/*
    * Licensed to the Apache Software Foundation (ASF) under one or more
    * contributor license agreements.  See the NOTICE file distributed with
    * this work for additional information regarding copyright ownership.
    * The ASF licenses this file to You 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.
   */
  package org.apache.commons.rng;
  
  import java.math.BigDecimal;
  import java.math.MathContext;
  import java.util.ArrayList;
  import java.util.Arrays;
  import java.util.List;
  import java.util.SplittableRandom;
  import java.util.concurrent.ThreadLocalRandom;
  import java.util.function.LongSupplier;
  import java.util.stream.Collectors;
  import java.util.stream.Stream;
  import org.junit.jupiter.api.Assertions;
  import org.junit.jupiter.api.Test;
  import org.junit.jupiter.params.ParameterizedTest;
  import org.junit.jupiter.params.provider.Arguments;
  import org.junit.jupiter.params.provider.CsvSource;
  import org.junit.jupiter.params.provider.MethodSource;
  import org.junit.jupiter.params.provider.ValueSource;
  
  /**
   * Tests for default method implementations in {@link UniformRandomProvider}.
   *
   * <p>This class verifies all exception conditions for the range methods.
   * Single value generation methods are asserted using a test of uniformity
   * from multiple samples.
   *
   * <p>Stream methods are tested {@link UniformRandomProviderStreamTest}.
   */
  class UniformRandomProviderTest {
      /** Sample size for statistical uniformity tests. */
      private static final int SAMPLE_SIZE = 1000;
      /** Sample size for statistical uniformity tests as a BigDecimal. */
      private static final BigDecimal SAMPLE_SIZE_BD = BigDecimal.valueOf(SAMPLE_SIZE);
      /** Relative error used to verify the sum of expected frequencies matches the sample size. */
      private static final double RELATIVE_ERROR = Math.ulp(1.0) * 2;
  
      /**
       * Dummy class for checking the behavior of the UniformRandomProvider.
       */
      private static class DummyGenerator implements UniformRandomProvider {
          /** An instance. */
          static final DummyGenerator INSTANCE = new DummyGenerator();
  
          @Override
          public long nextLong() {
              throw new UnsupportedOperationException("The nextLong method should not be invoked");
          }
      }
  
      static int[] invalidNextIntBound() {
          return new int[] {0, -1, Integer.MIN_VALUE};
      }
  
      static Stream<Arguments> invalidNextIntOriginBound() {
          return Stream.of(
              Arguments.of(1, 1),
              Arguments.of(2, 1),
              Arguments.of(-1, -1),
              Arguments.of(-1, -2)
          );
      }
  
      static long[] invalidNextLongBound() {
          return new long[] {0, -1, Long.MIN_VALUE};
      }
  
      static Stream<Arguments> invalidNextLongOriginBound() {
          return Stream.of(
              Arguments.of(1L, 1L),
              Arguments.of(2L, 1L),
              Arguments.of(-1L, -1L),
              Arguments.of(-1L, -2L)
          );
      }
  
      static float[] invalidNextFloatBound() {
          return new float[] {0, -1, -Float.MAX_VALUE, Float.NaN, Float.POSITIVE_INFINITY, Float.NEGATIVE_INFINITY};
      }
  
      static Stream<Arguments> invalidNextFloatOriginBound() {
          return Stream.of(
              Arguments.of(1f, 1f),
             Arguments.of(2f, 1f),
             Arguments.of(-1f, -1f),
             Arguments.of(-1f, -2f),
             Arguments.of(Float.NEGATIVE_INFINITY, 0f),
             Arguments.of(0f, Float.POSITIVE_INFINITY),
             Arguments.of(0f, Float.NaN),
             Arguments.of(Float.NaN, 1f)
         );
     }
 
     static double[] invalidNextDoubleBound() {
         return new double[] {0, -1, -Double.MAX_VALUE, Double.NaN, Double.POSITIVE_INFINITY, Double.NEGATIVE_INFINITY};
     }
 
     static Stream<Arguments> invalidNextDoubleOriginBound() {
         return Stream.of(
             Arguments.of(1, 1),
             Arguments.of(2, 1),
             Arguments.of(-1, -1),
             Arguments.of(-1, -2),
             Arguments.of(Double.NEGATIVE_INFINITY, 0),
             Arguments.of(0, Double.POSITIVE_INFINITY),
             Arguments.of(0, Double.NaN),
             Arguments.of(Double.NaN, 1)
         );
     }
 
