unmanic/media_date_cache.workingslow

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
media_date_cache.py

Builds and applies "best possible" dates to media files, caching choices in SQLite.
Sources, in priority order:

TV (kind=tv):
  1) Sonarr first-import date (history and/or episodeFile.dateAdded)
  2) Sonarr episode airDateUtc
  3) TMDB episode air_date (via folder IMDb ID)
  4) (optional) file mtime fallback if --fallback file_mtime

Movies (kind=movie):
  1) Radarr first-import (history and/or movieFile.dateAdded)
  2) Radarr release dates (digital -> physical -> theatrical)
  3) TMDB release dates (digital -> physical -> theatrical) via folder IMDb ID
  4) (optional) file mtime fallback if --fallback file_mtime

Environment (.env supported):
  SONARR_URL, SONARR_API_KEY
  RADARR_URL, RADARR_API_KEY
  TMDB_API_KEY
  TMDB_PRIMARY_COUNTRY=US
  GOTIFY_URL, GOTIFY_TOKEN, GOTIFY_PRIORITY=5
  GOTIFY_TICK_EVERY=500
  TV_ROOTS=/mnt/unionfs/Media/TV:/mnt/unionfs/Media/TV/tv:/mnt/unionfs/Media/TV/tv6
  MOVIE_ROOTS=/mnt/unionfs/Media/Movies:/mnt/unionfs/Media/Movies/movies:/mnt/unionfs/Media/Movies/movies6

Usage examples:
  python media_date_cache.py apply --kind tv --db /opt/scripts/unmanic/.cache/media_dates.db --root /mnt/unionfs/Media/TV --live-on-miss
  python media_date_cache.py apply --kind movie --db /opt/scripts/unmanic/.cache/media_dates.db --root /mnt/unionfs/Media/Movies --live-on-miss

Requires:
  pip install requests python-dotenv
"""

import argparse
import datetime as dt
import json
import os
import re
import sqlite3
import sys
import time
from pathlib import Path
from typing import Dict, List, Optional, Tuple

try:
    from dotenv import load_dotenv
except Exception:
    load_dotenv = None

import requests

VIDEO_EXTS = {".mkv", ".mp4", ".avi", ".m4v", ".mov", ".wmv", ".ts", ".m2ts"}

# ---------- Utilities ----------

def log(msg: str):
    now = dt.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    print(f"[{now}] {msg}", flush=True)

def load_env(env_file: Optional[str] = None) -> Dict[str, str]:
    if env_file and os.path.isfile(env_file):
        if load_dotenv:
            load_dotenv(env_file)
    else:
        # Auto-load .env from cwd if present
        local = Path(".env")
        if local.exists() and load_dotenv:
            load_dotenv(local)

    return dict(os.environ)

def env_get(env: Dict[str, str], key: str, default: Optional[str] = None) -> Optional[str]:
    val = env.get(key, default)
    return val

def env_get_int(env: Dict[str, str], key: str, default: int) -> int:
    raw = str(env.get(key, str(default)) or str(default))
    # tolerate trailing comments like: "5   # optional"
    raw = raw.strip().split()[0]
    try:
        return int(raw)
    except Exception:
        return default

def parse_utc(ts: str) -> Optional[int]:
    if not ts:
        return None
    # Try a few common formats
    for fmt in ("%Y-%m-%dT%H:%M:%S%z", "%Y-%m-%dT%H:%M:%S.%f%z",
                "%Y-%m-%dT%H:%M:%SZ", "%Y-%m-%d"):
        try:
            if fmt.endswith("Z"):
                dt_obj = dt.datetime.strptime(ts, fmt).replace(tzinfo=dt.timezone.utc)
            elif fmt == "%Y-%m-%d":
                # interpret bare dates as midnight UTC
                dt_obj = dt.datetime.strptime(ts, "%Y-%m-%d").replace(tzinfo=dt.timezone.utc)
            else:
                dt_obj = dt.datetime.strptime(ts, fmt)
            return int(dt_obj.timestamp())
        except Exception:
            continue
    return None

def dt_to_epoch_utc(d: dt.datetime) -> int:
    if d.tzinfo is None:
        d = d.replace(tzinfo=dt.timezone.utc)
    return int(d.timestamp())

def safe_join_paths(paths: List[str]) -> List[str]:
    out = []
    for p in paths:
        if p and os.path.isdir(p):
            out.append(p)
    return out

