from __future__ import annotations

import os
import shutil
import time
from dataclasses import dataclass
from datetime import datetime
from pathlib import Path

from azure.core.exceptions import ResourceNotFoundError
from azure.identity import ClientSecretCredential, DefaultAzureCredential
from azure.storage.filedatalake import DataLakeServiceClient

from racing_reports import vendor_env
from racing_reports.config import DEFAULT_SETTINGS, Settings
from racing_reports.utils import slugify


# Cache (a nivel módulo, persiste entre requests) del mapa de preprocessed en Azure.
_PRE_MAP_TTL = 1800.0
_PRE_MAP_CACHE: dict[str, object] = {"ts": 0.0, "value": None}


def _sas_expired(sas: str) -> bool:
    """True si el SAS tiene un 'se' (expiry) ya vencido. Si no se puede parsear,
    asume válido (False) para no romper tokens sin ese campo."""
    from datetime import datetime, timezone
    from urllib.parse import parse_qs

    try:
        se = parse_qs(sas.lstrip("?")).get("se", [None])[0]
        if not se:
            return False
        exp = datetime.fromisoformat(se.replace("Z", "+00:00"))
        return exp < datetime.now(timezone.utc)
    except Exception:
        return False


def _azure_credential():
    """Orden de auth: SAS token (vigente) > Service Principal > DefaultAzureCredential.

    - ``AZURE_SAS_TOKEN``: SAS de Storage Explorer/portal. Si está **vencido** se
      ignora (parseando su ``se=``), para que NO bloquee el fallback al Service
      Principal — error frecuente cuando el SAS dura poco.
    - Service Principal (3 env vars): "production-grade", no vence tan seguido.
    - DefaultAzureCredential: en local con ``az login``.
    """
    sas = os.getenv("AZURE_SAS_TOKEN")
    if sas and not _sas_expired(sas):
        # Las query strings de SAS pueden venir con o sin '?' inicial.
        return sas.lstrip("?")
    tenant = os.getenv("AZURE_TENANT_ID")
    client_id = os.getenv("AZURE_CLIENT_ID")
    client_secret = os.getenv("AZURE_CLIENT_SECRET")
    if tenant and client_id and client_secret:
        return ClientSecretCredential(
            tenant_id=tenant, client_id=client_id, client_secret=client_secret
        )
    return DefaultAzureCredential()


@dataclass
class DataStore:
    settings: Settings = DEFAULT_SETTINGS

    def __post_init__(self) -> None:
        self.settings.preprocessed_cache_dir.mkdir(parents=True, exist_ok=True)
        self.settings.artifacts_cache_dir.mkdir(parents=True, exist_ok=True)

    # ── ubicaciones en cache ──────────────────────────────────────────────
    def _csv_path(self, league: str, season: str, name: str) -> Path:
        return (
            self.settings.preprocessed_cache_dir
            / slugify(league)
            / season
            / name
        )

    def preprocessed_path(self, league: str, season: str) -> Path:
        return self._csv_path(league, season, self._preprocessed_filename(league, season))

    def labeled_path(self, league: str, season: str) -> Path:
        return self._csv_path(league, season, self.settings.labeled_preprocessed_name)

    def artifact_path(self, relative_path: str) -> Path:
        return self.settings.artifacts_cache_dir / relative_path

    # ── modo local (sin Azure, para dev/test) ─────────────────────────────
    @property
    def local_data_dir(self) -> Path | None:
        raw = self.settings.local_data_dir
        return Path(raw).expanduser() if raw else None

    # ── requires (sin descargar) ──────────────────────────────────────────
    def require_preprocessed(self, league: str, season: str) -> Path:
        local = self.local_data_dir
        if local is not None:
            cand = local / self.settings.preprocessed_name
            if cand.exists():
                return cand
        path = self.preprocessed_path(league, season)
        if not path.exists():
            raise FileNotFoundError(
                f"No existe preprocessed cacheado para {league} {season}: {path}. "
                f"Corré: racing-reports sync --league {league!r} --season {season}"
            )
        return path

    def require_labeled(self, league: str, season: str) -> Path:
        """Devuelve el labeled si existe; si no, cae al preprocessed base.

