from math import isclose

import numpy


def test_var_deterministic():
    x = numpy.array([-2.0, 2.0])
    expected = 4.0
    assert isclose(numpy.var(x), expected)


def test_var_fuzzing():
    rand_state = numpy.random.RandomState(8393)

    N, D = 100000, 5
    # Goal variances: [0.1 ,  0.45,  0.8 ,  1.15,  1.5]
    expected = numpy.linspace(0.1, 1.5, D)

    # Generate random, D-dimensional data
    x = rand_state.randn(N, D) * numpy.sqrt(expected)
    variance = numpy.var(x, axis=0)
    numpy.testing.assert_allclose(variance, expected, rtol=1e-2)