"""Linear SVM classifier - single-stage binary cyberbullying detection.

Uses ``LinearSVC`` wrapped in ``CalibratedClassifierCV(method='sigmoid', cv=5)`` so
that ``predict_proba`` returns proper probabilities (Platt scaling). RBF kernel
is deferred - see README.md § Models.
"""

from __future__ import annotations

import logging
from pathlib import Path

import joblib
import numpy as np
from sklearn.calibration import CalibratedClassifierCV
from sklearn.svm import LinearSVC

logger = logging.getLogger(__name__)


class SVMModel:
    """Linear SVM with Platt-calibrated probabilities."""

    # Target hyperparams to sweep during tuning:
    #   C ∈ [0.1, 1, 10]   (RBF + gamma sweep deferred)
    def __init__(
        self,
        C: float = 1.0,
        class_weight: str = "balanced",
        random_state: int = 42,
        max_iter: int = 10000,
        calibration_cv: int = 5,
    ) -> None:
        base = LinearSVC(
            C=C,
            class_weight=class_weight,
            random_state=random_state,
            max_iter=max_iter,
        )
        self.model = CalibratedClassifierCV(base, method="sigmoid", cv=calibration_cv)

    def fit(self, X_train, y_train, X_val=None, y_val=None) -> None:
        """Fit on TF-IDF features. ``X_val``/``y_val`` accepted for API parity but unused."""
        self.model.fit(X_train, y_train)
        logger.info("LinearSVC (calibrated) fitted on %d samples", X_train.shape[0])

    def predict(self, X) -> np.ndarray:
        """Predict 0/1 labels."""
        return self.model.predict(X)

    def predict_proba(self, X) -> np.ndarray:
        """Predict Platt-scaled class probabilities; shape ``(n, 2)``."""
        return self.model.predict_proba(X)

    def save(self, path: Path) -> None:
        """Persist the fitted estimator to ``path`` via joblib."""
        joblib.dump(self.model, str(path))

    @classmethod
    def load(cls, path: Path) -> "SVMModel":
        """Load a previously saved estimator and return a fresh wrapper."""
        instance = cls.__new__(cls)
        instance.model = joblib.load(str(path))
        return instance