"""Prediction model adapters used through a small dependency-injection boundary."""

from __future__ import annotations

from dataclasses import dataclass
from typing import Protocol

import pandas as pd

from final_pipeline import (
    DiagnosticModels,
    predict_test,
    train_models,
    validate_submission,
)

from .errors import InferenceError


@dataclass(frozen=True)
class PredictionBundle:
    submission: pd.DataFrame
    diagnostics: pd.DataFrame


class PredictionModel(Protocol):
    def predict(self, frame: pd.DataFrame) -> PredictionBundle: ...


class SklearnPredictionModel:
    """Thin adapter around the frozen, validated competition pipeline."""

    def __init__(self, models: DiagnosticModels) -> None:
        self._models = models

    @classmethod
    def train(
        cls,
        reference_frame: pd.DataFrame,
        *,
        random_state: int = 42,
        n_jobs: int = -1,
    ) -> "SklearnPredictionModel":
        try:
            models = train_models(
                reference_frame,
                random_state=random_state,
                n_jobs=n_jobs,
            )
        except Exception as exc:
            raise InferenceError(
                "Nie udało się przygotować modelu referencyjnego.",
                hint="Sprawdź kompletność val.csv oraz zgodność wersji zależności.",
            ) from exc
        return cls(models)

    def predict(self, frame: pd.DataFrame) -> PredictionBundle:
        try:
            submission, diagnostics = predict_test(self._models, frame)
            validate_submission(submission, frame)
        except Exception as exc:
            raise InferenceError(
                "Model odrzucił pomiary podczas predykcji.",
                hint="Zweryfikuj kompletność silników oraz brak nietypowych wartości w widmie.",
            ) from exc
        return PredictionBundle(submission, diagnostics)


class PrecomputedPredictionModel:
    """Read-only demo fallback; accepts only the exact precomputed key set."""

    def __init__(self, submission: pd.DataFrame, diagnostics: pd.DataFrame) -> None:
        self._submission = submission.reset_index(drop=True).copy()
        self._diagnostics = diagnostics.reset_index(drop=True).copy()

    def predict(self, frame: pd.DataFrame) -> PredictionBundle:
        keys = ["engine_id", "cylinder"]
        if not frame[keys].reset_index(drop=True).equals(self._submission[keys]):
            raise InferenceError(
                "Tryb awaryjny obsługuje wyłącznie dołączony zestaw demonstracyjny.",
                hint="Przywróć val.csv i uruchom ponownie aplikację, aby diagnozować własne pliki.",
            )
        return PredictionBundle(
            self._submission.copy(),
            self._diagnostics.copy(),
        )