     /**
      * Creates a functional random generator by implementing the
      * {@link UniformRandomProvider#nextLong} method with a high quality source of randomness.
      *
      * @param seed Seed for the generator.
      * @return the random generator
      */
     private static UniformRandomProvider createRNG(long seed) {
         // The algorithm for SplittableRandom with the default increment passes:
         // - Test U01 BigCrush
         // - PractRand with at least 2^42 bytes (4 TiB) of output
         return new SplittableRandom(seed)::nextLong;
     }
 
     @Test
     void testNextBytesThrows() {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(NullPointerException.class, () -> rng.nextBytes(null));
         Assertions.assertThrows(NullPointerException.class, () -> rng.nextBytes(null, 0, 1));
         // Invalid range
         final int length = 10;
         final byte[] bytes = new byte[length];
         Assertions.assertThrows(IndexOutOfBoundsException.class, () -> rng.nextBytes(bytes, -1, 1), "start < 0");
         Assertions.assertThrows(IndexOutOfBoundsException.class, () -> rng.nextBytes(bytes, length, 1), "start >= length");
         Assertions.assertThrows(IndexOutOfBoundsException.class, () -> rng.nextBytes(bytes, 0, -1), "len < 0");
         Assertions.assertThrows(IndexOutOfBoundsException.class, () -> rng.nextBytes(bytes, 5, 10), "start + len > length");
         Assertions.assertThrows(IndexOutOfBoundsException.class, () -> rng.nextBytes(bytes, 5, Integer.MAX_VALUE), "start + len > length, taking into account integer overflow");
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextIntBound"})
     void testNextIntBoundThrows(int bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextInt(bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextIntOriginBound"})
     void testNextIntOriginBoundThrows(int origin, int bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextInt(origin, bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextLongBound"})
     void testNextLongBoundThrows(long bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextLong(bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextLongOriginBound"})
     void testNextLongOriginBoundThrows(long origin, long bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextLong(origin, bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextFloatBound"})
     void testNextFloatBoundThrows(float bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextFloat(bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextFloatOriginBound"})
     void testNextFloatOriginBoundThrows(float origin, float bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextFloat(origin, bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextDoubleBound"})
     void testNextDoubleBoundThrows(double bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextDouble(bound));
     }
 
     @ParameterizedTest
     @MethodSource(value = {"invalidNextDoubleOriginBound"})
     void testNextDoubleOriginBoundThrows(double origin, double bound) {
         final UniformRandomProvider rng = DummyGenerator.INSTANCE;
         Assertions.assertThrows(IllegalArgumentException.class, () -> rng.nextDouble(origin, bound));
     }
 
     @Test
     void testNextFloatExtremes() {
         final UniformRandomProvider rng = new DummyGenerator() {
             private int i;
             private final int[] values = {0, -1, 1 << 8};
             @Override
             public int nextInt() {
                 return values[i++];
             }
         };
         Assertions.assertEquals(0, rng.nextFloat(), "Expected zero bits to return 0");
         Assertions.assertEquals(Math.nextDown(1f), rng.nextFloat(), "Expected all bits to return ~1");
         // Assumes the value is created from the high 24 bits of nextInt
         Assertions.assertEquals(1f - Math.nextDown(1f), rng.nextFloat(), "Expected 1 bit (shifted) to return ~0");
     }
 
     @ParameterizedTest
     @ValueSource(floats = {Float.MIN_NORMAL, Float.MIN_NORMAL / 2})
     void testNextFloatBoundRounding(float bound) {
         // This method will be used
         final UniformRandomProvider rng = new DummyGenerator() {
             @Override
             public float nextFloat() {
                 return Math.nextDown(1.0f);
             }
         };
         Assertions.assertEquals(bound, rng.nextFloat() * bound, "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextFloat(bound), "Result was not correctly rounded down");
     }
 
