gerbeur/api/remote/youtube-playlist.ts

/**
 * Gives the global player a tracklist for a YouTube playlist, which the embed
 * plays but doesn't list. It resolves the playlist to its entries' titles and
 * lengths, for `GET /api/remote/youtube`.
 *
 * They are read from the `ytInitialData` JSON of the public playlist page,
 * whose undocumented shape changes, so both the current `lockupViewModel` rows
 * and the older `playlistVideoRenderer` ones are accepted. Only the first 100
 * entries are there; the client adds the rest, untitled, as the player reports
 * them.
 *
 * @module
 */

import { APIException } from "../lib/api-exception.ts";
import { readTextCapped } from "./fetch.ts";
import type {
  YouTubePlaylist,
  YouTubePlaylistTrack,
} from "../../shared/player.ts";

/** Titles and lengths rarely change, and a stale one only mislabels a row. */
const TTL_MS = 6 * 60 * 60 * 1000;
/** The oldest write goes first. */
const MAX_ENTRIES = 200;

const cache = new Map<string, { value: YouTubePlaylist; at: number }>();

/**
 * Admits every kind of list id (`PL…`, `OLAK5uy_…`, `RD…`) and nothing else.
 */
const LIST_ID_RE = /^[\w-]{2,64}$/;

const HEADERS = {
  "User-Agent":
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/140.0.0.0 Safari/537.36",
  "Accept": "text/html",
  "Accept-Language": "en-US,en;q=0.8",
  // Declines the consent prompt, which EU requests are otherwise redirected
  // to instead of the playlist.
  "Cookie": "SOCS=CAI",
};

/** Converts a clock like "3:55" or "1:02:03" to seconds. */
function parseClock(text: string): number | undefined {
  if (!/^\d+(?::\d{1,2}){1,2}$/.test(text)) return undefined;
  return text.split(":").reduce((acc, part) => acc * 60 + Number(part), 0);
}

/** Collects, depth first, every value stored under `key` into `out`. */
function collect(node: unknown, key: string, out: unknown[]): void {
  if (Array.isArray(node)) {
    for (const child of node) collect(child, key, out);
  } else if (node && typeof node === "object") {
    for (const [k, v] of Object.entries(node)) {
      if (k === key) out.push(v);
      else collect(v, key, out);
    }
  }
}

/** Returns the first string stored under `key`, at any depth. */
function firstString(node: unknown, key: string): string | undefined {
  const found: unknown[] = [];
  collect(node, key, found);
  return found.find((v): v is string => typeof v === "string");
}

/** Returns the value at `path` in `node`; undefined when a step is missing. */
function at(node: unknown, ...path: (string | number)[]): unknown {
  let cur = node;
  for (const key of path) {
    if (cur === null || typeof cur !== "object") return undefined;
    cur = (cur as Record<string | number, unknown>)[key];
  }
  return cur;
}

/** Reads a track from a current `lockupViewModel` row; null for a non-video. */
function fromLockup(lockup: unknown): YouTubePlaylistTrack | null {
  if (at(lockup, "contentType") !== "LOCKUP_CONTENT_TYPE_VIDEO") return null;
  const videoId = at(lockup, "contentId");
  const title = at(
    lockup,
    "metadata",
    "lockupMetadataViewModel",
    "title",
    "content",
  );
  if (typeof videoId !== "string" || typeof title !== "string") return null;
  const badges: unknown[] = [];
  collect(at(lockup, "contentImage"), "thumbnailBadgeViewModel", badges);
  const duration = badges
    .map((b) => at(b, "text"))
    .filter((t): t is string => typeof t === "string")
    .map(parseClock)
    .find((d) => d !== undefined);
  return { videoId, title, duration };
}

/** Reads a track from an older `playlistVideoRenderer` row. */
function fromLegacyRenderer(r: unknown): YouTubePlaylistTrack | null {
  const videoId = at(r, "videoId");
  const title = at(r, "title", "runs", 0, "text") ??
    at(r, "title", "simpleText");
  if (typeof videoId !== "string" || typeof title !== "string") return null;
  const secs = Number(at(r, "lengthSeconds"));
  return {
    videoId,
    title,
    duration: Number.isFinite(secs) && secs > 0 ? secs : undefined,
  };
}

/**
 * Returns the playlist `listId` as read from its page `html`; null when the
 * page has no `ytInitialData` or no track in it.
 */
export function parsePlaylistPage(
  html: string,
  listId: string,
): YouTubePlaylist | null {
  const start = html.indexOf("var ytInitialData = ");
  if (start < 0) return null;
  const end = html.indexOf(";</script>", start);
  if (end < 0) return null;

  let data: unknown;
  try {
    data = JSON.parse(html.slice(start + "var ytInitialData = ".length, end));
  } catch {
    return null;
  }

  const lockups: unknown[] = [];
  collect(at(data, "contents"), "lockupViewModel", lockups);
  const legacy: unknown[] = [];
  collect(at(data, "contents"), "playlistVideoRenderer", legacy);

  // A video listed twice keeps both rows, so that row n is the embed's entry n.
  const tracks = [
    ...lockups.map(fromLockup),
    ...legacy.map(fromLegacyRenderer),
  ].filter((t): t is YouTubePlaylistTrack => t !== null);
  if (tracks.length === 0) return null;

  const metaTitle = at(data, "metadata", "playlistMetadataRenderer", "title");
  const title = typeof metaTitle === "string"
    ? metaTitle
    : firstString(at(data, "header"), "content");
  return { listId, title, tracks };
}

/**
 * Returns the tracklist of the playlist `listId`, from the cache when it was
 * resolved in the last 6 hours.
 *
 * @throws {APIException} 400 when `listId` isn't a playlist id, 502 when
 * YouTube can't be reached or answers with an error, 404 when the playlist has
 * no tracks to read (private, deleted, or a mix).
 */
export async function resolveYouTubePlaylist(
  listId: string,
): Promise<YouTubePlaylist> {
  if (!LIST_ID_RE.test(listId)) {
    throw new APIException(400, "Invalid playlist ID");
  }

  const hit = cache.get(listId);
  if (hit && Date.now() - hit.at < TTL_MS) return hit.value;
  if (hit) cache.delete(listId);

  const controller = new AbortController();
  const timer = setTimeout(() => controller.abort(), 8000);
  let html: string;
  try {
    // youtube.com needs no address check, and fetchWithTimeout can't send the
    // consent cookie.
    const res = await fetch(
      `https://www.youtube.com/playlist?list=${encodeURIComponent(listId)}`,
      { headers: HEADERS, signal: controller.signal },
    );
    if (!res.ok) {
      await res.body?.cancel();
      throw new APIException(502, `YouTube returned ${res.status}`);
    }
    html = await readTextCapped(res);
  } catch (err) {
    if (err instanceof APIException) throw err;
    throw new APIException(502, "Could not reach YouTube");
  } finally {
    clearTimeout(timer);
  }

  const parsed = parsePlaylistPage(html, listId);
  if (!parsed) {
    // The client then shows the embed without a tracklist.
    throw new APIException(404, "No tracks on this playlist");
  }

  if (cache.size >= MAX_ENTRIES) {
    const oldest = cache.keys().next().value;
    if (oldest !== undefined) cache.delete(oldest);
  }
  cache.set(listId, { value: parsed, at: Date.now() });
  return parsed;
}