fix(movies): obsada i studio z motywu "vid", pandamovies z powrotem

Mangoporn, streamporn.li i pandamovies przeszly w pazdzierniku na nowy motyw:
obsada tylko w JSON-LD, studio, rok i gatunki w chipach bez rel="tag" (czasem
bez koncowego ukosnika). Parser po cichu przestal czytac obsade i studio, a
mangoporn dostawal jedynie tag "Porn Movies". Pandamovies dziala znowu pod
pandamovies.org, wiec wraca do ingestu. Skrypty: pelne zaciagniecie zrodla
po przerwie i uzupelnienie metadanych filmow z pazdziernika.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

app/connectors/__init__.py

@@ -35,6 +35,7 @@ def get_movie_connectors() -> list[tuple[str, type]]:
     """
     from app.connectors.dooplay import (
         MangopornConnector,
+        PandamoviesConnector,
         StreampornVipConnector,
     )
     from app.connectors.freeomovie import FreeoMovieConnector
@@ -55,17 +56,11 @@ def get_movie_connectors() -> list[tuple[str, type]]:
         ("paradisehill", ParadisehillConnector),
         ("mangoporn", MangopornConnector),
         ("streampornvip", StreampornVipConnector),
-        # pandamovies WYLACZONY 2026-08-20: serwis padl, zostal im tylko zapamietany
-        # fronton. Strona glowna oddaje 291 kB i wypisuje 56 filmow, ale KAZDA inna
-        # sciezka to 404: listing /movies/ (ten, ktory czytal konektor), /movies/page/2/
-        # linkowana przez ich wlasny szablon, /genre/* i /release-year/* z ich menu,
-        # sitemapa, a przede wszystkim same strony filmow (0 na 6 dzialajacych).
-        # To nie blokada na nasze IP: przez residential proxy wychodzi identyczne 404
-        # (te same 796 bajtow). Konektor konczyl runy jako success z zerem rekordow,
-        # co falszowalo obraz w monitoringu. Playback tego, co juz mamy, dziala dalej,
-        # bo zrodla wskazuja na hostery (doply, luluvid, seekplayer), nie na ten serwis.
-        # Klasa zostaje w dooplay.py dla backrefu origins `pandamovies:*`.
-        # ("pandamovies", PandamoviesConnector),
+        # pandamovies: wylaczony 2026-08-20 (serwis padl, kazda sciezka poza strona
+        # glowna dawala 404), wlaczony z powrotem 2026-10-16 po przeprowadzce na
+        # pandamovies.org: listing /movies/ 27 filmow na strone, 7 hosterow na film,
+        # dziala wprost z VPS bez proxy. Ten sam motyw "vid" co mangoporn i streamporn.li.
+        ("pandamovies", PandamoviesConnector),
         # freeomovie.to (bestia theme, nie dooplay), mirror: świeże filmy DVD,
         # title-trigram attach do canonical + playback z TABS. Ocena 2026-07-02.
         ("freeomovie", FreeoMovieConnector),

