torrent/src/core/tracker/HttpTracker.ts

import { type BencodeDict, type BencodeList, decode } from "../bencode/decoder.js";
import type { AnnounceParams, ITracker, ScrapeResult } from "./ITracker.js";
import type { TrackerPeer, TrackerResponse } from "./TrackerResponse.js";

/** Maximum response body size from trackers (1 MB) */
const MAX_RESPONSE_SIZE = 1024 * 1024;

/** Tracker HTTP request timeout (15 seconds) */
const REQUEST_TIMEOUT_MS = 15_000;

/**
 * HTTP/HTTPS tracker client.
 *
 * Key correctness concern: info_hash and peer_id are raw binary — they must be
 * percent-encoded byte-by-byte (not hex, not base64). We build the query string
 * manually instead of using URLSearchParams to avoid double-encoding.
 */
export class HttpTracker implements ITracker {
  readonly url: string;
  private trackerId?: string;
  private userAgent = "";

  constructor(url: string) {
    this.url = url;
  }

  async announce(params: AnnounceParams): Promise<TrackerResponse> {
    this.userAgent = params.userAgent;
    const qs = buildQueryString(params, this.trackerId);
    const sep = this.url.includes("?") ? "&" : "?";
    const fullUrl = `${this.url}${sep}${qs}`;

    const res = await fetch(fullUrl, {
      headers: {
        "User-Agent": params.userAgent,
        "Accept-Encoding": "gzip",
      },
      signal: AbortSignal.timeout(REQUEST_TIMEOUT_MS),
    });

    if (!res.ok) {
      throw new Error(`HTTP tracker returned ${res.status} for ${this.url}`);
    }

    const arrayBuf = await res.arrayBuffer();
    if (arrayBuf.byteLength > MAX_RESPONSE_SIZE) {
      throw new Error(`Tracker response too large (${arrayBuf.byteLength} bytes)`);
    }

    const body = Buffer.from(arrayBuf);
    const decoded = decode(body);
    if (!decoded || typeof decoded !== "object" || Array.isArray(decoded)) {
      throw new Error("Tracker returned invalid bencode (expected dictionary)");
    }
    const dict = decoded as BencodeDict;

    if (dict["failure reason"]) {
      const raw = dict["failure reason"];
      const reason = Buffer.isBuffer(raw) ? raw.toString("utf8") : String(raw);
      throw new Error(`Tracker failure: ${reason}`);
    }

    if (dict["tracker id"]) {
      const raw = dict["tracker id"];
      this.trackerId = Buffer.isBuffer(raw) ? raw.toString("utf8") : String(raw);
    }

    return parseResponse(dict);
  }

  async stop(params: Omit<AnnounceParams, "event">): Promise<void> {
    try {
      await this.announce({ ...params, event: "stopped" });
    } catch {
      // best-effort
    }
  }

  /**
   * Scrape the tracker for current seeder/leecher counts without announcing.
   * Derives the scrape URL by replacing the last `/announce` path segment with `/scrape`.
   * Returns null if the tracker doesn't support scrape or the request fails.
   */
  async scrape(infoHash: Buffer): Promise<ScrapeResult | null> {
    // Only supported if announce URL contains /announce in the path
    const scrapeUrl = this.url.replace(/(\/announce)([^/]*)$/, "/scrape$2");
    if (scrapeUrl === this.url) return null;

    try {
      const scrapeQSep = scrapeUrl.includes("?") ? "&" : "?";
      const fullUrl = `${scrapeUrl}${scrapeQSep}info_hash=${percentEncode(infoHash)}`;
      const res = await fetch(fullUrl, {
        headers: { "User-Agent": this.userAgent || "Mozilla/5.0" },
        signal: AbortSignal.timeout(REQUEST_TIMEOUT_MS),
      });
      if (!res.ok) return null;

      const body = Buffer.from(await res.arrayBuffer());
      const decoded = decode(body);
      if (!decoded || typeof decoded !== "object" || Array.isArray(decoded)) return null;
      const dict = decoded as BencodeDict;
      if (!dict.files || typeof dict.files !== "object") return null;

      const entries = Object.values(dict.files as BencodeDict);
      if (entries.length === 0) return null;

      const entry = entries[0];
      if (!entry || typeof entry !== "object" || Array.isArray(entry)) return null;
      const e = entry as BencodeDict;

