// Video platforms plus the social-post platforms (see common/lib/posts.ts).
// Widening this allowlist is purely additive: it cannot invalidate existing
// cached entries, so no transcriptStore.ts DB_VERSION bump is needed. Posts get
// their own IDB store rather than sharing the transcript cache.
export type Platform =
  | "youtube"
  | "rumble"
  | "odysee"
  | "twitch"
  | "kick"
  // archive.org items (lib/archiveOrgId.ts): a whole item, or one file inside
  // a multi-file item.
  | "archiveorg"
  // BitChute videos (bitchute.com/video/<id>/) and channels.
  | "bitchute"
  | "twitter"
  | "bluesky"
  | "xenforo";

export const PLATFORM_VALUES: ReadonlyArray<Platform> = [
  "youtube",
  "rumble",
  "odysee",
  "twitch",
  "kick",
  "archiveorg",
  "bitchute",
  "twitter",
  "bluesky",
  "xenforo",
];

// The subset that carries social posts rather than videos. A channel on one of
// these is a `sourceKind: "social"` channel — the video scan skips it and the
// posts pipeline picks it up.
export const SOCIAL_PLATFORM_VALUES: ReadonlyArray<Platform> = [
  "twitter",
  "bluesky",
  "xenforo",
];

export function isSocialPlatform(platform: Platform | null | undefined): boolean {
  return platform === "twitter" || platform === "bluesky" || platform === "xenforo";
}

// Host → platform. Lives in `detectPlatform.mjs` (plain JS so umtool's `.mjs`
// scripts can import the same copy); every TS caller imports it from here.
import { detectPlatform, sameXenforoForum } from "./detectPlatform.mjs";
export { detectPlatform, sameXenforoForum };

// Best-effort canonical webpage URL for a video given its platform and
// canonical id. Used both as a metadata fallback in summarize() and to
// re-create a download URL when a video has no metadata.info.json. Note: for
// Odysee the canonical id is only the claim hash (the channel-name prefix is
// dropped by extractVideoId), so a reconstructed Odysee URL may not resolve.
export function defaultWebpageUrl(platform: Platform, id: string): string {
  if (platform === "rumble") return `https://rumble.com/${id}`;
  if (platform === "odysee") return `https://odysee.com/${id}`;
  if (platform === "twitch") return `https://www.twitch.tv/videos/${id}`;
  // Kick's canonical id is the VOD UUID; /video/<uuid> resolves to the VOD.
  if (platform === "kick") return `https://kick.com/video/${id}`;
  // A whole item's id IS its identifier. A file's id is a slug + hash the file
  // path cannot be recovered from, so this is the right page only for a whole
  // item; every archived record carries its own `webpage_url`.
  if (platform === "archiveorg") {
    const file = /^(.+?)__.*-[0-9a-f]{8}$/.exec(id);
    return `https://archive.org/details/${file ? file[1] : id}`;
  }
  // BitChute's canonical id is its video id, which is also yt-dlp's.
  if (platform === "bitchute") return `https://www.bitchute.com/video/${id}/`;
  // Social posts: /i/status/<id> resolves without knowing the handle. Bluesky
  // has no handle-free permalink, so this is only a last-resort fallback —
  // every archived post carries its own canonical `url` (see postPermalink).
  if (platform === "twitter") return `https://x.com/i/status/${id}`;
  if (platform === "bluesky") return `https://bsky.app/profile/${id}`;
  // A forum post's URL needs its forum's host, which an id alone does not
  // carry; every archived forum post has its own `url`. A thread URL passed as
  // the id is returned as it is.
  if (platform === "xenforo") return /^https?:\/\//.test(id) ? id : "";
  return `https://www.youtube.com/watch?v=${id}`;
}

export function platformQueueKey(
  platform: Platform | null | undefined,
): string {
  return `platform:${platform ?? "unknown"}`;
}

// Best-effort registrable domain for an arbitrary URL, used to route
// unrecognized sources to their own job queue. Uses a dependency-free
// "last two labels" heuristic (clips.twitch.tv -> twitch.tv,
// player.vimeo.com -> vimeo.com). Multi-part TLDs coarsen (example.co.uk
// -> co.uk), which is harmless for a queue key — at worst two unrelated
// sources share a serial queue. Returns null when the URL can't be parsed.
export function registrableDomain(url: string | undefined | null): string | null {
  if (!url) return null;
  try {
    const host = new URL(url).hostname.toLowerCase();
    if (!host) return null;
    const labels = host.split(".").filter(Boolean);
    if (labels.length <= 2) return labels.join(".") || null;
    return labels.slice(-2).join(".");
  } catch {
    return null;
  }
}

// Queue key for a channel/video URL. Known platforms get their canonical
// `platform:<name>` queue; unrecognized hosts fall back to a per-domain queue
// (e.g. `platform:vimeo.com`) instead of all colliding in `platform:unknown`.
export function queueKeyForUrl(url: string | undefined | null): string {
  const detected = detectPlatform(url);
  if (detected) return platformQueueKey(detected);
  const host = registrableDomain(url);
  return host ? `platform:${host}` : platformQueueKey(null);
}

// System-wide queue for resource-bound local jobs (whisper, ffmpeg).
// Channels share this queue so two heavy local jobs never run in parallel.
export const TRANSCRIPTION_QUEUE = "transcription";