IMDB_RX = re.compile(r"(?:^|\W)(?:imdb-)?(tt\d{6,9})(?:\W|$)", re.IGNORECASE)
SXXEYY_RX = re.compile(r"[Ss](\d{1,3})[ ._-]*[Ee](\d{1,3})")

def extract_imdb_from_path(path: str) -> Optional[str]:
    # Search each segment (show or movie folder often has it)
    for part in Path(path).parts[::-1]:
        m = IMDB_RX.search(part)
        if m:
            return m.group(1).lower()
    return None

def parse_sxxeyy(path: str) -> Optional[Tuple[int, int]]:
    m = SXXEYY_RX.search(path)
    if not m:
        return None
    try:
        return int(m.group(1)), int(m.group(2))
    except Exception:
        return None

# ---------- Gotify ----------

def gotify_send(env: Dict[str, str], title: str, message: str, priority: Optional[int] = None):
    url = env_get(env, "GOTIFY_URL")
    token = env_get(env, "GOTIFY_TOKEN")
    if not url or not token:
        return
    prio = priority if priority is not None else env_get_int(env, "GOTIFY_PRIORITY", 5)
    try:
        r = requests.post(
            f"{url.rstrip('/')}/message",
            headers={"X-Gotify-Key": token},
            json={"title": title, "message": message, "priority": prio},
            timeout=10,
        )
        r.raise_for_status()
    except Exception as e:
        log(f"GOTIFY WARN: {e}")

# ---------- SQLite ----------

DDL_MEDIA_DATES = """
CREATE TABLE IF NOT EXISTS media_dates (
  path TEXT PRIMARY KEY,
  kind TEXT NOT NULL,
  chosen_ts INTEGER NOT NULL,
  chosen_src TEXT NOT NULL,
  extra_json TEXT,
  first_seen INTEGER NOT NULL,
  last_update INTEGER NOT NULL
);
"""
IDX_MEDIA_DATES_KIND = "CREATE INDEX IF NOT EXISTS idx_media_dates_kind ON media_dates(kind);"

def ensure_db(db_path: str):
    os.makedirs(os.path.dirname(db_path), exist_ok=True)
    with sqlite3.connect(db_path) as cx:
        cx.execute(DDL_MEDIA_DATES)
        cx.execute(IDX_MEDIA_DATES_KIND)
        cx.commit()

def db_get_row(cx: sqlite3.Connection, path: str) -> Optional[Tuple]:
    cur = cx.execute("SELECT path, kind, chosen_ts, chosen_src, extra_json, first_seen, last_update FROM media_dates WHERE path=?", (path,))
    return cur.fetchone()

def db_upsert(cx: sqlite3.Connection, path: str, kind: str, chosen_ts: int, chosen_src: str, extra: dict):
    now = int(time.time())
    prev = db_get_row(cx, path)
    if prev:
        first_seen = prev[5]
        cx.execute(
            "UPDATE media_dates SET chosen_ts=?, chosen_src=?, extra_json=?, last_update=? WHERE path=?",
            (chosen_ts, chosen_src, json.dumps(extra, ensure_ascii=False), now, path),
        )
    else:
        first_seen = now
        cx.execute(
            "INSERT INTO media_dates (path, kind, chosen_ts, chosen_src, extra_json, first_seen, last_update) VALUES (?,?,?,?,?,?,?)",
            (path, kind, chosen_ts, chosen_src, json.dumps(extra, ensure_ascii=False), first_seen, now),
        )

# ---------- API clients (Sonarr, Radarr, TMDB) ----------

def _req_json(method: str, url: str, headers=None, params=None, timeout=30):
    r = requests.request(method, url, headers=headers or {}, params=params or {}, timeout=timeout)
    r.raise_for_status()
    return r.json()

class SonarrClient:
    def __init__(self, base: Optional[str], key: Optional[str]):
        self.base = base.rstrip("/") if base else None
        self.key = key
        self._series = None  # list
        self._episodes = {}  # seriesId -> list
        self._history = {}   # seriesId -> list
        self._epfiles = {}   # seriesId -> list

    def ok(self) -> bool:
        return bool(self.base and self.key)