     @Test
     void testNextFloatOriginBoundInfiniteRange() {
         // This method will be used
         final UniformRandomProvider rng = new DummyGenerator() {
             private int i;
             private final float[] values = {0, 0.25f, 0.5f, 0.75f, 1};
             @Override
             public float nextFloat() {
                 return values[i++];
             }
         };
         final float x = Float.MAX_VALUE;
         Assertions.assertEquals(-x, rng.nextFloat(-x, x));
         Assertions.assertEquals(-x / 2, rng.nextFloat(-x, x), Math.ulp(x / 2));
         Assertions.assertEquals(0, rng.nextFloat(-x, x));
         Assertions.assertEquals(x / 2, rng.nextFloat(-x, x), Math.ulp(x / 2));
         Assertions.assertEquals(Math.nextDown(x), rng.nextFloat(-x, x));
     }
 
     @Test
     void testNextFloatOriginBoundRounding() {
         // This method will be used
         final float v = Math.nextDown(1.0f);
         final UniformRandomProvider rng = new DummyGenerator() {
             @Override
             public float nextFloat() {
                 return v;
             }
         };
         float origin;
         float bound;
 
         // Simple case
         origin = 3.5f;
         bound = 4.5f;
         Assertions.assertEquals(bound, origin + v * (bound - origin), "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextFloat(origin, bound), "Result was not correctly rounded down");
 
         // Alternate formula:
         // requires sub-normal result for 'origin * v' to force rounding
         origin = -Float.MIN_NORMAL / 2;
         bound = Float.MIN_NORMAL / 2;
         Assertions.assertEquals(bound, origin * (1 - v) + v * bound, "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextFloat(origin, bound), "Result was not correctly rounded down");
     }
 
     @Test
     void testNextDoubleExtremes() {
         final UniformRandomProvider rng = new DummyGenerator() {
             private int i;
             private final long[] values = {0, -1, 1L << 11};
             @Override
             public long nextLong() {
                 return values[i++];
             }
         };
         Assertions.assertEquals(0, rng.nextDouble(), "Expected zero bits to return 0");
         Assertions.assertEquals(Math.nextDown(1.0), rng.nextDouble(), "Expected all bits to return ~1");
         // Assumes the value is created from the high 53 bits of nextInt
         Assertions.assertEquals(1.0 - Math.nextDown(1.0), rng.nextDouble(), "Expected 1 bit (shifted) to return ~0");
     }
 
     @ParameterizedTest
     @ValueSource(doubles = {Double.MIN_NORMAL, Double.MIN_NORMAL / 2})
     void testNextDoubleBoundRounding(double bound) {
         // This method will be used
         final UniformRandomProvider rng = new DummyGenerator() {
             @Override
             public double nextDouble() {
                 return Math.nextDown(1.0);
             }
         };
         Assertions.assertEquals(bound, rng.nextDouble() * bound, "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextDouble(bound), "Result was not correctly rounded down");
     }
 
     @Test
     void testNextDoubleOriginBoundInfiniteRange() {
         // This method will be used
         final UniformRandomProvider rng = new DummyGenerator() {
             private int i;
             private final double[] values = {0, 0.25, 0.5, 0.75, 1};
             @Override
             public double nextDouble() {
                 return values[i++];
             }
         };
         final double x = Double.MAX_VALUE;
         Assertions.assertEquals(-x, rng.nextDouble(-x, x));
         Assertions.assertEquals(-x / 2, rng.nextDouble(-x, x), Math.ulp(x / 2));
         Assertions.assertEquals(0, rng.nextDouble(-x, x));
         Assertions.assertEquals(x / 2, rng.nextDouble(-x, x), Math.ulp(x / 2));
         Assertions.assertEquals(Math.nextDown(x), rng.nextDouble(-x, x));
     }
 
     @Test
     void testNextDoubleOriginBoundRounding() {
         // This method will be used
         final double v = Math.nextDown(1.0);
         final UniformRandomProvider rng = new DummyGenerator() {
             @Override
             public double nextDouble() {
                 return v;
             }
         };
         double origin;
         double bound;
 
         // Simple case
         origin = 3.5;
         bound = 4.5;
         Assertions.assertEquals(bound, origin + v * (bound - origin), "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextDouble(origin, bound), "Result was not correctly rounded down");
 
         // Alternate formula:
         // requires sub-normal result for 'origin * v' to force rounding
         origin = -Double.MIN_NORMAL / 2;
         bound = Double.MIN_NORMAL / 2;
         Assertions.assertEquals(bound, origin * (1 - v) + v * bound, "Expected result to be rounded up");
         Assertions.assertEquals(Math.nextDown(bound), rng.nextDouble(origin, bound), "Result was not correctly rounded down");
     }
 
