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
|
import bisect
import itertools
from random import Random
from typing import Generator, Iterable, Optional, Sequence, TypeVar
from faker.generator import random as mod_random
def random_sample(random: Optional[Random] = None) -> float:
if random is None:
random = mod_random
return random.uniform(0.0, 1.0)
def cumsum(it: Iterable[float]) -> Generator[float, None, None]:
total: float = 0
for x in it:
total += x
yield total
T = TypeVar("T")
def choices_distribution_unique(
a: Sequence[T],
p: Optional[Sequence[float]],
random: Optional[Random] = None,
length: int = 1,
) -> Sequence[T]:
# As of Python 3.7, there isn't a way to sample unique elements that takes
# weight into account.
if random is None:
random = mod_random
assert p is not None
assert len(a) == len(p)
assert len(a) >= length, "You can't request more unique samples than elements in the dataset."
choices = []
items = list(a)
probabilities = list(p)
for i in range(length):
cdf = tuple(cumsum(probabilities))
normal = cdf[-1]
cdf2 = [i / normal for i in cdf]
uniform_sample = random_sample(random=random)
idx = bisect.bisect_right(cdf2, uniform_sample)
item = items[idx]
choices.append(item)
probabilities.pop(idx)
items.pop(idx)
return choices
def choices_distribution(
a: Sequence[T],
p: Optional[Sequence[float]],
random: Optional[Random] = None,
length: int = 1,
) -> Sequence[T]:
if random is None:
random = mod_random
if p is not None:
assert len(a) == len(p)
if hasattr(random, "choices"):
if length == 1 and p is None:
return [random.choice(a)]
else:
return random.choices(a, weights=p, k=length)
else:
choices = []
if p is None:
p = itertools.repeat(1, len(a)) # type: ignore
cdf = list(cumsum(p)) # type: ignore
normal = cdf[-1]
cdf2 = [i / normal for i in cdf]
for i in range(length):
uniform_sample = random_sample(random=random)
idx = bisect.bisect_right(cdf2, uniform_sample)
item = a[idx]
choices.append(item)
return choices
|