      const complete = e.complete;
      const incomplete = e.incomplete;
      const downloaded = e.downloaded;
      return {
        seeders: typeof complete === "number" ? complete : 0,
        leechers: typeof incomplete === "number" ? incomplete : 0,
        completed: typeof downloaded === "number" ? downloaded : 0,
      };
    } catch {
      return null;
    }
  }
}

function buildQueryString(params: AnnounceParams, trackerId?: string): string {
  const parts: string[] = [
    `info_hash=${percentEncode(params.infoHash)}`,
    `peer_id=${percentEncode(params.peerId)}`,
    `port=${params.port}`,
    `uploaded=${params.uploaded}`,
    `downloaded=${params.downloaded}`,
    `left=${params.left}`,
    `compact=${params.compact !== false ? 1 : 0}`,
    `no_peer_id=1`,
  ];

  if (params.event) {
    parts.push(`event=${params.event}`);
  }

  if (params.numWant !== undefined) {
    parts.push(`numwant=${params.numWant}`);
  }

  if (params.key) {
    parts.push(`key=${encodeURIComponent(params.key)}`);
  }

  if (trackerId) {
    parts.push(`trackerid=${encodeURIComponent(trackerId)}`);
  }

  return parts.join("&");
}

/**
 * Percent-encode each byte of a Buffer individually.
 * e.g. 0x1a → %1A  (uppercase hex, safe chars left as-is... but for binary data
 * we encode everything non-alphanumeric to be safe).
 */
function percentEncode(buf: Buffer): string {
  let result = "";
  for (let i = 0; i < buf.length; i++) {
    const byte = buf[i];
    // safe chars: A-Z a-z 0-9 - _ . ~
    if (
      (byte >= 0x41 && byte <= 0x5a) || // A-Z
      (byte >= 0x61 && byte <= 0x7a) || // a-z
      (byte >= 0x30 && byte <= 0x39) || // 0-9
      byte === 0x2d || // -
      byte === 0x5f || // _
      byte === 0x2e || // .
      byte === 0x7e // ~
    ) {
      result += String.fromCharCode(byte);
    } else {
      result += `%${byte.toString(16).padStart(2, "0").toUpperCase()}`;
    }
  }
  return result;
}

function parseResponse(dict: BencodeDict): TrackerResponse {
  const rawInterval = dict.interval;
  const interval = typeof rawInterval === "number" ? rawInterval : 1800;
  const rawMinInterval = dict["min interval"];
  const minInterval = typeof rawMinInterval === "number" ? rawMinInterval : undefined;
  const rawSeeders = dict.complete;
  const seeders = typeof rawSeeders === "number" ? rawSeeders : undefined;
  const rawLeechers = dict.incomplete;
  const leechers = typeof rawLeechers === "number" ? rawLeechers : undefined;
  const rawWarning = dict["warning message"];
  const warning = Buffer.isBuffer(rawWarning) ? rawWarning.toString("utf8") : undefined;
  const rawTrackerId = dict["tracker id"];
  const trackerId = Buffer.isBuffer(rawTrackerId) ? rawTrackerId.toString("utf8") : undefined;

  const peers = parsePeers(dict.peers);

  return { interval, minInterval, seeders, leechers, peers, warning, trackerId };
}

function parsePeers(raw: unknown): TrackerPeer[] {
  if (!raw) return [];

  // Compact format: 6 bytes per peer (4 IP + 2 port)
  if (Buffer.isBuffer(raw)) {
    const peers: TrackerPeer[] = [];
    for (let i = 0; i + 6 <= raw.length; i += 6) {
      const ip = `${raw[i]}.${raw[i + 1]}.${raw[i + 2]}.${raw[i + 3]}`;
      const port = raw.readUInt16BE(i + 4);
      peers.push({ ip, port });
    }
    return peers;
  }

  // Dict format (older trackers)
  if (Array.isArray(raw)) {
    return (raw as BencodeList)
      .filter((p): p is BencodeDict => !!p && typeof p === "object" && !Array.isArray(p))
      .map((pd) => ({
        ip: Buffer.isBuffer(pd.ip) ? pd.ip.toString("ascii") : String(pd.ip ?? "0.0.0.0"),
        port: typeof pd.port === "number" ? pd.port : 0,
      }));
  }

  return [];
}