     @Test
     void testNextBooleanIsIntSignBit() {
         final int size = 10;
         final int[] values = ThreadLocalRandom.current().ints(size).toArray();
         final UniformRandomProvider rng = new DummyGenerator() {
             private int i;
             @Override
             public int nextInt() {
                 return values[i++];
             }
         };
         for (int i = 0; i < size; i++) {
             Assertions.assertEquals(values[i] < 0, rng.nextBoolean());
         }
     }
 
     // Statistical tests for uniform distribution
 
     // Monobit tests
 
     @ParameterizedTest
     @ValueSource(longs = {263784628482L, -2563472, -2367482842368L})
     void testNextIntMonobit(long seed) {
         final UniformRandomProvider rng = createRNG(seed);
         int bitCount = 0;
         final int n = 100;
         for (int i = 0; i < n; i++) {
             bitCount += Integer.bitCount(rng.nextInt());
         }
         final int numberOfBits = n * Integer.SIZE;
         assertMonobit(bitCount, numberOfBits);
     }
 
     @ParameterizedTest
     @ValueSource(longs = {263784628L, 253674, -23568426834L})
     void testNextDoubleMonobit(long seed) {
         final UniformRandomProvider rng = createRNG(seed);
         int bitCount = 0;
         final int n = 100;
         for (int i = 0; i < n; i++) {
             bitCount += Long.bitCount((long) (rng.nextDouble() * (1L << 53)));
         }
         final int numberOfBits = n * 53;
         assertMonobit(bitCount, numberOfBits);
     }
 
     @ParameterizedTest
     @ValueSource(longs = {265342L, -234232, -672384648284L})
     void testNextBooleanMonobit(long seed) {
         final UniformRandomProvider rng = createRNG(seed);
         int bitCount = 0;
         final int n = 1000;
         for (int i = 0; i < n; i++) {
             if (rng.nextBoolean()) {
                 bitCount++;
             }
         }
         final int numberOfBits = n;
         assertMonobit(bitCount, numberOfBits);
     }
 
     @ParameterizedTest
     @ValueSource(longs = {-1526731, 263846, 4545})
     void testNextFloatMonobit(long seed) {
         final UniformRandomProvider rng = createRNG(seed);
         int bitCount = 0;
         final int n = 100;
         for (int i = 0; i < n; i++) {
             bitCount += Integer.bitCount((int) (rng.nextFloat() * (1 << 24)));
         }
         final int numberOfBits = n * 24;
         assertMonobit(bitCount, numberOfBits);
     }
 
     @ParameterizedTest
     @CsvSource({
         "-2638468223894, 16",
         "235647674, 17",
         "-928738475, 23",
     })
     void testNextBytesMonobit(long seed, int range) {
         final UniformRandomProvider rng = createRNG(seed);
         final byte[] bytes = new byte[range];
         int bitCount = 0;
         final int n = 100;
         for (int i = 0; i < n; i++) {
             rng.nextBytes(bytes);
             for (final byte b1 : bytes) {
                 bitCount += Integer.bitCount(b1 & 0xff);
             }
         }
         final int numberOfBits = n * Byte.SIZE * range;
         assertMonobit(bitCount, numberOfBits);
     }
 
     /**
      * Assert that the number of 1 bits is approximately 50%. This is based upon a
      * fixed-step "random walk" of +1/-1 from zero.
      *
      * <p>The test is equivalent to the NIST Monobit test with a fixed p-value of
      * 0.01. The number of bits is recommended to be above 100.</p>
      *
      * @see <A
      * href="https://csrc.nist.gov/publications/detail/sp/800-22/rev-1a/final">Bassham,
      * et al (2010) NIST SP 800-22: A Statistical Test Suite for Random and
      * Pseudorandom Number Generators for Cryptographic Applications. Section
      * 2.1.</a>
      *
      * @param bitCount The bit count.
      * @param numberOfBits Number of bits.
      */
     private static void assertMonobit(int bitCount, int numberOfBits) {
         // Convert the bit count into a number of +1/-1 steps.
         final double sum = 2.0 * bitCount - numberOfBits;
         // The reference distribution is Normal with a standard deviation of sqrt(n).
         // Check the absolute position is not too far from the mean of 0 with a fixed
         // p-value of 0.01 taken from a 2-tailed Normal distribution. Computation of
         // the p-value requires the complimentary error function.
         final double absSum = Math.abs(sum);
         final double max = Math.sqrt(numberOfBits) * 2.5758293035489004;
         Assertions.assertTrue(absSum <= max,
             () -> "Walked too far astray: " + absSum + " > " + max +
                   " (test will fail randomly about 1 in 100 times)");
     }
 
