# Licensed under a 3-clause BSD style license - see LICENSE.rst

import pytest
import numpy as np
from numpy.testing import assert_equal, assert_allclose

from astropy.stats.jackknife import jackknife_resampling, jackknife_stats
from astropy.utils.compat.optional_deps import HAS_SCIPY  # noqa


def test_jackknife_resampling():
    data = np.array([1, 2, 3, 4])
    answer = np.array([[2, 3, 4], [1, 3, 4], [1, 2, 4], [1, 2, 3]])
    assert_equal(answer, jackknife_resampling(data))


# test jackknife stats, except confidence interval
@pytest.mark.skipif('not HAS_SCIPY')
def test_jackknife_stats():
    # Test from the third example of Ref.[3]
    data = np.array((115, 170, 142, 138, 280, 470, 480, 141, 390))
    # true estimate, bias, and std_err
    answer = (258.4444, 0.0, 50.25936)
    assert_allclose(answer, jackknife_stats(data, np.mean)[0:3], atol=1e-4)


# test jackknife stats, including confidence intervals
@pytest.mark.skipif('not HAS_SCIPY')
def test_jackknife_stats_conf_interval():
    # Test from the first example of Ref.[3]
    data = np.array([48, 42, 36, 33, 20, 16, 29, 39, 42, 38, 42, 36, 20, 15,
                    42, 33, 22, 20, 41, 43, 45, 34, 14, 22, 6, 7, 0, 15, 33,
                    34, 28, 29, 34, 41, 4, 13, 32, 38, 24, 25, 47, 27, 41, 41,
                    24, 28, 26, 14, 30, 28, 41, 40])
    data = np.reshape(data, (-1, 2))
    data = data[:, 1]

    # true estimate, bias, and std_err
    answer = (113.7862, -4.376391, 22.26572)

    # calculate the mle of the variance (biased estimator!)
    def mle_var(x): return np.sum((x - np.mean(x))*(x - np.mean(x)))/len(x)

    assert_allclose(answer, jackknife_stats(data, mle_var, 0.95)[0:3],
                    atol=1e-4)

    # test confidence interval
    answer = np.array((70.14615, 157.42616))
    assert_allclose(answer, jackknife_stats(data, mle_var, 0.95)[3], atol=1e-4)


def test_jackknife_stats_exceptions():
    with pytest.raises(ValueError):
        jackknife_stats(np.arange(2), np.mean, confidence_level=42)