codekingpro/portable-devtools

1import sys
2
3import pytest
4
5import numpy as np
6from numpy import random
7from numpy.testing import assert_, assert_array_equal, assert_raises
8
9
10class TestRegression:
11
12    def test_VonMises_range(self):
13        # Make sure generated random variables are in [-pi, pi].
14        # Regression test for ticket #986.
15        for mu in np.linspace(-7., 7., 5):
16            r = random.vonmises(mu, 1, 50)
17            assert_(np.all(r > -np.pi) and np.all(r <= np.pi))
18
19    def test_hypergeometric_range(self):
20        # Test for ticket #921
21        assert_(np.all(random.hypergeometric(3, 18, 11, size=10) < 4))
22        assert_(np.all(random.hypergeometric(18, 3, 11, size=10) > 0))
23
24        # Test for ticket #5623
25        args = [
26            (2**20 - 2, 2**20 - 2, 2**20 - 2),  # Check for 32-bit systems
27        ]
28        is_64bits = sys.maxsize > 2**32
29        if is_64bits and sys.platform != 'win32':
30            # Check for 64-bit systems
31            args.append((2**40 - 2, 2**40 - 2, 2**40 - 2))
32        for arg in args:
33            assert_(random.hypergeometric(*arg) > 0)
34
35    def test_logseries_convergence(self):
36        # Test for ticket #923
37        N = 1000
38        random.seed(0)
39        rvsn = random.logseries(0.8, size=N)
40        # these two frequency counts should be close to theoretical
41        # numbers with this large sample
42        # theoretical large N result is 0.49706795
43        freq = np.sum(rvsn == 1) / N
44        msg = f'Frequency was {freq:f}, should be > 0.45'
45        assert_(freq > 0.45, msg)
46        # theoretical large N result is 0.19882718
47        freq = np.sum(rvsn == 2) / N
48        msg = f'Frequency was {freq:f}, should be < 0.23'
49        assert_(freq < 0.23, msg)
50
51    def test_shuffle_mixed_dimension(self):
52        # Test for trac ticket #2074
53        for t in [[1, 2, 3, None],
54                  [(1, 1), (2, 2), (3, 3), None],
55                  [1, (2, 2), (3, 3), None],
56                  [(1, 1), 2, 3, None]]:
57            rng = random.RandomState(12345)
58            shuffled = list(t)
59            rng.shuffle(shuffled)
60            expected = np.array([t[0], t[3], t[1], t[2]], dtype=object)
61            assert_array_equal(np.array(shuffled, dtype=object), expected)
62
63    def test_call_within_randomstate(self):
64        # Check that custom RandomState does not call into global state
65        m = random.RandomState()
66        res = np.array([0, 8, 7, 2, 1, 9, 4, 7, 0, 3])
67        for i in range(3):
68            random.seed(i)
69            m.seed(4321)
70            # If m.state is not honored, the result will change
71            assert_array_equal(m.choice(10, size=10, p=np.ones(10) / 10.), res)
72
73    def test_multivariate_normal_size_types(self):
74        # Test for multivariate_normal issue with 'size' argument.
75        # Check that the multivariate_normal size argument can be a
76        # numpy integer.
77        random.multivariate_normal([0], [[0]], size=1)
78        random.multivariate_normal([0], [[0]], size=np.int_(1))
79        random.multivariate_normal([0], [[0]], size=np.int64(1))
80
81    def test_beta_small_parameters(self):
82        # Test that beta with small a and b parameters does not produce
83        # NaNs due to roundoff errors causing 0 / 0, gh-5851
84        random.seed(1234567890)
85        x = random.beta(0.0001, 0.0001, size=100)
86        assert_(not np.any(np.isnan(x)), 'Nans in random.beta')
87
88    def test_choice_sum_of_probs_tolerance(self):
89        # The sum of probs should be 1.0 with some tolerance.
90        # For low precision dtypes the tolerance was too tight.
91        # See numpy github issue 6123.
92        random.seed(1234)
93        a = [1, 2, 3]
94        counts = [4, 4, 2]
95        for dt in np.float16, np.float32, np.float64:
96            probs = np.array(counts, dtype=dt) / sum(counts)
97            c = random.choice(a, p=probs)
98            assert_(c in a)
99            assert_raises(ValueError, random.choice, a, p=probs * 0.9)
100
101    def test_shuffle_of_array_of_different_length_strings(self):
102        # Test that permuting an array of different length strings
103        # will not cause a segfault on garbage collection
104        # Tests gh-7710
105        random.seed(1234)
106
107        a = np.array(['a', 'a' * 1000])
108
109        for _ in range(100):
110            random.shuffle(a)
111
112        # Force Garbage Collection - should not segfault.
