ranger/core/cache.py

"""
Cache SQLite de Ranger.

Trois usages :
  - availability : statut cache des hashes par débrideur (évite de re-vérifier
    les mêmes hashes à chaque requête Stremio)
  - searches     : résultats de recherche par tracker (les épisodes suivants
    d'une même série réutilisent la recherche)
  - meta         : métadonnées TMDB/Cinemeta par ID

Accès synchrone (opérations courtes, WAL activé) — suffisant pour un addon
mono-processus.
"""

import hashlib
import json
import logging
import os
import sqlite3
import threading
import time


def fingerprint(*secrets):
    """
    Empreinte courte et non réversible d'identifiants (clé API, login...),
    utilisée pour namespacer les clés de cache par compte. Sans ça, deux
    utilisateurs distincts d'une même instance Ranger (plusieurs profils
    d'une même famille, par ex.) avec des credentials différents pourraient
    se voir servir les résultats/liens de l'un ou l'autre pendant la fenêtre
    de cache. Retourne "" si aucun secret fourni (cache partagé, volontaire
    pour les trackers publics sans credential).
    """
    raw = "|".join(str(s) for s in secrets if s)
    if not raw:
        return ""
    return hashlib.sha256(raw.encode("utf-8")).hexdigest()[:12]

DB_PATH = os.getenv("RANGER_DB", os.path.join(os.path.dirname(os.path.dirname(__file__)), "data", "ranger.db"))

# TTLs (secondes), surchargables par variables d'environnement
TTL_AVAIL_CACHED = int(os.getenv("RANGER_TTL_CACHED", 6 * 3600))      # torrent vu en cache
TTL_AVAIL_MISS = int(os.getenv("RANGER_TTL_UNCACHED", 20 * 60))       # torrent vu non-caché
TTL_SEARCH = int(os.getenv("RANGER_TTL_SEARCH", 30 * 60))
# Un résultat de recherche VIDE est mis en cache moins longtemps qu'un résultat
# non-vide : sinon, un tracker en accroc passager (timeout, blip réseau, API
# qui répond 200 avec 0 item) verrouille "aucun résultat" pour TOUT le monde
# jusqu'à expiration du TTL_SEARCH complet (30 min par défaut), y compris une
# fois le tracker redevenu normal. Voir core/search.py::_cached_search.
TTL_SEARCH_EMPTY = int(os.getenv("RANGER_TTL_SEARCH_EMPTY", 3 * 60))
TTL_META = int(os.getenv("RANGER_TTL_META", 7 * 24 * 3600))
TTL_LINK = int(os.getenv("RANGER_TTL_LINK", 15 * 60))                 # lien débrideur résolu

_lock = threading.Lock()
_conn = None

# Compteurs de performance du cache (depuis le démarrage)
_metrics = {"search_hit": 0, "search_miss": 0, "avail_hit": 0, "avail_miss": 0}


def _get_conn():
    global _conn
    if _conn is None:
        os.makedirs(os.path.dirname(DB_PATH), exist_ok=True)
        _conn = sqlite3.connect(DB_PATH, check_same_thread=False)
        _conn.execute("PRAGMA journal_mode=WAL")
        _conn.execute("PRAGMA synchronous=NORMAL")
        _conn.executescript(
            """
            CREATE TABLE IF NOT EXISTS availability (
                service TEXT NOT NULL,
                hash TEXT NOT NULL,
                cached INTEGER NOT NULL,
                checked_at INTEGER NOT NULL,
                PRIMARY KEY (service, hash)
            );
            CREATE TABLE IF NOT EXISTS searches (
                key TEXT PRIMARY KEY,
                results TEXT NOT NULL,
                created_at INTEGER NOT NULL
            );
            CREATE TABLE IF NOT EXISTS meta (
                key TEXT PRIMARY KEY,
                value TEXT NOT NULL,
                created_at INTEGER NOT NULL
            );
            CREATE TABLE IF NOT EXISTS links (
                key TEXT PRIMARY KEY,
                url TEXT NOT NULL,
                created_at INTEGER NOT NULL
            );
            """
        )
        _conn.commit()
        logging.info(f"Cache SQLite initialisé : {DB_PATH}")
    return _conn


