/* -------------------------------------------------------------------------- *
 *                       Simbody(tm): SimTKcommon                             *
 * -------------------------------------------------------------------------- *
 * This is part of the SimTK biosimulation toolkit originating from           *
 * Simbios, the NIH National Center for Physics-Based Simulation of           *
 * Biological Structures at Stanford, funded under the NIH Roadmap for        *
 * Medical Research, grant U54 GM072970. See https://simtk.org/home/simbody.  *
 *                                                                            *
 * Portions copyright (c) 2007-12 Stanford University and the Authors.        *
 * Authors: Peter Eastman                                                     *
 * Contributors:                                                              *
 *                                                                            *
 * Licensed under the Apache License, Version 2.0 (the "License"); you may    *
 * not use this file except in compliance with the License. You may obtain a  *
 * copy of the License at http://www.apache.org/licenses/LICENSE-2.0.         *
 *                                                                            *
 * Unless required by applicable law or agreed to in writing, software        *
 * distributed under the License is distributed on an "AS IS" BASIS,          *
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.   *
 * See the License for the specific language governing permissions and        *
 * limitations under the License.                                             *
 * -------------------------------------------------------------------------- */

#include "SimTKcommon.h"

#include <iostream>

using std::cout;
using std::endl;
using std::sqrt;
using namespace SimTK;

/**
 * Given the number of values expected and found in a set of bins, verify that the distribution is correct.
 */

void verifyDistribution(int expected[], int found[], int bins) {
    for (int i = 0; i < bins; ++i) {
        Real dev = sqrt((Real) expected[i]);
        SimTK_TEST(found[i] >= expected[i]-4*dev && found[i] <= expected[i]+4*dev)
    }
}

/**
 * Given a set of Reals, verify that they satisfy a uniform distribution between 0 and 1.
 */

void verifyUniformDistribution(Real min, Real max, Real value[], int length) {
    int expected[10], found[10];
    for (int i = 0; i < 10; ++i) {
        expected[i] = length/10;
        found[i] = 0;
    }
    for (int i = 0; i < length; ++i) {
        SimTK_TEST(value[i] >= min)
        SimTK_TEST(value[i] < max)
        int index = (int) ((value[i]-min)*10/(max-min));
        found[index]++;
    }
    verifyDistribution(expected, found, 10);
}

/**
 * Given a set of ints, verify that they satisfy a uniform distribution between 0 and max.
 */

void verifyUniformDistribution(int min, int max, int value[], int length) {
    int range = max-min;
    int* expected = new int[range];
    int* found = new int[range];
    for (int i = 0; i < range; ++i) {
        expected[i] = length/range;
        found[i] = 0;
    }
    for (int i = 0; i < length; ++i) {
        SimTK_TEST(value[i] >= min)
        SimTK_TEST(value[i] < max)
        found[value[i]-min]++;
    }
    verifyDistribution(expected, found, range);
    delete[] expected;
    delete[] found;
}

/**
 * Given a set of values, verify that they satisfy a Gaussian distribution.
 */

void verifyGaussianDistribution(Real mean, Real stddev, Real value[], int length) {
    int expected[6], found[6];
    expected[0] = expected[5] = (int) (0.0228*length);
    expected[1] = expected[4] = (int) (0.1587*length-expected[0]);
    expected[2] = expected[3] = (int) (0.5*length-expected[1]);
    for (int i = 0; i < 6; ++i)
        found[i] = 0;
    for (int i = 0; i < length; ++i) {
        Real val = (value[i]-mean)/stddev;
        if (val < -2)
            found[0]++;
        else if (val < -1)
            found[1]++;
        else if (val < 0)
            found[2]++;
        else if (val < 1)
            found[3]++;
        else if (val < 2)
            found[4]++;
        else
            found[5]++;
    }
    verifyDistribution(expected, found, 6);
}

void testUniform() {
    Random::Uniform rand;
    SimTK_TEST(rand.getMin() == 0.0)
    SimTK_TEST(rand.getMax() == 1.0)

    // Try generating a bunch of random numbers, and make sure they are distributed uniformly between 0 and 1.
    
