File: test_float_nastiness.py

package info (click to toggle)
python-hypothesis 6.138.0-1
  • links: PTS, VCS
  • area: main
  • in suites: sid
  • size: 15,272 kB
  • sloc: python: 62,853; ruby: 1,107; sh: 253; makefile: 41; javascript: 6
file content (276 lines) | stat: -rw-r--r-- 8,442 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
# This file is part of Hypothesis, which may be found at
# https://github.com/HypothesisWorks/hypothesis/
#
# Copyright the Hypothesis Authors.
# Individual contributors are listed in AUTHORS.rst and the git log.
#
# This Source Code Form is subject to the terms of the Mozilla Public License,
# v. 2.0. If a copy of the MPL was not distributed with this file, You can
# obtain one at https://mozilla.org/MPL/2.0/.

import math
import sys
import warnings

import pytest

from hypothesis import assume, given, strategies as st
from hypothesis.errors import InvalidArgument
from hypothesis.internal.floats import (
    float_of,
    float_to_int,
    int_to_float,
    is_negative,
    next_down,
    next_up,
)

from tests.common.debug import find_any, minimal

try:
    import numpy
except ImportError:
    numpy = None


@pytest.mark.parametrize(
    ("lower", "upper"),
    [
        # Exact values don't matter, but they're large enough so that x + y = inf.
        (9.9792015476736e291, 1.7976931348623157e308),
        (-sys.float_info.max, sys.float_info.max),
    ],
)
@given(data=st.data())
def test_floats_are_in_range(data, lower, upper):
    t = data.draw(st.floats(lower, upper))
    assert lower <= t <= upper


@pytest.mark.parametrize("sign", [-1, 1])
def test_can_generate_both_zeros(sign):
    assert minimal(st.floats(), lambda x: math.copysign(1, x) == sign) == sign * 0.0


@pytest.mark.parametrize(
    ("l", "r"),
    [(-1.0, 1.0), (-0.0, 1.0), (-1.0, 0.0), (-sys.float_info.min, sys.float_info.min)],
)
@pytest.mark.parametrize("sign", [-1, 1])
def test_can_generate_both_zeros_when_in_interval(l, r, sign):
    assert minimal(st.floats(l, r), lambda x: math.copysign(1, x) == sign) == sign * 0.0


@given(st.floats(0.0, 1.0))
def test_does_not_generate_negative_if_right_boundary_is_positive(x):
    assert math.copysign(1, x) == 1


@given(st.floats(-1.0, -0.0))
def test_does_not_generate_positive_if_right_boundary_is_negative(x):
    assert math.copysign(1, x) == -1


def test_half_bounded_generates_zero():
    find_any(st.floats(min_value=-1.0), lambda x: x == 0.0)
    find_any(st.floats(max_value=1.0), lambda x: x == 0.0)


@given(st.floats(max_value=-0.0))
def test_half_bounded_respects_sign_of_upper_bound(x):
    assert math.copysign(1, x) == -1


@given(st.floats(min_value=0.0))
def test_half_bounded_respects_sign_of_lower_bound(x):
    assert math.copysign(1, x) == 1


@given(st.floats(allow_nan=False))
def test_filter_nan(x):
    assert not math.isnan(x)


@given(st.floats(allow_infinity=False))
def test_filter_infinity(x):
    assert not math.isinf(x)


def test_can_guard_against_draws_of_nan():
    """In this test we create a NaN value that naturally "tries" to shrink into
    the first strategy, where it is not permitted. This tests a case that is
    very unlikely to happen in random generation: When the unconstrained first
    branch of generating a float just happens to produce a NaN value.

    Here what happens is that we get a NaN from the *second* strategy,
    but this then shrinks into its unconstrained branch. The natural
    thing to happen is then to try to zero the branch parameter of the
    one_of, but that will put an illegal value there, so it's not
    allowed to happen.
    """
    tagged_floats = st.one_of(
        st.tuples(st.just(0), st.floats(allow_nan=False)),
        st.tuples(st.just(1), st.floats(allow_nan=True)),
    )

    tag, f = minimal(tagged_floats, lambda x: math.isnan(x[1]))
    assert tag == 1


def test_very_narrow_interval():
    upper_bound = -1.0
    lower_bound = int_to_float(float_to_int(upper_bound) + 10)
    assert lower_bound < upper_bound

    @given(st.floats(lower_bound, upper_bound))
    def test(f):
        assert lower_bound <= f <= upper_bound

    test()


@given(st.floats())
def test_up_means_greater(x):
    hi = next_up(x)
    if not x < hi:
        assert (
            (math.isnan(x) and math.isnan(hi))
            or (x > 0 and math.isinf(x))
            or (x == hi == 0 and is_negative(x) and not is_negative(hi))
        )