        El preprocessed base ya trae ``phaseLabel`` (columnas ``w_*`` y
        ``label_after_rules`` se derivan en el reporte). Mantenemos el labeled
        como override por compatibilidad con flujos viejos.
        """
        local = self.local_data_dir
        if local is not None:
            cand = local / self.settings.labeled_preprocessed_name
            if cand.exists():
                return cand
        path = self.labeled_path(league, season)
        if path.exists():
            return path
        return self.require_preprocessed(league, season)

    # ── sync desde Azure ──────────────────────────────────────────────────
    def _filesystem_client(self):
        service = DataLakeServiceClient(
            account_url=self.settings.account_url, credential=_azure_credential()
        )
        return service.get_file_system_client(self.settings.azure_filesystem)

    def available_preprocessed_map(self) -> dict[str, dict[str, str]] | None:
        """{liga: {temporada: nombre_archivo}} de los preprocessed subidos a Azure.

        El nombre del preprocessed es por liga (``preprocessed_<ABBR>_<temporada>.csv``),
        así que lo descubrimos listando Azure en vez de asumir un nombre fijo.
        Cacheado a nivel módulo (TTL). Devuelve None si no se puede listar
        (sin credenciales / dev local / error): el caller lo interpreta como
        "desconocido, no filtrar".
        """
        if self.local_data_dir is not None:
            return None
        now = time.time()
        if _PRE_MAP_CACHE["value"] is not None and now - float(_PRE_MAP_CACHE["ts"]) < _PRE_MAP_TTL:
            return _PRE_MAP_CACHE["value"]  # type: ignore[return-value]
        root = self.settings.azure_preprocessed_root.strip("/")
        try:
            fs = self._filesystem_client()
            found: dict[str, dict[str, str]] = {}
            for p in fs.get_paths(path=root, recursive=True):
                if getattr(p, "is_directory", False):
                    continue
                path_name = getattr(p, "name", "") or ""
                rel = path_name[len(root):].lstrip("/") if path_name.startswith(root) else path_name
                parts = rel.split("/")
                if len(parts) < 3:
                    continue
                fname = parts[-1]
                if not (fname.startswith("preprocessed_") and fname.endswith(".csv")):
                    continue
                if "etiquetado" in fname:  # variante labeled, no la base
                    continue
                found.setdefault(parts[-3], {})[parts[-2]] = fname
            result = found or None
        except Exception:
            result = None
        _PRE_MAP_CACHE["value"] = result
        _PRE_MAP_CACHE["ts"] = now
        return result

    def available_league_seasons(self) -> dict[str, list[str]] | None:
        """{liga: [temporadas]} con preprocessed en Azure (None si desconocido)."""
        m = self.available_preprocessed_map()
        if not m:
            return None
        return {lg: sorted(seasons.keys(), reverse=True) for lg, seasons in sorted(m.items())}

    def _preprocessed_filename(self, league: str, season: str) -> str:
        """Nombre real del preprocessed para esa liga/temporada (fallback al fijo)."""
        m = self.available_preprocessed_map()
        if m:
            fname = m.get(league, {}).get(str(season))
            if fname:
                return fname
        return self.settings.preprocessed_name

    def _sync_named_csv(self, league: str, season: str, name: str, force: bool) -> Path:
        local = self._csv_path(league, season, name)
        if local.exists() and not force:
            return local
        if self.local_data_dir is not None:
            src = self.local_data_dir / name
            if src.exists():
                local.parent.mkdir(parents=True, exist_ok=True)
                if local.resolve() != src.resolve():
                    shutil.copy2(src, local)
                return local
        remote = f"{self.settings.azure_preprocessed_root}/{league}/{season}/{name}"
        fs = self._filesystem_client()
        self._download(fs, remote, local)
        return local

    def sync_preprocessed(self, league: str, season: str, force: bool = False) -> Path:
        return self._sync_named_csv(
            league, season, self._preprocessed_filename(league, season), force
        )

    def sync_labeled(self, league: str, season: str, force: bool = False) -> Path:
        """Sincroniza el labeled si existe en Azure; si no, cae al preprocessed base.