    def _hdr(self):
        return {"X-Api-Key": self.key}

    def series(self) -> List[dict]:
        if self._series is not None:
            return self._series
        if not self.ok():
            self._series = []
            return self._series
        log("Sonarr: loading series index…")
        js = _req_json("GET", f"{self.base}/api/v3/series", headers=self._hdr())
        if isinstance(js, list):
            self._series = js
        else:
            # defensive
            self._series = list(js) if js else []
        log(f"Sonarr: indexed {len(self._series)} series")
        return self._series

    def episodes_for(self, series_id: int) -> List[dict]:
        if series_id in self._episodes:
            return self._episodes[series_id]
        if not self.ok():
            self._episodes[series_id] = []
            return self._episodes[series_id]
        js = _req_json("GET", f"{self.base}/api/v3/episode", headers=self._hdr(), params={"seriesId": series_id})
        if isinstance(js, list):
            out = js
        elif isinstance(js, dict) and "records" in js:
            out = js["records"]
        else:
            out = []
        self._episodes[series_id] = out
        return out

    def episode_files_for(self, series_id: int) -> List[dict]:
        if series_id in self._epfiles:
            return self._epfiles[series_id]
        if not self.ok():
            self._epfiles[series_id] = []
            return self._epfiles[series_id]
        js = _req_json("GET", f"{self.base}/api/v3/episodefile", headers=self._hdr(), params={"seriesId": series_id})
        self._epfiles[series_id] = js if isinstance(js, list) else []
        return self._epfiles[series_id]

    def history_for(self, series_id: int) -> List[dict]:
        if series_id in self._history:
            return self._history[series_id]
        if not self.ok():
            self._history[series_id] = []
            return self._history[series_id]
        # series history paginated
        page = 1
        page_size = 500
        records: List[dict] = []
        while True:
            js = _req_json(
                "GET",
                f"{self.base}/api/v3/history/series",
                headers=self._hdr(),
                params={"seriesId": series_id, "includeEpisode": "true", "page": page, "pageSize": page_size},
            )
            chunk = []
            if isinstance(js, dict):
                chunk = js.get("records") or js.get("results") or []
                total = js.get("totalRecords") or js.get("total") or 0
            elif isinstance(js, list):
                chunk = js
                total = len(js)
            records.extend(chunk)
            if isinstance(js, dict) and js.get("page") is not None:
                # If paginated, stop when we fetched all
                if len(records) >= total:
                    break
                page += 1
            else:
                # Not paginated
                break
        self._history[series_id] = records
        return records

class RadarrClient:
    def __init__(self, base: Optional[str], key: Optional[str]):
        self.base = base.rstrip("/") if base else None
        self.key = key
        self._movies = None
        self._history = None
        self._files_by_movie = {}

    def ok(self) -> bool:
        return bool(self.base and self.key)

    def _hdr(self):
        return {"X-Api-Key": self.key}

    def movies(self) -> List[dict]:
        if self._movies is not None:
            return self._movies
        if not self.ok():
            self._movies = []
            return self._movies
        js = _req_json("GET", f"{self.base}/api/v3/movie", headers=self._hdr())
        self._movies = js if isinstance(js, list) else []
        return self._movies

    def history(self) -> List[dict]:
        if self._history is not None:
            return self._history
        if not self.ok():
            self._history = []
            return self._history
        page = 1
        page_size = 500
        records: List[dict] = []
        while True:
            js = _req_json("GET", f"{self.base}/api/v3/history", headers=self._hdr(), params={"page": page, "pageSize": page_size})
            chunk = []
            total = 0
            if isinstance(js, dict):
                chunk = js.get("records") or js.get("results") or []
                total = js.get("totalRecords") or js.get("total") or 0
            elif isinstance(js, list):
                chunk = js
                total = len(js)
            records.extend(chunk)
            if isinstance(js, dict) and js.get("page") is not None:
                if len(records) >= total:
                    break
                page += 1
            else:
                break
        self._history = records
        return records

    def moviefile_for(self, movie_id: int) -> Optional[dict]:
        if movie_id in self._files_by_movie:
            return self._files_by_movie[movie_id]
        if not self.ok():
            self._files_by_movie[movie_id] = None
            return None
        js = _req_json("GET", f"{self.base}/api/v3/moviefile", headers=self._hdr(), params={"movieId": movie_id})
        # could be list, pick first if any
        mf = js[0] if isinstance(js, list) and js else None
        self._files_by_movie[movie_id] = mf
        return mf

