: /videos/<id> or legacy /<channel>/v/<id>; clips:
      // clips.twitch.tv/<slug> or /<channel>/clip/<slug>. Grab the segment
      // after the marker; otherwise fall back to the last path segment.
      const segs = u.pathname.split("/").filter(Boolean);
      for (const marker of ["videos", "v", "clip", "clips"]) {
        const idx = segs.indexOf(marker);
        if (idx >= 0 && segs[idx + 1]) return segs[idx + 1];
      }
      return segs.pop() ?? null;
    }
    if (host.endsWith("kick.com")) {
      // VODs: /<channel>/videos/<uuid> or /video/<uuid>; clips:
      // /<channel>/clips/<slug> or /clip/<slug>. The UUID after the marker is
      // yt-dlp's native Kick id; fall back to the last segment otherwise.
      const segs = u.pathname.split("/").filter(Boolean);
      for (const marker of ["videos", "video", "clips", "clip"]) {
        const idx = segs.indexOf(marker);
        if (idx >= 0 && segs[idx + 1]) return segs[idx + 1];
      }
      return segs.pop() ?? null;
    }
    const seg = u.pathname.split("/").filter(Boolean).pop();
    return seg ?? null;
  } catch {
    return null;
  }
}