chore(scripts): scalanie tych samych odcinkow hentaimama i xanimeporn

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

scripts/merge_hentai_episode_dupes.py

@@ -0,0 +1,79 @@
+"""Scala te same odcinki hentai z hentaimama i xanimeporn (2026-10-15).
+
+Tytuly hentaimama mialy do 15.10 ogon "- Watch Online Free in HD - Hentaimama", a xanime
+dokleja "[Sub-ENG]", wiec ten sam odcinek byl dwiema scenami. Para = scena tylko z
+hentaimama i scena tylko z xanimeporn o tym samym title_normalized po zdjeciu
+"sub eng"/"raw"/"uncensored" z konca. Scalamy wylacznie pary jeden do jednego
+(scena hentaimama zostaje, ma dlugosc i miniature).
+
+    python scripts/merge_hentai_episode_dupes.py            # podglad
+    python scripts/merge_hentai_episode_dupes.py --apply
+"""
+from __future__ import annotations
+
+import argparse
+import logging
+from collections import defaultdict
+
+import sqlalchemy as sa
+
+from app.db import session_scope
+from app.resolve.scene_merge import merge_scenes
+
+log = logging.getLogger(__name__)
+
+_PAIRS_SQL = """
+with x as (
+    select s.id, regexp_replace(s.title_normalized, '\\s+(sub eng|eng sub|english sub|raw|uncensored)$', '') k
+    from scenes s
+    where exists (select 1 from playback_sources p where p.scene_id = s.id and p.origin = 'tube:xanimeporn')
+      and not exists (select 1 from playback_sources p where p.scene_id = s.id and p.origin = 'tube:hentaimama')
+), h as (
+    select s.id, s.title_normalized k
+    from scenes s
+    where exists (select 1 from playback_sources p where p.scene_id = s.id and p.origin = 'tube:hentaimama')
+      and not exists (select 1 from playback_sources p where p.scene_id = s.id and p.origin = 'tube:xanimeporn')
+)
+select h.id as keep_id, x.id as drop_id, h.k from x join h on h.k = x.k
+"""
+
+
+def run(apply: bool) -> dict[str, int]:
+    with session_scope() as s:
+        rows = s.execute(sa.text(_PAIRS_SQL)).all()
+    by_keep: dict = defaultdict(list)
+    by_drop: dict = defaultdict(list)
+    for keep_id, drop_id, _k in rows:
+        by_keep[keep_id].append(drop_id)
+        by_drop[drop_id].append(keep_id)
+    pairs = [(k, d, t) for k, d, t in rows if len(by_keep[k]) == 1 and len(by_drop[d]) == 1]
+    c = {"candidates": len(rows), "one_to_one": len(pairs), "merged": 0, "errors": 0}
+    for keep_id, drop_id, title in pairs[:5]:
+        print(f"{title!r}: keep {keep_id} <- {drop_id}")
+    if not apply:
+        return c
+    for keep_id, drop_id, _t in pairs:
+        try:
+            with session_scope() as s:
+                merge_scenes(s, keep_id=keep_id, drop_id=drop_id, resolved_by="merge_hentai_episode_dupes")
+                s.commit()
+            c["merged"] += 1
+        except Exception as e:
+            c["errors"] += 1
+            log.warning("merge %s <- %s padl: %s", keep_id, drop_id, e)
+        if c["merged"] % 200 == 0 and c["merged"]:
+            log.info("hentai dupes %s", c)
+    return c
+
+
+def main() -> int:
+    logging.basicConfig(level=logging.INFO)
+    p = argparse.ArgumentParser()
+    p.add_argument("--apply", action="store_true")
+    args = p.parse_args()
+    print(run(args.apply))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())