class TMDBClient:
    def __init__(self, api_key: Optional[str], primary_country: str = "US"):
        self.key = api_key
        self.primary_country = (primary_country or "US").upper()
        self.base = "https://api.themoviedb.org/3"
        self._tv_id_cache: Dict[str, int] = {}      # imdb_id -> tmdb_tv_id
        self._movie_id_cache: Dict[str, int] = {}   # imdb_id -> tmdb_movie_id
        self._season_cache: Dict[Tuple[int, int], dict] = {}  # (tv_id, season)-> episodes mapping

    def ok(self) -> bool:
        return bool(self.key)

    def _get(self, path: str, params: dict = None):
        if not self.ok():
            return None
        params = dict(params or {})
        params["api_key"] = self.key
        return _req_json("GET", f"{self.base}{path}", params=params)

    def tv_id_by_imdb(self, imdb_id: str) -> Optional[int]:
        imdb_id = imdb_id.lower()
        if imdb_id in self._tv_id_cache:
            return self._tv_id_cache[imdb_id]
        js = self._get(f"/find/{imdb_id}", {"external_source": "imdb_id"})
        tid = None
        if isinstance(js, dict):
            tvres = js.get("tv_results") or []
            if tvres:
                tid = tvres[0].get("id")
        self._tv_id_cache[imdb_id] = tid
        return tid

    def movie_id_by_imdb(self, imdb_id: str) -> Optional[int]:
        imdb_id = imdb_id.lower()
        if imdb_id in self._movie_id_cache:
            return self._movie_id_cache[imdb_id]
        js = self._get(f"/find/{imdb_id}", {"external_source": "imdb_id"})
        mid = None
        if isinstance(js, dict):
            mres = js.get("movie_results") or []
            if mres:
                mid = mres[0].get("id")
        self._movie_id_cache[imdb_id] = mid
        return mid

    def season_airdates(self, tv_id: int, season_number: int) -> Dict[int, str]:
        key = (tv_id, season_number)
        if key in self._season_cache:
            return self._season_cache[key]
        js = self._get(f"/tv/{tv_id}/season/{season_number}")
        mapping: Dict[int, str] = {}
        if isinstance(js, dict):
            for ep in js.get("episodes") or []:
                epno = ep.get("episode_number")
                ad = ep.get("air_date")
                if epno and ad:
                    mapping[int(epno)] = ad
        self._season_cache[key] = mapping
        return mapping

    def movie_best_release(self, movie_id: int) -> Optional[Tuple[str, str]]:
        """
        Returns (src_key, release_date_iso) where src_key in {'tmdb_digital','tmdb_physical','tmdb_theatrical'}
        Preference order: digital(4) -> physical(5) -> theatrical(3), prioritizing primary_country, else any.
        """
        js = self._get(f"/movie/{movie_id}/release_dates")
        if not isinstance(js, dict):
            return None
        results = js.get("results") or []
        def pick(for_country: Optional[str]) -> Optional[Tuple[str, str]]:
            best = {}
            for entry in results:
                if for_country and (entry.get("iso_3166_1") != for_country):
                    continue
                for rel in entry.get("release_dates") or []:
                    rtype = int(rel.get("type") or 0)
                    rdate = rel.get("release_date")
                    if not rdate:
                        continue
                    if rtype == 4:   # digital
                        best.setdefault("tmdb_digital", []).append(rdate)
                    elif rtype == 5: # physical
                        best.setdefault("tmdb_physical", []).append(rdate)
                    elif rtype == 3: # theatrical
                        best.setdefault("tmdb_theatrical", []).append(rdate)
            # choose earliest of preferred buckets
            for key in ("tmdb_digital", "tmdb_physical", "tmdb_theatrical"):
                if key in best:
                    return key, sorted(best[key])[0]
            return None
        # Try primary country first, then any
        chosen = pick(self.primary_country) or pick(None)
        return chosen