113        import gc
114        gc.collect()
115
116    def test_shuffle_of_array_of_objects(self):
117        # Test that permuting an array of objects will not cause
118        # a segfault on garbage collection.
119        # See gh-7719
120        random.seed(1234)
121        a = np.array([np.arange(1), np.arange(4)], dtype=object)
122
123        for _ in range(1000):
124            random.shuffle(a)
125
126        # Force Garbage Collection - should not segfault.
127        import gc
128        gc.collect()
129
130    def test_permutation_subclass(self):
131        class N(np.ndarray):
132            pass
133
134        rng = random.RandomState(1)
135        orig = np.arange(3).view(N)
136        perm = rng.permutation(orig)
137        assert_array_equal(perm, np.array([0, 2, 1]))
138        assert_array_equal(orig, np.arange(3).view(N))
139
140        class M:
141            a = np.arange(5)
142
143            def __array__(self, dtype=None, copy=None):
144                return self.a
145
146        rng = random.RandomState(1)
147        m = M()
148        perm = rng.permutation(m)
149        assert_array_equal(perm, np.array([2, 1, 4, 0, 3]))
150        assert_array_equal(m.__array__(), np.arange(5))
151
152    def test_warns_byteorder(self):
153        # GH 13159
154        other_byteord_dt = '<i4' if sys.byteorder == 'big' else '>i4'
155        with pytest.deprecated_call(match='non-native byteorder is not'):
156            random.randint(0, 200, size=10, dtype=other_byteord_dt)
157
158    def test_named_argument_initialization(self):
159        # GH 13669
160        rs1 = np.random.RandomState(123456789)
161        rs2 = np.random.RandomState(seed=123456789)
162        assert rs1.randint(0, 100) == rs2.randint(0, 100)
163
164    def test_choice_retun_dtype(self):
165        # GH 9867, now long since the NumPy default changed.
166        c = np.random.choice(10, p=[.1] * 10, size=2)
167        assert c.dtype == np.dtype(np.long)
168        c = np.random.choice(10, p=[.1] * 10, replace=False, size=2)
169        assert c.dtype == np.dtype(np.long)
170        c = np.random.choice(10, size=2)
171        assert c.dtype == np.dtype(np.long)
172        c = np.random.choice(10, replace=False, size=2)
173        assert c.dtype == np.dtype(np.long)
174
175    @pytest.mark.skipif(np.iinfo('l').max < 2**32,
176                        reason='Cannot test with 32-bit C long')
177    def test_randint_117(self):
178        # GH 14189
179        rng = random.RandomState(0)
180        expected = np.array([2357136044, 2546248239, 3071714933, 3626093760,
181                             2588848963, 3684848379, 2340255427, 3638918503,
182                             1819583497, 2678185683], dtype='int64')
183        actual = rng.randint(2**32, size=10)
184        assert_array_equal(actual, expected)
185
186    def test_p_zero_stream(self):
187        # Regression test for gh-14522.  Ensure that future versions
188        # generate the same variates as version 1.16.
189        rng = random.RandomState(12345)
190        assert_array_equal(rng.binomial(1, [0, 0.25, 0.5, 0.75, 1]),
191                           [0, 0, 0, 1, 1])
192
193    def test_n_zero_stream(self):
194        # Regression test for gh-14522.  Ensure that future versions
195        # generate the same variates as version 1.16.
196        rng = random.RandomState(8675309)
197        expected = np.array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
198                             [3, 4, 2, 3, 3, 1, 5, 3, 1, 3]])
199        assert_array_equal(rng.binomial([[0], [10]], 0.25, size=(2, 10)),
200                           expected)
201
202
203def test_multinomial_empty():
204    # gh-20483
205    # Ensure that empty p-vals are correctly handled
206    assert random.multinomial(10, []).shape == (0,)
207    assert random.multinomial(3, [], size=(7, 5, 3)).shape == (7, 5, 3, 0)
208
209
210def test_multinomial_1d_pval():
211    # gh-20483
212    with pytest.raises(TypeError, match="pvals must be a 1-d"):
213        random.multinomial(10, 0.3)
214