def get_availability(service, hashes):
    """
    Retourne (known, unknown) :
      known   : {hash: bool} pour les hashes dont le statut en cache est encore valide
      unknown : liste des hashes à re-vérifier auprès du débrideur
    """
    if not hashes:
        return {}, []
    now = int(time.time())
    known = {}
    with _lock:
        conn = _get_conn()
        placeholders = ",".join("?" * len(hashes))
        rows = conn.execute(
            f"SELECT hash, cached, checked_at FROM availability WHERE service = ? AND hash IN ({placeholders})",
            [service] + list(hashes),
        ).fetchall()
    for h, cached, checked_at in rows:
        ttl = TTL_AVAIL_CACHED if cached else TTL_AVAIL_MISS
        if now - checked_at <= ttl:
            known[h] = bool(cached)
    unknown = [h for h in hashes if h not in known]
    _metrics["avail_hit"] += len(known)
    _metrics["avail_miss"] += len(unknown)
    return known, unknown


def set_availability(service, availability):
    """Enregistre un dict {hash: bool} pour un débrideur."""
    if not availability:
        return
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        conn.executemany(
            "INSERT OR REPLACE INTO availability (service, hash, cached, checked_at) VALUES (?, ?, ?, ?)",
            [(service, h, 1 if c else 0, now) for h, c in availability.items()],
        )
        conn.commit()


def mark_cached(service, info_hash):
    """Marque un hash comme caché (après un débridage réussi)."""
    set_availability(service, {info_hash: True})


def get_search(key):
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        row = conn.execute("SELECT results, created_at FROM searches WHERE key = ?", (key,)).fetchone()
    if row:
        try:
            results = json.loads(row[0])
            ttl = TTL_SEARCH if results else TTL_SEARCH_EMPTY
            if now - row[1] <= ttl:
                _metrics["search_hit"] += 1
                return results
        except Exception:
            pass
    _metrics["search_miss"] += 1
    return None


def set_search(key, results):
    with _lock:
        conn = _get_conn()
        conn.execute(
            "INSERT OR REPLACE INTO searches (key, results, created_at) VALUES (?, ?, ?)",
            (key, json.dumps(results, ensure_ascii=False), int(time.time())),
        )
        conn.commit()


def get_link(key):
    """Lien débrideur résolu, encore valide (TTL court)."""
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        row = conn.execute("SELECT url, created_at FROM links WHERE key = ?", (key,)).fetchone()
    if row and now - row[1] <= TTL_LINK:
        return row[0]
    return None


def set_link(key, url):
    if not url:
        return
    with _lock:
        conn = _get_conn()
        conn.execute(
            "INSERT OR REPLACE INTO links (key, url, created_at) VALUES (?, ?, ?)",
            (key, url, int(time.time())),
        )
        conn.commit()


def get_meta(key):
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        row = conn.execute("SELECT value, created_at FROM meta WHERE key = ?", (key,)).fetchone()
    if row and now - row[1] <= TTL_META:
        try:
            return json.loads(row[0])
        except Exception:
            return None
    return None


def set_meta(key, value):
    with _lock:
        conn = _get_conn()
        conn.execute(
            "INSERT OR REPLACE INTO meta (key, value, created_at) VALUES (?, ?, ?)",
            (key, json.dumps(value, ensure_ascii=False), int(time.time())),
        )
        conn.commit()


def cleanup():
    """Purge les entrées expirées (appelé périodiquement)."""
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        conn.execute("DELETE FROM availability WHERE checked_at < ?", (now - max(TTL_AVAIL_CACHED, TTL_AVAIL_MISS),))
        conn.execute("DELETE FROM searches WHERE created_at < ?", (now - TTL_SEARCH,))
        conn.execute("DELETE FROM meta WHERE created_at < ?", (now - TTL_META,))
        conn.execute("DELETE FROM links WHERE created_at < ?", (now - TTL_LINK,))
        conn.commit()


# ============================================================================
# Fonctions d'administration (panel admin)
# ============================================================================

def get_metrics():
    """Compteurs hit/miss du cache depuis le démarrage."""
    m = dict(_metrics)
    s_tot = m["search_hit"] + m["search_miss"]
    a_tot = m["avail_hit"] + m["avail_miss"]
    m["search_hit_rate"] = round(m["search_hit"] / s_tot * 100, 1) if s_tot else 0.0
    m["avail_hit_rate"] = round(m["avail_hit"] / a_tot * 100, 1) if a_tot else 0.0
    return m


def stats():
    """Statistiques globales du cache SQLite pour le dashboard admin."""
    with _lock:
        conn = _get_conn()
        avail_total = conn.execute("SELECT COUNT(*) FROM availability").fetchone()[0]
        avail_cached = conn.execute("SELECT COUNT(*) FROM availability WHERE cached = 1").fetchone()[0]
        by_service = conn.execute(
            "SELECT service, COUNT(*), SUM(cached) FROM availability GROUP BY service ORDER BY COUNT(*) DESC"
        ).fetchall()
        searches_total = conn.execute("SELECT COUNT(*) FROM searches").fetchone()[0]
        by_source = conn.execute(
            "SELECT substr(key, 8, instr(substr(key, 8), ':') - 1) AS src, COUNT(*) "
            "FROM searches GROUP BY src ORDER BY COUNT(*) DESC"
        ).fetchall()
        meta_total = conn.execute("SELECT COUNT(*) FROM meta").fetchone()[0]
        links_total = conn.execute("SELECT COUNT(*) FROM links").fetchone()[0]