# ---------- Matching helpers ----------

def longest_prefix_match(target: str, prefix_map: Dict[str, dict]) -> Optional[dict]:
    """
    prefix_map: series_path -> series_obj (Sonarr), or movie_path -> movie_obj (Radarr)
    Returns the mapped object whose path is the longest prefix of target.
    """
    candidates = []
    t = target.lower()
    for p, obj in prefix_map.items():
        pl = p.lower()
        if t.startswith(pl.rstrip("/") + "/") or t == pl:
            candidates.append((len(pl), obj))
    if not candidates:
        return None
    candidates.sort(key=lambda x: x[0], reverse=True)
    return candidates[0][1]

def expand_roots_from_env(env: Dict[str, str], kind: str, cli_roots: List[str]) -> List[str]:
    roots = []
    if cli_roots:
        roots.extend(cli_roots)
    else:
        env_key = "TV_ROOTS" if kind == "tv" else "MOVIE_ROOTS"
        val = env_get(env, env_key)
        if val:
            roots.extend([p.strip() for p in val.split(":") if p.strip()])
    # Add known child dirs automatically if they exist
    expanded = []
    for r in roots:
        expanded.append(r)
        # Only add children for top-level TV/Movie roots
        if kind == "tv":
            for child in ("tv", "tv6"):
                c = os.path.join(r, child)
                if os.path.isdir(c):
                    expanded.append(c)
        else:
            for child in ("movies", "movies6"):
                c = os.path.join(r, child)
                if os.path.isdir(c):
                    expanded.append(c)
    # dedupe and only keep existing dirs
    uniq = []
    seen = set()
    for p in expanded:
        if os.path.isdir(p) and p not in seen:
            uniq.append(p)
            seen.add(p)
    return uniq

def walk_media_files(roots: List[str]) -> List[str]:
    files = []
    for root in roots:
        for dirpath, _, filenames in os.walk(root):
            for fn in filenames:
                if Path(fn).suffix.lower() in VIDEO_EXTS:
                    files.append(os.path.join(dirpath, fn))
    return files

# ---------- Choosing timestamps ----------

def sonarr_choose_for_file(sc: SonarrClient, file_path: str) -> Optional[Tuple[int, str, dict]]:
    """
    If the file belongs to a Sonarr series (by path prefix), try to determine a timestamp:
      - earliest import (history or episodeFile.dateAdded)
      - else episode airDateUtc
    Returns (ts, src, extra) or None
    """
    if not sc.ok():
        return None
    series_index = {s.get("path") or "": s for s in sc.series() if (s.get("path") or "").strip()}
    series = longest_prefix_match(file_path, series_index)
    if not series:
        return None

    sid = series.get("id")
    eps = sc.episodes_for(sid)
    epfiles = sc.episode_files_for(sid)
    hist = sc.history_for(sid)

    # Build lookup maps
    # Map (season,episode) -> episode id and airDateUtc
    ep_map: Dict[Tuple[int, int], dict] = {}
    for e in eps:
        try:
            k = (int(e.get("seasonNumber") or 0), int(e.get("episodeNumber") or 0))
            ep_map[k] = e
        except Exception:
            continue

    # episodeId -> earliest import date (from history)
    import_map: Dict[int, int] = {}
    for r in hist:
        try:
            ev = (r.get("eventType") or r.get("event") or "")
            evl = str(ev).lower()
            if "import" not in evl:
                continue
            edate = r.get("date")
            ts = parse_utc(edate)
            if ts is None:
                continue
            ep_obj = r.get("episode") or {}
            eid = ep_obj.get("id")
            if not eid:
                continue
            prev = import_map.get(eid)
            if prev is None or ts < prev:
                import_map[eid] = ts
        except Exception:
            continue

    # episodeFile dateAdded (another hint)
    # episodeFile doesn't directly map to episodeId always, but often the filename/relativePath implies SxxEyy.
    # We'll only use dateAdded as a general earliest among files that share the same S/E when we can parse S/E.
    # We'll handle per-file below.

    # Parse S/E from the path
    se = parse_sxxeyy(file_path)
    if not se:
        return None
    s_num, e_num = se
    ep = ep_map.get((s_num, e_num))
    if not ep:
        return None
    eid = ep.get("id")

    # Candidate 1: history first import for this episode
    candidates: List[Tuple[int, str]] = []
    if eid and eid in import_map:
        candidates.append((import_map[eid], "sonarr_import"))

    # Candidate 2: episodeFile.dateAdded for matching S/E if any
    # Try to match on season/episode parsed from episodeFile.relativePath
    earliest_epfile_ts = None
    for ef in epfiles:
        rel = ef.get("relativePath") or ""
        se2 = parse_sxxeyy(rel)
        if se2 == (s_num, e_num):
            ts2 = parse_utc(ef.get("dateAdded"))
            if ts2 is not None and (earliest_epfile_ts is None or ts2 < earliest_epfile_ts):
                earliest_epfile_ts = ts2
    if earliest_epfile_ts is not None:
        candidates.append((earliest_epfile_ts, "episodefile_dateAdded"))

