1 // Copyright 2017 The Abseil Authors.
2 //
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
6 //
7 // https://www.apache.org/licenses/LICENSE-2.0
8 //
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
14
15 #include "absl/random/uniform_int_distribution.h"
16
17 #include <cmath>
18 #include <cstdint>
19 #include <iterator>
20 #include <random>
21 #include <sstream>
22 #include <vector>
23
24 #include "gmock/gmock.h"
25 #include "gtest/gtest.h"
26 #include "absl/base/internal/raw_logging.h"
27 #include "absl/random/internal/chi_square.h"
28 #include "absl/random/internal/distribution_test_util.h"
29 #include "absl/random/internal/pcg_engine.h"
30 #include "absl/random/internal/sequence_urbg.h"
31 #include "absl/random/random.h"
32 #include "absl/strings/str_cat.h"
33
34 namespace {
35
36 template <typename IntType>
37 class UniformIntDistributionTest : public ::testing::Test {};
38
39 using IntTypes = ::testing::Types<int8_t, uint8_t, int16_t, uint16_t, int32_t,
40 uint32_t, int64_t, uint64_t>;
41 TYPED_TEST_SUITE(UniformIntDistributionTest, IntTypes);
42
TYPED_TEST(UniformIntDistributionTest,ParamSerializeTest)43 TYPED_TEST(UniformIntDistributionTest, ParamSerializeTest) {
44 // This test essentially ensures that the parameters serialize,
45 // not that the values generated cover the full range.
46 using Limits = std::numeric_limits<TypeParam>;
47 using param_type =
48 typename absl::uniform_int_distribution<TypeParam>::param_type;
49 const TypeParam kMin = std::is_unsigned<TypeParam>::value ? 37 : -105;
50 const TypeParam kNegOneOrZero = std::is_unsigned<TypeParam>::value ? 0 : -1;
51
52 constexpr int kCount = 1000;
53 absl::InsecureBitGen gen;
54 for (const auto& param : {
55 param_type(),
56 param_type(2, 2), // Same
57 param_type(9, 32),
58 param_type(kMin, 115),
59 param_type(kNegOneOrZero, Limits::max()),
60 param_type(Limits::min(), Limits::max()),
61 param_type(Limits::lowest(), Limits::max()),
62 param_type(Limits::min() + 1, Limits::max() - 1),
63 }) {
64 const auto a = param.a();
65 const auto b = param.b();
66 absl::uniform_int_distribution<TypeParam> before(a, b);
67 EXPECT_EQ(before.a(), param.a());
68 EXPECT_EQ(before.b(), param.b());
69
70 {
71 // Initialize via param_type
72 absl::uniform_int_distribution<TypeParam> via_param(param);
73 EXPECT_EQ(via_param, before);
74 }
75
76 // Initialize via iostreams
77 std::stringstream ss;
78 ss << before;
79
80 absl::uniform_int_distribution<TypeParam> after(Limits::min() + 3,
81 Limits::max() - 5);
82
83 EXPECT_NE(before.a(), after.a());
84 EXPECT_NE(before.b(), after.b());
85 EXPECT_NE(before.param(), after.param());
86 EXPECT_NE(before, after);
87
88 ss >> after;
89
90 EXPECT_EQ(before.a(), after.a());
91 EXPECT_EQ(before.b(), after.b());
92 EXPECT_EQ(before.param(), after.param());
93 EXPECT_EQ(before, after);
94
95 // Smoke test.
96 auto sample_min = after.max();
97 auto sample_max = after.min();
98 for (int i = 0; i < kCount; i++) {
99 auto sample = after(gen);
100 EXPECT_GE(sample, after.min());
101 EXPECT_LE(sample, after.max());
102 if (sample > sample_max) {
103 sample_max = sample;
104 }
105 if (sample < sample_min) {
106 sample_min = sample;
107 }
108 }
109 std::string msg = absl::StrCat("Range: ", +sample_min, ", ", +sample_max);
110 ABSL_RAW_LOG(INFO, "%s", msg.c_str());
111 }
112 }
113
TYPED_TEST(UniformIntDistributionTest,ViolatesPreconditionsDeathTest)114 TYPED_TEST(UniformIntDistributionTest, ViolatesPreconditionsDeathTest) {
115 #if GTEST_HAS_DEATH_TEST
116 // Hi < Lo
117 EXPECT_DEBUG_DEATH({ absl::uniform_int_distribution<TypeParam> dist(10, 1); },
118 "");
119 #endif // GTEST_HAS_DEATH_TEST
120 #if defined(NDEBUG)
121 // opt-mode, for invalid parameters, will generate a garbage value,
122 // but should not enter an infinite loop.
123 absl::InsecureBitGen gen;
124 absl::uniform_int_distribution<TypeParam> dist(10, 1);
125 auto x = dist(gen);
126
127 // Any value will generate a non-empty string.
128 EXPECT_FALSE(absl::StrCat(+x).empty()) << x;
129 #endif // NDEBUG
130 }
131
TYPED_TEST(UniformIntDistributionTest,TestMoments)132 TYPED_TEST(UniformIntDistributionTest, TestMoments) {
133 constexpr int kSize = 100000;
134 using Limits = std::numeric_limits<TypeParam>;
135 using param_type =
136 typename absl::uniform_int_distribution<TypeParam>::param_type;
137
138 // We use a fixed bit generator for distribution accuracy tests. This allows
139 // these tests to be deterministic, while still testing the qualify of the
140 // implementation.