    db_size = os.path.getsize(DB_PATH) if os.path.exists(DB_PATH) else 0
    for wal in (DB_PATH + "-wal", DB_PATH + "-shm"):
        if os.path.exists(wal):
            db_size += os.path.getsize(wal)

    return {
        "db_path": DB_PATH,
        "db_size_bytes": db_size,
        "availability": {
            "total": avail_total,
            "cached": avail_cached,
            "uncached": avail_total - avail_cached,
            "by_service": [
                {"service": s, "total": t, "cached": c or 0} for s, t, c in by_service
            ],
        },
        "searches": {
            "total": searches_total,
            "by_source": [{"source": s or "?", "total": t} for s, t in by_source],
        },
        "meta": {"total": meta_total},
        "links": {"total": links_total},
        "ttl": {
            "cached": TTL_AVAIL_CACHED,
            "uncached": TTL_AVAIL_MISS,
            "search": TTL_SEARCH,
            "meta": TTL_META,
            "link": TTL_LINK,
        },
        "metrics": get_metrics(),
    }


def _parse_search_key(key):
    # search:{source}:{type}:{imdb}[:season:episode]                 (trackers publics, cache partagé)
    # search:{source}:{fingerprint}:{type}:{imdb}[:season:episode]   (trackers à clé, cache namespacé par compte)
    parts = key.split(":")
    if len(parts) < 4 or parts[0] != "search":
        return {"key": key}
    source = parts[1]
    if parts[2] in ("movie", "series"):
        offset, scoped = 2, False
    else:
        offset, scoped = 3, True
    if len(parts) < offset + 2:
        return {"key": key, "source": source}
    info = {"key": key, "source": source, "type": parts[offset], "imdb": parts[offset + 1], "scoped": scoped}
    if len(parts) >= offset + 4:
        info["season"], info["episode"] = parts[offset + 2], parts[offset + 3]
    return info


def list_searches(limit=300):
    """Liste les entrées du cache de recherche (clé parsée + âge)."""
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        rows = conn.execute(
            "SELECT key, created_at, LENGTH(results) FROM searches ORDER BY created_at DESC LIMIT ?",
            (limit,),
        ).fetchall()
    out = []
    for key, created, size in rows:
        info = _parse_search_key(key)
        info["age_seconds"] = now - created
        info["expires_in"] = max(0, TTL_SEARCH - (now - created))
        info["bytes"] = size
        out.append(info)
    return out


def list_meta(limit=300):
    now = int(time.time())
    with _lock:
        conn = _get_conn()
        rows = conn.execute(
            "SELECT key, created_at FROM meta ORDER BY created_at DESC LIMIT ?", (limit,)
        ).fetchall()
    return [{"key": k, "age_seconds": now - c, "expires_in": max(0, TTL_META - (now - c))} for k, c in rows]


def delete_key(table, key):
    """Supprime une entrée précise (table = searches|meta)."""
    if table not in ("searches", "meta"):
        return 0
    with _lock:
        conn = _get_conn()
        cur = conn.execute(f"DELETE FROM {table} WHERE key = ?", (key,))
        conn.commit()
        return cur.rowcount


def refresh_media(imdb_id):
    """
    Force le refresh d'un média : supprime ses recherches et métadonnées en
    cache. Le prochain fetch Stremio ré-interrogera les trackers et TMDB.
    Retourne le nombre d'entrées supprimées.
    """
    imdb_id = (imdb_id or "").strip()
    if not imdb_id:
        return {"searches": 0, "meta": 0}
    like = f"%{imdb_id}%"
    with _lock:
        conn = _get_conn()
        s = conn.execute("DELETE FROM searches WHERE key LIKE ?", (like,)).rowcount
        m = conn.execute("DELETE FROM meta WHERE key LIKE ?", (like,)).rowcount
        conn.commit()
    return {"searches": s, "meta": m}


def flush(table):
    """Vide une table (searches|availability|meta|links|all)."""
    targets = ["searches", "availability", "meta", "links"] if table == "all" else [table]
    deleted = {}
    with _lock:
        conn = _get_conn()
        for t in targets:
            if t in ("searches", "availability", "meta", "links"):
                deleted[t] = conn.execute(f"DELETE FROM {t}").rowcount
        conn.commit()
    return deleted