     // Uniformity tests
 
     @ParameterizedTest
     @CsvSource({
         "263746283, 23, 0, 23",
         "-126536861889, 16, 0, 16",
         "617868181124, 1234, 567, 89",
         "-56788, 512, 0, 233",
         "6787535424, 512, 233, 279",
     })
     void testNextBytesUniform(long seed,
                               int length, int start, int size) {
         final UniformRandomProvider rng = createRNG(seed);
         final byte[] buffer = new byte[length];
 
         final Runnable nextMethod = start == 0 && size == length ?
                 () -> rng.nextBytes(buffer) :
                 () -> rng.nextBytes(buffer, start, size);
 
         final int last = start + size;
         Assertions.assertTrue(isUniformNextBytes(buffer, start, last, nextMethod),
                               "Expected uniform bytes");
 
         // The parts of the buffer where no values are put should be zero.
         for (int i = 0; i < start; i++) {
             Assertions.assertEquals(0, buffer[i], () -> "Filled < start: " + start);
         }
         for (int i = last; i < length; i++) {
             Assertions.assertEquals(0, buffer[i], () -> "Filled >= last: " + last);
         }
     }
 
     /**
      * Checks that the generator values can be placed into 256 bins with
      * approximately equal number of counts.
      * Test allows to select only part of the buffer for performing the
      * statistics.
      *
      * @param buffer Buffer to be filled.
      * @param first First element (included) of {@code buffer} range for
      * which statistics must be taken into account.
      * @param last Last element (excluded) of {@code buffer} range for
      * which statistics must be taken into account.
      * @param nextMethod Method that fills the given {@code buffer}.
      * @return {@code true} if the distribution is uniform.
      */
     private static boolean isUniformNextBytes(byte[] buffer,
                                               int first,
                                               int last,
                                               Runnable nextMethod) {
         final int sampleSize = 10000;
 
         // Number of possible values (do not change).
         final int byteRange = 256;
         // Chi-square critical value with 255 degrees of freedom
         // and 1% significance level.
         final double chi2CriticalValue = 310.45738821990585;
 
         // Bins.
         final long[] observed = new long[byteRange];
         final double[] expected = new double[byteRange];
 
         Arrays.fill(expected, sampleSize * (last - first) / (double) byteRange);
 
         for (int k = 0; k < sampleSize; k++) {
             nextMethod.run();
 
             for (int i = first; i < last; i++) {
                 // Convert byte to an index in [0, 255]
                 ++observed[buffer[i] & 0xff];
             }
         }
 
         // Compute chi-square.
         double chi2 = 0;
         for (int k = 0; k < byteRange; k++) {
             final double diff = observed[k] - expected[k];
             chi2 += diff * diff / expected[k];
         }
 
         // Statistics check.
         return chi2 <= chi2CriticalValue;
     }
 
     // Add range tests
 
     @ParameterizedTest
     @CsvSource({
         // No lower bound
         "2673846826, 0, 10",
         "-23658268, 0, 12",
         "263478624, 0, 31",
         "1278332, 0, 32",
         "99734765, 0, 2016128993",
         "-63485384, 0, 1834691456",
         "3876457638, 0, 869657561",
         "-126784782, 0, 1570504788",
         "2637846, 0, 2147483647",
         // Small range
         "2634682, 567576, 567586",
         "-56757798989, -1000, -100",
         "-97324785, -54656, 12",
         "23423235, -526783468, 257",
         // Large range
         "-2634682, -1073741824, 1073741824",
         "6786868132, -1263842626, 1237846372",
         "-263846723, -368268, 2147483647",
         "7352352, -2147483648, 61523457",
     })
     void testNextIntUniform(long seed, int origin, int bound) {
         final UniformRandomProvider rng = createRNG(seed);
         final LongSupplier nextMethod = origin == 0 ?
                 () -> rng.nextInt(bound) :
                 () -> rng.nextInt(origin, bound);
         checkNextInRange("nextInt", origin, bound, nextMethod);
     }
 