@given(st.floats())
def test_down_means_lesser(x):
    lo = next_down(x)
    if not x > lo:
        assert (
            (math.isnan(x) and math.isnan(lo))
            or (x < 0 and math.isinf(x))
            or (x == lo == 0 and is_negative(lo) and not is_negative(x))
        )


@given(st.floats(allow_nan=False, allow_infinity=False))
def test_updown_roundtrip(val):
    assert val == next_up(next_down(val))
    assert val == next_down(next_up(val))


@given(st.floats(width=32, allow_infinity=False))
def test_float32_can_exclude_infinity(x):
    assert not math.isinf(x)


@given(st.floats(width=16, allow_infinity=False))
def test_float16_can_exclude_infinity(x):
    assert not math.isinf(x)


@pytest.mark.parametrize(
    "kwargs",
    [
        {"min_value": 10**5, "width": 16},
        {"max_value": 10**5, "width": 16},
        {"min_value": 10**40, "width": 32},
        {"max_value": 10**40, "width": 32},
        {"min_value": 10**400, "width": 64},
        {"max_value": 10**400, "width": 64},
        {"min_value": 10**400},
        {"max_value": 10**400},
    ],
)
def test_out_of_range(kwargs):
    with pytest.raises(OverflowError):
        st.floats(**kwargs).validate()


def test_disallowed_width():
    with pytest.raises(InvalidArgument):
        st.floats(width=128).validate()


def test_no_single_floats_in_range():
    low = 2.0**25 + 1
    high = low + 2
    st.floats(low, high).validate()  # Note: OK for 64bit floats
    with warnings.catch_warnings():
        # Unrepresentable bounds are deprecated, but we're not testing that here
        warnings.simplefilter("ignore")
        with pytest.raises(InvalidArgument):
            st.floats(low, high, width=32).validate()


# If the floats() strategy adds random floats to a value as large as 10^304
# without handling overflow, we are very likely to generate infinity.
@given(st.floats(min_value=1e304, allow_infinity=False))
def test_finite_min_bound_does_not_overflow(x):
    assert not math.isinf(x)


@given(st.floats(max_value=-1e304, allow_infinity=False))
def test_finite_max_bound_does_not_overflow(x):
    assert not math.isinf(x)


@given(st.floats(0, 1, exclude_min=True, exclude_max=True))
def test_can_exclude_endpoints(x):
    assert 0 < x < 1


@given(st.floats(-math.inf, -1e307, exclude_min=True))
def test_can_exclude_neg_infinite_endpoint(x):
    assert not math.isinf(x)


@given(st.floats(1e307, math.inf, exclude_max=True))
def test_can_exclude_pos_infinite_endpoint(x):
    assert not math.isinf(x)


def test_exclude_infinite_endpoint_is_invalid():
    with pytest.raises(InvalidArgument):
        st.floats(min_value=math.inf, exclude_min=True).validate()
    with pytest.raises(InvalidArgument):
        st.floats(max_value=-math.inf, exclude_max=True).validate()


@pytest.mark.parametrize("lo,hi", [(True, False), (False, True), (True, True)])
@given(bound=st.floats(allow_nan=False, allow_infinity=False).filter(bool))
def test_exclude_entire_interval(lo, hi, bound):
    with pytest.raises(InvalidArgument, match="exclude_min=.+ and exclude_max="):
        st.floats(bound, bound, exclude_min=lo, exclude_max=hi).validate()


def test_zero_intervals_are_OK():
    st.floats(0.0, 0.0).validate()
    st.floats(-0.0, 0.0).validate()
    st.floats(-0.0, -0.0).validate()


@pytest.mark.parametrize("lo", [0.0, -0.0])
@pytest.mark.parametrize("hi", [0.0, -0.0])
@pytest.mark.parametrize("exmin,exmax", [(True, False), (False, True), (True, True)])
def test_cannot_exclude_endpoint_with_zero_interval(lo, hi, exmin, exmax):
    with pytest.raises(InvalidArgument):
        st.floats(lo, hi, exclude_min=exmin, exclude_max=exmax).validate()


WIDTHS = (64, 32, 16)


@pytest.mark.parametrize("nonfloat", [st.nothing(), st.none()])
@given(data=st.data(), width=st.sampled_from(WIDTHS))
def test_fuzzing_floats_bounds(data, width, nonfloat):
    lo = data.draw(nonfloat | st.floats(allow_nan=False, width=width), label="lo")
    hi = data.draw(nonfloat | st.floats(allow_nan=False, width=width), label="hi")
    if lo is not None and hi is not None and lo > hi:
        lo, hi = hi, lo
    assume(lo != 0 or hi != 0)
    value = data.draw(
        st.floats(min_value=lo, max_value=hi, width=width, allow_nan=False),
        label="value",
    )
    assert value == float_of(value, width=width)
    assert lo is None or lo <= value
    assert hi is None or value <= hi