    # Candidate 3: airDateUtc
    air_ts = parse_utc(ep.get("airDateUtc") or ep.get("airDate"))
    if air_ts is not None:
        candidates.append((air_ts, "tv_airdate"))

    if not candidates:
        return None
    candidates.sort(key=lambda x: x[0])
    chosen_ts, chosen_src = candidates[0]
    extra = {
        "seriesId": sid,
        "seriesTitle": series.get("title"),
        "season": s_num,
        "episode": e_num,
        "candidates": [{"ts": ts, "src": src} for ts, src in candidates],
    }
    return chosen_ts, chosen_src, extra

def radarr_choose_for_file(rc: RadarrClient, file_path: str) -> Optional[Tuple[int, str, dict]]:
    """
    Determine movie timestamp using Radarr:
      - earliest import from history/movieFile.dateAdded
      - else Radarr release dates (digital->physical->theatrical)
    """
    if not rc.ok():
        return None
    movies_index = {m.get("path") or "": m for m in rc.movies() if (m.get("path") or "").strip()}
    movie = longest_prefix_match(file_path, movies_index)
    if not movie:
        return None
    mid = movie.get("id")

    candidates: List[Tuple[int, str]] = []

    # History import events
    for r in rc.history():
        try:
            ev = (r.get("eventType") or r.get("event") or "")
            evl = str(ev).lower()
            if "import" not in evl:
                continue
            mobj = r.get("movie") or {}
            if mobj.get("id") != mid:
                continue
            ts = parse_utc(r.get("date"))
            if ts is not None:
                candidates.append((ts, "radarr_import"))
        except Exception:
            continue
    # Movie file dateAdded
    mf = rc.moviefile_for(mid)
    mf_ts = parse_utc(mf.get("dateAdded")) if mf else None
    if mf_ts is not None:
        candidates.append((mf_ts, "moviefile_dateAdded"))

    # Radarr release dates on the movie object itself
    # Radarr's movie object has fields like 'digitalRelease', 'physicalRelease', 'inCinemas'
    for key, src in (("digitalRelease", "movie_digital"),
                     ("physicalRelease", "movie_physical"),
                     ("inCinemas", "movie_cinema")):
        ts = parse_utc(movie.get(key))
        if ts is not None:
            candidates.append((ts, src))

    if not candidates:
        return None
    candidates.sort(key=lambda x: x[0])
    chosen_ts, chosen_src = candidates[0]
    extra = {
        "movieId": mid,
        "movieTitle": movie.get("title"),
        "candidates": [{"ts": ts, "src": src} for ts, src in candidates],
    }
    return chosen_ts, chosen_src, extra

def tmdb_tv_choose(file_path: str, tmdb: TMDBClient) -> Optional[Tuple[int, str, dict]]:
    """
    For a TV episode when Sonarr lacks it:
      - Extract IMDb ID from path
      - Map to TMDB tv_id
      - Parse season/episode from filename
      - Fetch season's episodes and get air_date for the episode
    """
    if not tmdb.ok():
        return None
    imdb_id = extract_imdb_from_path(file_path)
    if not imdb_id:
        return None
    se = parse_sxxeyy(file_path)
    if not se:
        return None
    s_num, e_num = se
    tv_id = tmdb.tv_id_by_imdb(imdb_id)
    if not tv_id:
        return None
    airdates = tmdb.season_airdates(tv_id, s_num)
    iso = airdates.get(e_num)
    if not iso:
        return None
    ts = parse_utc(iso)
    if ts is None:
        return None
    extra = {
        "imdb": imdb_id,
        "tmdb_tv_id": tv_id,
        "season": s_num,
        "episode": e_num,
        "tmdb_air_date": iso,
    }
    return ts, "tmdb_airdate", extra

def tmdb_movie_choose(file_path: str, tmdb: TMDBClient) -> Optional[Tuple[int, str, dict]]:
    if not tmdb.ok():
        return None
    imdb_id = extract_imdb_from_path(file_path)
    if not imdb_id:
        return None
    movie_id = tmdb.movie_id_by_imdb(imdb_id)
    if not movie_id:
        return None
    picked = tmdb.movie_best_release(movie_id)
    if not picked:
        return None
    src, iso = picked
    ts = parse_utc(iso)
    if ts is None:
        return None
    extra = {"imdb": imdb_id, "tmdb_movie_id": movie_id, "tmdb_release": {"src": src, "date": iso}}
    return ts, src, extra

