# pyright: basic

from __future__ import annotations

import numpy as np

DEFAULT_BACKENDS = ["ensemble", "bayesian"]
DEFAULT_POSITIVE_CLASS_INDEX = 0
DEFAULT_CALIBRATION_BINS = 10
DEFAULT_BAYESIAN_MC_PASSES = 20
DEFAULT_DECISION_THRESHOLD = 0.5

THRESHOLD_START = 0.5
THRESHOLD_STOP = 0.95
THRESHOLD_STEP = 0.05

NOISE_FACTOR_MIN = 0.0
NOISE_FACTOR_MAX = 7.0
NOISE_FACTOR_COUNT = 21

UNCERTAINTY_CUTOFF_COUNT = 21


def threshold_grid() -> np.ndarray:
    if THRESHOLD_STEP <= 0:
        raise ValueError("THRESHOLD_STEP must be > 0")
    if THRESHOLD_STOP < THRESHOLD_START:
        raise ValueError("THRESHOLD_STOP must be >= THRESHOLD_START")

    n = int(round((THRESHOLD_STOP - THRESHOLD_START) / THRESHOLD_STEP))
    return np.linspace(THRESHOLD_START, THRESHOLD_STOP, num=n + 1, dtype=float)


def uncertainty_cutoff_percentiles() -> np.ndarray:
    return np.linspace(0.0, 100.0, num=UNCERTAINTY_CUTOFF_COUNT, dtype=float)


def noise_factor_grid() -> list[float]:
    factors = np.linspace(
        NOISE_FACTOR_MIN,
        NOISE_FACTOR_MAX,
        num=NOISE_FACTOR_COUNT,
        dtype=float,
    )
    return [float(v) for v in factors]