goon/app/extractors/tubes/porntrex.py

"""porntrex.com — KVS engine direct stream extractor.

2026-05-22: VPS Hetzner IP znów dociera do porntrex (HTTP 200) — wcześniej blokada
trzymała `porntrexcom` na `_vps_blocked_fallback`. Patrz [[goon_porntrex_vps_unblocked]].

KVS player: detail page ma `flashvars` z `video_url` / `video_alt_url` / `video_alt_url2`
(480p / 720p / 1080p), każdy to `get_file/<srv>/<token>/<path>.mp4/` URL.

`get_file` 302 → `cdn.pcdn.cloudswitches.com/...mp4?expires=<ts>&md5=<sig>` — to
**time-bound signed URL** (nie IP-bound, NIE cookie-bound) → po rozwiązaniu jest
portable: mobile gra direct z CDN, zero VPS bandwidth.

REVISION 2026-05-31 (bug usera "porntrex wolno + brak wyboru jakości + chyba proxy"):
Wcześniejsze założenie "mobile zużyje get_file sam" było BŁĘDNE — `get_file` token jest
**cookie/session-bound**: działa tylko w tej samej sesji curl_cffi która pobrała stronę.
Osobny request mobile (ExoPlayer, bez cookies) → 410 → mobile spadał na VPS proxy
(stąd flicker = nav.replace + wolne odtwarzanie). Zweryfikowane: same-session follow
get_file → 200 (streamuje wideo); fresh session → 410. Finalny CDN url (cloudswitches,
expires+md5) jest natomiast portable (fresh session → 206).

FIX: resolvujemy 302 NA BACKENDZIE (w tej samej sesji co fetch strony) i oddajemy
**finalny CDN url** per jakość. Mobile gra direct, multi-quality picker działa, zero proxy.
Token get_file zużywamy raz tu; CDN url jest time-bound (nie single-use) → starcza na
sesję odtwarzania.
"""
from __future__ import annotations

import logging
import re
import time

from app.extractors._fetch import _DEFAULT_IMPERSONATE, _DEFAULT_UA, _HAS_CURL_CFFI, fetch_tube_html
from app.extractors._models import HosterDead, StreamSource

log = logging.getLogger(__name__)

_BASE = "https://www.porntrex.com"

# Porntrex soft-delete: usunięte wideo zwraca HTTP 200 ze stroną z komunikatem (np.
# "this video was deleted per copyright owner request") zamiast playera → extract bez
# tego zwracał [] = transient → źródło NIGDY nie oznaczone dead → user wciąż klika
# martwy link (zgłoszenie 75dbf53e). Match → raise HosterDead → resolve mark-dead.
_DEAD_RE = re.compile(
    r"this video (?:was|has been) deleted|video (?:was|has been) removed"
    r"|no longer available|video is unavailable",
    re.IGNORECASE,
)

# flashvars: `video_url: 'https://.../get_file/...mp4/'` + `video_url_text: '480p'`.
# Warianty: video_url, video_alt_url, video_alt_url2, video_alt_url3...
_URL_RE = re.compile(
    r"(video(?:_alt)?_url\d*)\s*:\s*'(https?://[^']+/get_file/[^']+)'",
    re.IGNORECASE,
)
_TEXT_RE = re.compile(
    r"(video(?:_alt)?_url\d*)_text\s*:\s*'([^']*)'",
    re.IGNORECASE,
)


def _quality_rank(label: str | None) -> int:
    """`1080p` → 1080, `720p HD` → 720. Do sortowania malejąco."""
    if not label:
        return -1
    m = re.search(r"(\d{3,4})\s*p", label, re.IGNORECASE)
    return int(m.group(1)) if m else -1


def _resolve_get_file(session, get_file_url: str, timeout: float) -> str | None:
    """Follow get_file 302 → finalny portable CDN url (w sesji która ma cookies strony).

    `?rnd=` cache-bust jak kt_player. stream=True + Range → łapiemy tylko nagłówki/finalny
    URL po redirectach, NIE pobieramy 644MB body. Zwraca None gdy resolve padł."""
    sep = "&" if "?" in get_file_url else "?"
    url = f"{get_file_url}{sep}rnd={int(time.time() * 1000)}"
    try:
        r = session.get(
            url,
            timeout=timeout,
            allow_redirects=True,
            stream=True,
            headers={"Referer": _BASE + "/", "Range": "bytes=0-1"},
        )
        final = str(r.url)
        status = r.status_code
        r.close()
    except Exception as e:
        log.info("porntrex: get_file resolve failed (%s): %s", get_file_url[:60], e)
        return None
    if status >= 400 or "/get_file/" in final:
        log.info("porntrex: get_file resolve bad status=%s final=%s", status, final[:70])
        return None
    return final


def extract(page_url: str, *, timeout: float = 60.0) -> list[StreamSource] | None:
    # Wspólna sesja: get_file token jest cookie/session-bound, więc 302 MUSI być
    # rozwiązany w tej samej sesji curl_cffi co fetch strony (patrz docstring).
    session = None
    if _HAS_CURL_CFFI:
        from curl_cffi import requests as _cf_requests
        session = _cf_requests.Session(impersonate=_DEFAULT_IMPERSONATE)
        try:
            resp = session.get(
                page_url,
                headers={"User-Agent": _DEFAULT_UA, "Accept": "text/html,application/xhtml+xml"},
                timeout=timeout,
                allow_redirects=True,
            )
            html = resp.text if resp.status_code < 400 else ""
        except Exception as e:
            log.info("porntrex: page fetch failed %s: %s", page_url, e)
            html = ""
        if not html:
            html = fetch_tube_html(page_url, timeout=timeout)
            session = None  # fetch_tube_html użył innej sesji → nie resolvuj w `session`
    else:
        html = fetch_tube_html(page_url, timeout=timeout)

    # Soft-delete: strona żyje (200) ale wideo skasowane → mark dead (nie transient).
    # Tylko gdy html niepuste (puste = fetch fail = transient, NIE dead).
    if html and _DEAD_RE.search(html):
        raise HosterDead(f"porntrex {page_url}: video deleted/removed")

    # Mapa <var_name> → quality label (np. video_alt_url → "720p HD").
    quality_by_var: dict[str, str] = {}
    for m in _TEXT_RE.finditer(html):
        quality_by_var[m.group(1).lower()] = m.group(2).strip()

    seen: set[str] = set()
    result: list[StreamSource] = []
    for m in _URL_RE.finditer(html):
        var_name = m.group(1).lower()
        url = m.group(2)
        if url in seen:
            continue
        seen.add(url)
        quality = quality_by_var.get(var_name)
        # Rozwiąż get_file → portable CDN url (w sesji ze stroną). Gdy resolve padnie,
        # oddaj get_file jako fallback (mobile spróbuje direct → ewentualnie proxy).
        final_link = url
        if session is not None:
            resolved = _resolve_get_file(session, url, timeout)
            if resolved:
                final_link = resolved
        result.append(
            StreamSource(
                link=final_link,
                type="mp4",
                quality=quality or None,
                referer=_BASE + "/",
                # Finalny CDN url (cloudswitches) jest time-bound (expires+md5), nie
                # cookie/IP-bound → mobile gra direct, zero VPS proxy bandwidth.
                raw={"mobile_direct_ok": True},
            )
        )

    if not result:
        log.info("porntrex: no KVS video_url in flashvars on %s", page_url)
        return None

    result.sort(key=lambda s: _quality_rank(s.quality), reverse=True)
    return result