app/connectors/dooplay.py

@@ -191,9 +191,22 @@ _ST_DURATION_RE = re.compile(
 _ST_DATE_RE = re.compile(
     r'class="st st--date"[^>]*>.*?<time\s+datetime="(\d{4}-\d{2}-\d{2})', re.IGNORECASE | re.DOTALL
 )
-_YEAR_LOOSE_RE = re.compile(r'/(?:year|release-year)/(\d{4})/"', re.IGNORECASE)
+# pandamovies.org 2026-10: chip `/release-year/2026` bez koncowego ukosnika.
+_YEAR_LOOSE_RE = re.compile(r'/(?:year|release-year)/(\d{4})/?"', re.IGNORECASE)
+# Motyw "vid" (mangoporn, pandamovies, streamporn.li, 2026-10): studio w chipie bez
+# rel="tag", pod /studios/, /movies/studios/ albo /director/, czasem bez ukosnika.
+_STUDIO_CHIP_RE = re.compile(
+    r'<a\s+class="chip"\s+href="https?://[^/]+/(?:movies/)?(?:studios?|director)/([a-z0-9-]+)/?"[^>]*>([^<]+)</a>',
+    re.IGNORECASE,
+)
+# Ten sam motyw nie ma juz sekcji obsady z rel="tag"; obsada filmu jest w JSON-LD
+# (`"actor":[{"name":..., "url":".../actor/<slug>"}]`). Linki do aktorow w HTML bywaja
+# tez w widzetach bocznych, wiec bierzemy JSON-LD, ktory opisuje tylko ten film.
+_JSONLD_ACTORS_RE = re.compile(r'"actor"\s*:\s*(\[[^\]]*\])')
+# mangoporn ma gatunki pod /movies/genre/, pandamovies bez koncowego ukosnika, a chip
+# kategorii /genres/porn-movies/ to nie gatunek (byl jedynym tagiem filmow mangoporn).
 _GENRE_CHIP_RE = re.compile(
-    r'<a\s+class="chip"\s+href="https?://[^/]+/genres?/([a-z0-9-]+)/"[^>]*>([^<]+)</a>',
+    r'<a\s+class="chip"\s+href="https?://[^/]+/(?:movies/)?genres?/(?!porn-movies/?")([a-z0-9-]+)/?"[^>]*>([^<]+)</a>',
     re.IGNORECASE,
 )
 _DATA_SERVERS_RE = re.compile(r'data-servers="([^"]*)"', re.IGNORECASE)
@@ -348,6 +361,30 @@ class DooplayConnector(BaseMovieConnector):
         )
 
 
+def _jsonld_performers(html: str, source_name: str) -> list[RawPerformer]:
+    """Obsada z JSON-LD filmu (motyw "vid" 2026-10). Id jak dawniej: slug z adresu aktora,
+    wiec te same osoby nie dubluja sie wzgledem filmow sprzed zmiany motywu."""
+    m = _JSONLD_ACTORS_RE.search(html)
+    if not m:
+        return []
+    try:
+        people = json.loads(m.group(1))
+    except ValueError:
+        return []
+    out: list[RawPerformer] = []
+    seen: set[str] = set()
+    for p in people:
+        if not isinstance(p, dict):
+            continue
+        name = _decode_html((p.get("name") or "").strip())
+        slug = (p.get("url") or "").rstrip("/").rsplit("/", 1)[-1].lower()
+        if not name or not slug or slug in seen:
+            continue
+        seen.add(slug)
+        out.append(RawPerformer(external_id=f"{source_name}:{slug}", name=name))
+    return out
+
+
 def _parse_dooplay_detail(
     *, slug: str, html: str, source_name: str, base_url: str, page_url: str | None = None
 ) -> RawMovie | None:
@@ -381,7 +418,7 @@ def _parse_dooplay_detail(
     release_year = int(m_year.group(1)) if m_year else vid_year
 
     studio: RawStudio | None = None
-    m_studio = _STUDIO_RE.search(html)
+    m_studio = _STUDIO_RE.search(html) or _STUDIO_CHIP_RE.search(html)
     if m_studio:
         studio_slug = m_studio.group(1)
         studio_name = _decode_html(m_studio.group(2).strip())
@@ -468,6 +505,8 @@ def _parse_dooplay_detail(
         )
         for m in _PORNSTAR_RE.finditer(html)
     ]
+    if not performers:
+        performers = _jsonld_performers(html, source_name)
 