141 absl::random_internal::pcg64_2018_engine rng{0x2B7E151628AED2A6};
142
143 std::vector<double> values(kSize);
144 for (const auto& param :
145 {param_type(0, Limits::max()), param_type(13, 127)}) {
146 absl::uniform_int_distribution<TypeParam> dist(param);
147 for (int i = 0; i < kSize; i++) {
148 const auto sample = dist(rng);
149 ASSERT_LE(dist.param().a(), sample);
150 ASSERT_GE(dist.param().b(), sample);
151 values[i] = sample;
152 }
153
154 auto moments = absl::random_internal::ComputeDistributionMoments(values);
155 const double a = dist.param().a();
156 const double b = dist.param().b();
157 const double n = (b - a + 1);
158 const double mean = (a + b) / 2;
159 const double var = ((b - a + 1) * (b - a + 1) - 1) / 12;
160 const double kurtosis = 3 - 6 * (n * n + 1) / (5 * (n * n - 1));
161
162 // TODO(ahh): this is not the right bound
163 // empirically validated with --runs_per_test=10000.
164 EXPECT_NEAR(mean, moments.mean, 0.01 * var);
165 EXPECT_NEAR(var, moments.variance, 0.015 * var);
166 EXPECT_NEAR(0.0, moments.skewness, 0.025);
167 EXPECT_NEAR(kurtosis, moments.kurtosis, 0.02 * kurtosis);
168 }
169 }
170
TYPED_TEST(UniformIntDistributionTest,ChiSquaredTest50)171 TYPED_TEST(UniformIntDistributionTest, ChiSquaredTest50) {
172 using absl::random_internal::kChiSquared;
173
174 constexpr size_t kTrials = 1000;
175 constexpr int kBuckets = 50; // inclusive, so actally +1
176 constexpr double kExpected =
177 static_cast<double>(kTrials) / static_cast<double>(kBuckets);
178
179 // Empirically validated with --runs_per_test=10000.
180 const int kThreshold =
181 absl::random_internal::ChiSquareValue(kBuckets, 0.999999);
182
183 const TypeParam min = std::is_unsigned<TypeParam>::value ? 37 : -37;
184 const TypeParam max = min + kBuckets;
185
186 // We use a fixed bit generator for distribution accuracy tests. This allows
187 // these tests to be deterministic, while still testing the qualify of the
188 // implementation.
189 absl::random_internal::pcg64_2018_engine rng{0x2B7E151628AED2A6};
190
191 absl::uniform_int_distribution<TypeParam> dist(min, max);
192
193 std::vector<int32_t> counts(kBuckets + 1, 0);
194 for (size_t i = 0; i < kTrials; i++) {
195 auto x = dist(rng);
196 counts[x - min]++;
197 }
198 double chi_square = absl::random_internal::ChiSquareWithExpected(
199 std::begin(counts), std::end(counts), kExpected);
200 if (chi_square > kThreshold) {
201 double p_value =
202 absl::random_internal::ChiSquarePValue(chi_square, kBuckets);
203
204 // Chi-squared test failed. Output does not appear to be uniform.
205 std::string msg;
206 for (const auto& a : counts) {
207 absl::StrAppend(&msg, a, "\n");
208 }
209 absl::StrAppend(&msg, kChiSquared, " p-value ", p_value, "\n");
210 absl::StrAppend(&msg, "High ", kChiSquared, " value: ", chi_square, " > ",
211 kThreshold);
212 ABSL_RAW_LOG(INFO, "%s", msg.c_str());
213 FAIL() << msg;
214 }
215 }
216
TEST(UniformIntDistributionTest,StabilityTest)217 TEST(UniformIntDistributionTest, StabilityTest) {
218 // absl::uniform_int_distribution stability relies only on integer operations.
219 absl::random_internal::sequence_urbg urbg(
220 {0x0003eb76f6f7f755ull, 0xFFCEA50FDB2F953Bull, 0xC332DDEFBE6C5AA5ull,
221 0x6558218568AB9702ull, 0x2AEF7DAD5B6E2F84ull, 0x1521B62829076170ull,
222 0xECDD4775619F1510ull, 0x13CCA830EB61BD96ull, 0x0334FE1EAA0363CFull,
223 0xB5735C904C70A239ull, 0xD59E9E0BCBAADE14ull, 0xEECC86BC60622CA7ull});
224
225 std::vector<int> output(12);
226
227 {
228 absl::uniform_int_distribution<int32_t> dist(0, 4);
229 for (auto& v : output) {
230 v = dist(urbg);
231 }
232 }
233 EXPECT_EQ(12, urbg.invocations());
234 EXPECT_THAT(output, testing::ElementsAre(4, 4, 3, 2, 1, 0, 1, 4, 3, 1, 3, 1));
235
236 {
237 urbg.reset();
238 absl::uniform_int_distribution<int32_t> dist(0, 100);
239 for (auto& v : output) {
240 v = dist(urbg);
241 }
242 }
243 EXPECT_EQ(12, urbg.invocations());
244 EXPECT_THAT(output, testing::ElementsAre(97, 86, 75, 41, 36, 16, 38, 92, 67,
245 30, 80, 38));
246
247 {
248 urbg.reset();
249 absl::uniform_int_distribution<int32_t> dist(0, 10000);
250 for (auto& v : output) {
251 v = dist(urbg);
252 }
253 }
254 EXPECT_EQ(12, urbg.invocations());
255 EXPECT_THAT(output, testing::ElementsAre(9648, 8562, 7439, 4089, 3571, 1602,
256 3813, 9195, 6641, 2986, 7956, 3765));
257 }
258
259 } // namespace
260