import numpy
import numpy.random

import csb.test as test

from csb.statistics import Cumulative
from csb.statistics import kurtosis, skewness, autocorrelation


@test.functional
class TestStatFunction(test.Case):


    def testCumulative(self):
        from scipy.stats import norm
        
        x = numpy.linspace(-5., 5., 200)
        samples = numpy.random.normal(size=100000)
        cumula = Cumulative(samples)
        c = cumula(x)
        
        cx = norm.cdf(x)
        for i in range(199):
            self.assertAlmostEqual(cx[i], c[i], delta=1e-2)
        

    def testKurtosis(self):
        samples = numpy.random.normal(size=100000)
        self.assertAlmostEqual(kurtosis(samples), 0., delta=1e-1)

        samples = numpy.random.uniform(-2., 2., size=100000)
        self.assertAlmostEqual(kurtosis(samples), -1.2, delta=1e-1)


    def testSkewness(self):
        samples = numpy.random.gamma(2., 0.5, size=100000)
        self.assertAlmostEqual(skewness(samples), 2. / numpy.sqrt(2.), delta=1e-1)

    def testAutorcorrelation(self):
        x = numpy.random.normal(size=1000) + numpy.sin(numpy.linspace(0., 2 * numpy.pi, 1000))
        n = 10
        ac = autocorrelation(x, n)
        self.assertAlmostEqual(ac[0], 1., delta=1e-1)
        
    def testEntropy(self):
        pass

    def testCircvar(self):
        pass

    def testCircmean(self):
        pass