     tags = [
         RawTag(

scripts/backfill_dooplay_metadata.py

@@ -0,0 +1,88 @@
+"""Uzupelnia obsade, studio i tagi filmow z serwisow dooplay po zmianie motywu (2026-10).
+
+Mangoporn, streamporn.li i pandamovies przeszly na motyw "vid": obsada tylko w JSON-LD,
+studio i rok w chipach bez rel="tag". Parser tego nie czytal, wiec filmy dodane od ok.
+5.10 przyszly bez obsady i czesto bez studia (cichy fail parsera, ingest konczyl sie
+"errors: 0"). Skrypt pobiera ich strony na nowo i puszcza przez resolve_movie, ktory przy
+znanym refie tylko uzupelnia puste pola i dopina obsade/tagi. Strony przez proxy, bo
+mangoporn z VPS bywa niedostepny (timeout, blad TLS).
+
+  python3 /srv/scripts/backfill_dooplay_metadata.py                 # dry-run: tylko liczy
+  python3 /srv/scripts/backfill_dooplay_metadata.py --apply [--since 2026-10-01]
+"""
+from __future__ import annotations
+
+import argparse
+import functools
+import logging
+
+import app.connectors.dooplay as dp
+from app.config import get_settings
+
+dp.browser_get = functools.partial(dp.browser_get, proxy=get_settings().brightdata_proxy_url)
+
+from sqlalchemy import text  # noqa: E402
+
+from app.db import session_scope  # noqa: E402
+from app.normalize.movies import normalize_movie  # noqa: E402
+from app.resolve.movie_resolver import resolve_movie  # noqa: E402
+
+_CONNECTORS = {
+    "mangoporn": dp.MangopornConnector,
+    "streampornvip": dp.StreampornVipConnector,
+    "pandamovies": dp.PandamoviesConnector,
+}
+
+_REFS_SQL = """
+    SELECT r.movie_id, r.external_id, r.url, so.id AS source_id, so.name
+    FROM movie_external_refs r
+    JOIN sources so ON so.id = r.source_id
+    JOIN movies m ON m.id = r.movie_id
+    WHERE so.name = ANY(:names) AND m.created_at >= :since AND r.url IS NOT NULL
+      AND (m.studio_id IS NULL
+           OR NOT EXISTS (SELECT 1 FROM movie_performers p WHERE p.movie_id = m.id))
+    ORDER BY m.created_at DESC
+"""
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--apply", action="store_true")
+    ap.add_argument("--since", default="2026-10-01")
+    args = ap.parse_args()
+    logging.basicConfig(level=logging.WARNING)
+
+    with session_scope() as s:
+        refs = s.execute(text(_REFS_SQL), {"names": list(_CONNECTORS), "since": args.since}).all()
+    print(f"refow do uzupelnienia: {len(refs)}", flush=True)
+    if not args.apply:
+        return
+
+    conns = {name: cls() for name, cls in _CONNECTORS.items()}
+    done = failed = with_cast = 0
+    for n, ref in enumerate(refs, 1):
+        try:
+            raw = conns[ref.name]._fetch_detail(ref.url)
+        except Exception as e:  # noqa: BLE001
+            failed += 1
+            print(f"  blad {ref.name} {ref.url}: {str(e)[:100]}", flush=True)
+            continue
+        if raw is None:
+            failed += 1
+            continue
+        norm = normalize_movie(raw)
+        if norm.external_id != ref.external_id:
+            failed += 1
+            print(f"  inny id {ref.external_id} -> {norm.external_id}, pomijam", flush=True)
+            continue
+        with session_scope() as s:
+            resolve_movie(s, norm=norm, source_id=ref.source_id)
+        done += 1
+        with_cast += bool(raw.performers)
+        if n % 25 == 0:
+            print(f"... {n}/{len(refs)}, uzupelnione {done}, z obsada {with_cast}, bledy {failed}", flush=True)
+    print(f"uzupelnione: {done}, z obsada: {with_cast}, bledy: {failed}", flush=True)
+
+
+if __name__ == "__main__":
+    main()

scripts/ingest_movies_backfill.py

@@ -0,0 +1,20 @@
+"""Pelne zaciagniecie filmow z jednego zrodla (bez delty), np. po przerwie w ingescie.
+
+Codzienny _job_movie_ingest idzie w trybie przyrostowym (dooplay: 3 strony listy), wiec
+po kilku tygodniach przestoju zrodla zaleglosci same sie nie nadrobia. Pelny tryb dooplay
+czyta do 50 stron. Uzycie (z pliku, w osobnym kontenerze):
+    docker compose run -d --name X worker python3 /srv/scripts/ingest_movies_backfill.py pandamovies [limit]
+"""
+import logging
+import sys
+
+from app.connectors import get_movie_connectors
+from app.ingest import ingest_movies_from_connector
+
+logging.basicConfig(level=logging.INFO)
+name = sys.argv[1]
+limit = int(sys.argv[2]) if len(sys.argv) > 2 else None
+connectors = dict(get_movie_connectors())
+if name not in connectors:
+    sys.exit(f"nie ma konektora filmow {name!r}, sa: {sorted(connectors)}")
+print(ingest_movies_from_connector(connectors[name](), limit=limit, use_delta=False))