    Real value[2001];
    value[2000] = 123.4;
    rand.setSeed(1);
    for (int i = 0; i < 2000; ++i)
        value[i] = rand.getValue();
    verifyUniformDistribution(0.0, 1.0, value, 2000);
    
    // Reset the random number generator, and make sure it produces the same values again.
    
    rand.setSeed(1);
    for (int i = 0; i < 2000; ++i)
        SimTK_TEST(value[i] == rand.getValue())
    
    // Now try asking for a whole array at a time, and verify that it still gives the same results.
    
    Real value2[2001];
    value2[2000] = 567.8;
    rand.setSeed(1);
    rand.fillArray(value2, 2000);
    for (int i = 0; i < 2000; ++i)
        SimTK_TEST(value[i] == value2[i])
    
    // Set the seed to a different value, and verify that the results are different.
    
    rand.setSeed(2);
    rand.fillArray(value2, 2000);
    for (int i = 0; i < 2000; ++i)
        SimTK_TEST(value[i] != value2[i])
    
    // Change the range and test the distribution.
    
    rand.setMin(5.0);
    rand.setMax(20.0);
    SimTK_TEST(rand.getMin() == 5.0)
    SimTK_TEST(rand.getMax() == 20.0)
    rand.fillArray(value2, 2000);
    verifyUniformDistribution(5.0, 20.0, value2, 2000);
    
    // Try generating uniform integers.
    
    int value3[2001];
    value3[2000] = -99;
    rand.setSeed(3);
    for (int i = 0; i < 2000; ++i)
        value3[i] = rand.getIntValue();
    verifyUniformDistribution(5, 20, value3, 2000);

    // Verify that if we do not explicitly set the seed, every Random object is initialized with a different seed.
    
    Random::Uniform rand1, rand2;
    rand1.fillArray(value, 2000);
    rand2.fillArray(value2, 2000);
    for (int i = 0; i < 2000; ++i)
        SimTK_TEST(value[i] != value2[i])

    // Make sure none of the above operations has overwritten the final array
    // element. On i386 the 32 bit representation of the non-integer values
    // might not be exact so test to a tolerance.
    
    SimTK_TEST_EQ(value[2000], 123.4)
    SimTK_TEST_EQ(value2[2000], 567.8)
    SimTK_TEST(value3[2000] == -99)
}

void testGaussian() {
    Random::Gaussian rand;
    SimTK_TEST(rand.getMean() == 0.0)
    SimTK_TEST(rand.getStdDev() == 1.0)
    
    // Try generating a bunch of Gaussian random numbers, and check the distribution.
    
    Real value[2001];
    value[2000] = 123.4;
    rand.setSeed(1);
    for (int i = 0; i < 2000; ++i)
        value[i] = rand.getValue();
    verifyGaussianDistribution(0.0, 1.0, value, 2000);
    
    // Try getting a whole array at a time.
    
    Real value2[2001];
    value2[2000] = 567.8;
    rand.setSeed(1);
    rand.fillArray(value2, 2000);
    for (int i = 0; i < 2000; ++i)
        SimTK_TEST(value[i] == value2[i])
    
    // Change the parameters and test the distribution.
    
    rand.setMean(10.0);
    rand.setStdDev(7.0);
    SimTK_TEST(rand.getMean() == 10.0)
    SimTK_TEST(rand.getStdDev() == 7.0)
    rand.fillArray(value2, 2000);
    verifyGaussianDistribution(10.0, 7.0, value2, 2000);

    // Make sure none of the above operations has overwritten the final array
    // element. On i386 the 32 bit representation of the non-integer values
    // might not be exact so test to a tolerance.

    SimTK_TEST_EQ(value[2000], 123.4)
    SimTK_TEST_EQ(value2[2000], 567.8)
}

int main() {
    try {
        testUniform();
        testGaussian();
    } catch(const std::exception& e) {
        cout << "exception: " << e.what() << endl;
        return 1;
    }
    cout << "Done" << endl;
    return 0;
}