        El base ya trae ``phaseLabel`` para derivar bloques en runtime.
        """
        try:
            return self._sync_named_csv(
                league, season, self.settings.labeled_preprocessed_name, force
            )
        except ResourceNotFoundError:
            return self.sync_preprocessed(league, season, force=force)

    def sync_artifact(self, relative_path: str, force: bool = False) -> Path:
        # 1) bundleado en el repo (CORE artifacts viven acá para no requerir Azure)
        if not force:
            bundled = vendor_env.bundled_artifact_path(relative_path)
            if bundled is not None:
                return bundled
        # 2) cache local previo
        local = self.artifact_path(relative_path)
        if local.exists() and not force:
            return local
        # 3) carpeta override para dev (RACING_LOCAL_DATA_DIR)
        if self.local_data_dir is not None:
            src = self.local_data_dir / relative_path
            if src.exists():
                local.parent.mkdir(parents=True, exist_ok=True)
                shutil.copy2(src, local)
                return local
        # 4) descargar de Azure
        remote = f"{self.settings.azure_report_artifacts_root}/{relative_path}".replace(
            "//", "/"
        )
        fs = self._filesystem_client()
        self._download(fs, remote, local)
        return local

    def sync_artifacts(self, force: bool = False) -> list[Path]:
        return [
            self.sync_artifact(rel, force=force)
            for rel in vendor_env.ARTIFACT_RELATIVE_PATHS
        ]

    # ── preparación de vendor/data para los scripts reales ────────────────
    def prepare_vendor(
        self, report_types: list[str], league: str, season: str, force: bool = False
    ) -> dict[str, Path]:
        """Asegura artifacts + CSV necesarios y los linkea en vendor/data.

        Devuelve paths resueltos: preprocessed_csv, labeled_csv, matches_csv.
        Descarga perezosa según el tipo de reporte pedido.
        """
        vendor_env.add_scripts_to_path()
        # artifacts chicos (siempre baratos) + linkeo a vendor/data
        for rel in vendor_env.ARTIFACT_RELATIVE_PATHS:
            src = self.sync_artifact(rel, force=force)
            vendor_env.link_artifact(rel, src)

        resolved: dict[str, Path] = {
            "matches_csv": vendor_env.DATA_DIR / "matches.csv",
        }
        need_pre = any(r in report_types for r in ("post_match", "triple_multitag", "triple_centros", "recuperaciones"))
        need_labeled = "block_zscore" in report_types
        if need_pre:
            resolved["preprocessed_csv"] = self.sync_preprocessed(
                league, season, force=force
            )
        if need_labeled:
            resolved["labeled_csv"] = self.sync_labeled(league, season, force=force)
        return resolved

    @staticmethod
    def _download(fs, remote_path: str, local_path: Path) -> None:
        local_path.parent.mkdir(parents=True, exist_ok=True)
        downloader = fs.get_file_client(remote_path).download_file()
        tmp = local_path.with_suffix(local_path.suffix + ".part")
        with tmp.open("wb") as handle:
            downloader.readinto(handle)
        tmp.replace(local_path)

    @staticmethod
    def _latest_csv(fs, remote_dir: str) -> str:
        paths = [
            p
            for p in fs.get_paths(path=remote_dir, recursive=False)
            if not p.is_directory and p.name.endswith(".csv")
        ]
        if not paths:
            raise FileNotFoundError(f"No se encontraron CSV en Azure: {remote_dir}")
        paths.sort(key=lambda p: p.last_modified or datetime.min, reverse=True)
        return paths[0].name