# ---------- Apply ----------

def apply_mtime(path: str, ts: int, dry_run: bool = False) -> bool:
    try:
        if dry_run:
            return True
        os.utime(path, (ts, ts))
        return True
    except Exception as e:
        log(f"ERROR utime failed for {path}: {e}")
        return False

def apply_command(args):
    env = load_env(args.env)
    ensure_db(args.db)

    # Clients
    sc = SonarrClient(env_get(env, "SONARR_URL"), env_get(env, "SONARR_API_KEY"))
    rc = RadarrClient(env_get(env, "RADARR_URL"), env_get(env, "RADARR_API_KEY"))
    tmdb = TMDBClient(env_get(env, "TMDB_API_KEY"), env_get(env, "TMDB_PRIMARY_COUNTRY") or "US")

    tick_every = env_get_int(env, "GOTIFY_TICK_EVERY", 500)

    # Roots
    roots = expand_roots_from_env(env, args.kind, args.root or [])
    if not roots:
        log(f"ERROR: No roots found for kind={args.kind}. Use --root or set TV_ROOTS/MOVIE_ROOTS.")
        sys.exit(2)

    files = walk_media_files(roots)
    log(f"Found {len(files)} media file(s) under {', '.join(roots)}")

    # If asked, prewarm Sonarr/Radarr series indexes (episodes/history lazy anyway)
    if args.kind == "tv" and args.live_on_miss:
        log("Initializing Sonarr (series index only)…")
        sc.series()  # index
    if args.kind == "movie" and args.live_on_miss:
        log("Initializing Radarr (movies index only)…")
        rc.movies()

    gotify_send(env, "Apply mtimes: start", f"kind={args.kind}, files={len(files)}")

    updated = 0
    checked = 0
    misses = 0
    with sqlite3.connect(args.db) as cx:
        for path in files:
            checked += 1
            # Existing cached?
            row = db_get_row(cx, path)
            if row and not args.refresh:
                # Already have it — apply to file (unless skipping)
                ts, src = row[2], row[3]
                ok = apply_mtime(path, ts, dry_run=args.dry_run)
                if ok:
                    updated += 1
                if checked % tick_every == 0:
                    gotify_send(env, "Apply mtimes: progress", f"{checked}/{len(files)} (cached)")
                continue

            # Need to compute live
            chosen: Optional[Tuple[int, str, dict]] = None

            if args.kind == "tv":
                if args.live_on_miss and sc.ok():
                    chosen = sonarr_choose_for_file(sc, path)
                if not chosen:
                    # TMDB fallback
                    chosen = tmdb_tv_choose(path, tmdb)
            else:
                # movies
                if args.live_on_miss and rc.ok():
                    chosen = radarr_choose_for_file(rc, path)
                if not chosen:
                    chosen = tmdb_movie_choose(path, tmdb)

            if not chosen and args.fallback == "file_mtime":
                try:
                    st = os.stat(path)
                    ts = int(st.st_mtime)
                    chosen = (ts, "file_mtime", {})
                except Exception:
                    chosen = None

            if not chosen:
                log(f"WARN: no import/release date for {path}")
                misses += 1
                if checked % tick_every == 0:
                    gotify_send(env, "Apply mtimes: progress", f"{checked}/{len(files)} (misses so far: {misses})")
                continue

            ts, src, extra = chosen
            # Upsert DB & apply
            db_upsert(cx, path, args.kind, ts, src, extra)
            ok = apply_mtime(path, ts, dry_run=args.dry_run)
            if ok:
                updated += 1

            if checked % tick_every == 0:
                gotify_send(env, "Apply mtimes: progress", f"{checked}/{len(files)} (updated: {updated})")

        cx.commit()

    gotify_send(env, "Apply mtimes: done", f"kind={args.kind}, updated={updated}, checked={checked}, misses={misses}")
    log(f"Done. kind={args.kind} checked={checked} updated={updated} misses={misses} dry_run={args.dry_run}")