     @ParameterizedTest
     @CsvSource({
         // No lower bound
         "2673846826, 0, 11",
         "-23658268, 0, 19",
         "263478624, 0, 31",
         "1278332, 0, 32",
         "99734765, 0, 2326378468282368421",
         "-63485384, 0, 4872347624242243222",
         "3876457638, 0, 6263784682638866843",
         "-126784782, 0, 7256684297832668332",
         "2637846, 0, 9223372036854775807",
         // Small range
         "2634682, 567576, 567586",
         "-56757798989, -1000, -100",
         "-97324785, -54656, 12",
         "23423235, -526783468, 257",
         // Large range
         "-2634682, -4611686018427387904, 4611686018427387904",
         "6786868132, -4962836478223688590, 6723648246224929947",
         "-263846723, -368268, 9223372036854775807",
         "7352352, -9223372036854775808, 61523457",
     })
     void testNextLongUniform(long seed, long origin, long bound) {
         final UniformRandomProvider rng = createRNG(seed);
         final LongSupplier nextMethod = origin == 0 ?
                 () -> rng.nextLong(bound) :
                 () -> rng.nextLong(origin, bound);
         checkNextInRange("nextLong", origin, bound, nextMethod);
     }
 
     @ParameterizedTest
     @CsvSource({
         // Note: If the range limits are integers above 2^24 (16777216) it is not possible
         // to represent all the values with a float. This has no effect on sampling into bins
         // but should be avoided when generating integers for use in production code.
 
         // No lower bound.
         "2673846826, 0, 11",
         "-23658268, 0, 19",
         "263478624, 0, 31",
         "1278332, 0, 32",
         "99734765, 0, 1234",
         "-63485384, 0, 578",
         "3876457638, 0, 10000",
         "-126784782, 0, 2983423",
         "2637846, 0, 16777216",
         // Range
         "2634682, 567576, 567586",
         "-56757798989, -1000, -100",
         "-97324785, -54656, 12",
         "23423235, -526783468, 257",
         "-2634682, -688689797, -516827",
         "6786868132, -67, 67",
         "-263846723, -5678, 42",
         "7352352, 678687, 61523457",
     })
     void testNextFloatUniform(long seed, float origin, float bound) {
         Assertions.assertEquals((long) origin, origin, "origin");
         Assertions.assertEquals((long) bound, bound, "bound");
         final UniformRandomProvider rng = createRNG(seed);
         // Note casting as long will round towards zero.
         // If the upper bound is negative then this can create a domain error so use floor.
         final LongSupplier nextMethod = origin == 0 ?
                 () -> (long) rng.nextFloat(bound) :
                 () -> (long) Math.floor(rng.nextFloat(origin, bound));
         checkNextInRange("nextFloat", (long) origin, (long) bound, nextMethod);
     }
 
 
     @ParameterizedTest
     @CsvSource({
         // Note: If the range limits are integers above 2^53 (9007199254740992) it is not possible
         // to represent all the values with a double. This has no effect on sampling into bins
         // but should be avoided when generating integers for use in production code.
 
         // No lower bound.
         "2673846826, 0, 11",
         "-23658268, 0, 19",
         "263478624, 0, 31",
         "1278332, 0, 32",
         "99734765, 0, 1234",
         "-63485384, 0, 578",
         "3876457638, 0, 10000",
         "-126784782, 0, 2983423",
         "2637846, 0, 9007199254740992",
         // Range
         "2634682, 567576, 567586",
         "-56757798989, -1000, -100",
         "-97324785, -54656, 12",
         "23423235, -526783468, 257",
         "-2634682, -688689797, -516827",
         "6786868132, -67, 67",
         "-263846723, -5678, 42",
         "7352352, 678687, 61523457",
     })
     void testNextDoubleUniform(long seed, double origin, double bound) {
         Assertions.assertEquals((long) origin, origin, "origin");
         Assertions.assertEquals((long) bound, bound, "bound");
         final UniformRandomProvider rng = createRNG(seed);
         // Note casting as long will round towards zero.
         // If the upper bound is negative then this can create a domain error so use floor.
         final LongSupplier nextMethod = origin == 0 ?
                 () -> (long) rng.nextDouble(bound) :
                 () -> (long) Math.floor(rng.nextDouble(origin, bound));
         checkNextInRange("nextDouble", (long) origin, (long) bound, nextMethod);
     }
 