# ---------- Build cache (optional prefetch) ----------

def build_cache_command(args):
    env = load_env(args.env)
    ensure_db(args.db)

    sc = SonarrClient(env_get(env, "SONARR_URL"), env_get(env, "SONARR_API_KEY"))
    rc = RadarrClient(env_get(env, "RADARR_URL"), env_get(env, "RADARR_API_KEY"))
    tmdb = TMDBClient(env_get(env, "TMDB_API_KEY"), env_get(env, "TMDB_PRIMARY_COUNTRY") or "US")

    gotify_send(env, "Cache build: start", f"kind={args.kind}")

    roots = expand_roots_from_env(env, args.kind, args.root or [])
    files = walk_media_files(roots) if roots else []
    if roots:
        log(f"Found {len(files)} media file(s) under {', '.join(roots)}")

    inserted = 0
    checked = 0
    with sqlite3.connect(args.db) as cx:
        if args.kind == "tv" and sc.ok():
            series = sc.series()
            total = len(series)
            for i, s in enumerate(series, start=1):
                log(f"Sonarr: loading episodes for series {s.get('id')} ({i}/{total})…")
                eps = sc.episodes_for(s.get("id"))
                log(f"Sonarr: loading history for series {s.get('id')}…")
                hist = sc.history_for(s.get("id"))
                epfiles = sc.episode_files_for(s.get("id"))
                # We are not mapping specific files here; build-cache is optional.
                # If roots provided, we will also pass over files and insert rows using live logic:
        # If roots provided, populate using the same chooser logic as apply but without touching mtimes.
        if files:
            for path in files:
                checked += 1
                if db_get_row(cx, path) and not args.refresh:
                    continue
                chosen = None
                if args.kind == "tv":
                    if sc.ok():
                        chosen = sonarr_choose_for_file(sc, path)
                    if not chosen:
                        chosen = tmdb_tv_choose(path, tmdb)
                else:
                    if rc.ok():
                        chosen = radarr_choose_for_file(rc, path)
                    if not chosen:
                        chosen = tmdb_movie_choose(path, tmdb)
                if chosen:
                    ts, src, extra = chosen
                    db_upsert(cx, path, args.kind, ts, src, extra)
                    inserted += 1
                elif args.fallback == "file_mtime":
                    try:
                        st = os.stat(path)
                        ts = int(st.st_mtime)
                        db_upsert(cx, path, args.kind, ts, "file_mtime", {})
                        inserted += 1
                    except Exception:
                        pass
                if checked % 500 == 0:
                    gotify_send(env, "Cache build: progress", f"{checked}/{len(files)} cached")
        cx.commit()

    gotify_send(env, "Cache build: done", f"kind={args.kind}, inserted={inserted}, scanned={checked or 'n/a'}")
    log(f"Cache build done. kind={args.kind} inserted={inserted} scanned={checked}")

# ---------- CLI ----------

def main():
    ap = argparse.ArgumentParser(description="Media date cache & mtime applier (Sonarr/Radarr + TMDB fallback)")
    sub = ap.add_subparsers(dest="cmd", required=True)

    common = argparse.ArgumentParser(add_help=False)
    common.add_argument("--db", required=True, help="SQLite database path")
    common.add_argument("--env", default=None, help="Path to .env (optional, defaults to ./.env if present)")
    common.add_argument("--kind", choices=["tv", "movie"], required=True, help="Kind of media this run targets")
    common.add_argument("--root", action="append", help="Root path(s) to scan; can repeat. If omitted, uses TV_ROOTS/MOVIE_ROOTS.")
    common.add_argument("--refresh", action="store_true", help="Ignore existing DB rows; recompute live")
    common.add_argument("--fallback", choices=["none", "file_mtime"], default="none", help="Enable last-resort fallback")
    common.add_argument("--dry-run", action="store_true", help="Compute and write DB, but do not change file mtimes")

    ap_apply = sub.add_parser("apply", parents=[common], help="Apply mtimes to files, caching choices")
    ap_apply.add_argument("--live-on-miss", action="store_true", help="If DB row missing, query Sonarr/Radarr/TMDB live")

    ap_build = sub.add_parser("build-cache", parents=[common], help="Build/populate cache without touching mtimes")
    # build-cache shares flags

    args = ap.parse_args()

    if args.cmd == "apply":
        apply_command(args)
    elif args.cmd == "build-cache":
        build_cache_command(args)
    else:
        ap.print_help()

if __name__ == "__main__":
    main()