     /**
      * Tests uniformity of the distribution produced by the given
      * {@code nextMethod}.
      * It performs a chi-square test of homogeneity of the observed
      * distribution with the expected uniform distribution.
      * Repeat tests are performed at the 1% level and the total number of failed
      * tests is tested at the 0.5% significance level.
      *
      * @param method Generator method.
      * @param origin Lower bound (inclusive).
      * @param bound Upper bound (exclusive).
      * @param nextMethod method to call.
      */
     private static void checkNextInRange(String method,
                                          long origin,
                                          long bound,
                                          LongSupplier nextMethod) {
         // Do not change
         // (statistical test assumes that 500 repeats are made with dof = 9).
         final int numTests = 500;
         final int numBins = 10; // dof = numBins - 1
 
         // Set up bins.
         final long[] binUpperBounds = new long[numBins];
         // Range may be above a positive long: step = (bound - origin) / bins
         final BigDecimal range = BigDecimal.valueOf(bound)
                 .subtract(BigDecimal.valueOf(origin));
         final double step = range.divide(BigDecimal.TEN).doubleValue();
         for (int k = 1; k < numBins; k++) {
             binUpperBounds[k - 1] = origin + (long) (k * step);
         }
         // Final bound
         binUpperBounds[numBins - 1] = bound;
 
         // Create expected frequencies
         final double[] expected = new double[numBins];
         long previousUpperBound = origin;
         final double scale = SAMPLE_SIZE_BD.divide(range, MathContext.DECIMAL128).doubleValue();
         double sum = 0;
         for (int k = 0; k < numBins; k++) {
             final long binWidth = binUpperBounds[k] - previousUpperBound;
             expected[k] = scale * binWidth;
             sum += expected[k];
             previousUpperBound = binUpperBounds[k];
         }
         Assertions.assertEquals(SAMPLE_SIZE, sum, SAMPLE_SIZE * RELATIVE_ERROR, "Invalid expected frequencies");
 
         final int[] observed = new int[numBins];
         // Chi-square critical value with 9 degrees of freedom
         // and 1% significance level.
         final double chi2CriticalValue = 21.665994333461924;
 
         // For storing chi2 larger than the critical value.
         final List<Double> failedStat = new ArrayList<>();
         try {
             final int lastDecileIndex = numBins - 1;
             for (int i = 0; i < numTests; i++) {
                 Arrays.fill(observed, 0);
                 SAMPLE: for (int j = 0; j < SAMPLE_SIZE; j++) {
                     final long value = nextMethod.getAsLong();
                     if (value < origin) {
                         Assertions.fail(String.format("Sample %d not within bound [%d, %d)",
                                                       value, origin, bound));
                     }
 
                     for (int k = 0; k < lastDecileIndex; k++) {
                         if (value < binUpperBounds[k]) {
                             ++observed[k];
                             continue SAMPLE;
                         }
                     }
                     if (value >= bound) {
                         Assertions.fail(String.format("Sample %d not within bound [%d, %d)",
                                                       value, origin, bound));
                     }
                     ++observed[lastDecileIndex];
                 }
 
                 // Compute chi-square.
                 double chi2 = 0;
                 for (int k = 0; k < numBins; k++) {
                     final double diff = observed[k] - expected[k];
                     chi2 += diff * diff / expected[k];
                 }
 
                 // Statistics check.
                 if (chi2 > chi2CriticalValue) {
                     failedStat.add(chi2);
                 }
             }
         } catch (final Exception e) {
             // Should never happen.
             throw new RuntimeException("Unexpected", e);
         }
 
         // The expected number of failed tests can be modelled as a Binomial distribution
         // B(n, p) with n=500, p=0.01 (500 tests with a 1% significance level).
         // The cumulative probability of the number of failed tests (X) is:
         // x     P(X>x)
         // 10    0.0132
         // 11    0.00521
         // 12    0.00190
 
         if (failedStat.size() > 11) { // Test will fail with 0.5% probability
             Assertions.fail(String.format(
                 "%s: Too many failures for n = %d, sample size = %d " +
                 "(%d out of %d tests failed, chi2 > %.3f=%s)",
                 method, bound, SAMPLE_SIZE, failedStat.size(), numTests, chi2CriticalValue,
                 failedStat.stream().map(d -> String.format("%.3f", d))
                           .collect(Collectors.joining(", ", "[", "]"))));
         }
     }
 }