Fix IndexError in choices_distribution when uniform() returns 1.0

[bysiber]

Summary

choices_distribution_unique (and the fallback path of choices_distribution) can crash with IndexError when random.uniform(0, 1.0) returns exactly 1.0.

The Bug

random.uniform(0, 1.0) can return 1.0 — from the Python docs: "The end-point value b may or may not be included" and the implementation note says a + (b-a) * random() can return b.

When this happens, bisect.bisect_right(cdf2, 1.0) returns len(cdf2) because the last CDF element is always 1.0 and bisect_right places equal values to the right. This index is one past the end:

uniform_sample = random_sample(random=random)  # can be 1.0
idx = bisect.bisect_right(cdf2, uniform_sample)  # returns len(cdf2)
item = items[idx]  # IndexError!

from faker.utils.distribution import choices_distribution_unique
from unittest.mock import MagicMock
from random import Random

mock_random = MagicMock(spec=Random)
mock_random.uniform.return_value = 1.0

choices_distribution_unique(["a", "b"], [0.5, 0.5], random=mock_random, length=1)
# IndexError: list index out of range

This is intermittent in production — uniform() returning exactly 1.0 is rare but will happen given enough calls.

The Fix

Clamp the index to the valid range:

idx = bisect.bisect_right(cdf2, uniform_sample)
idx = min(idx, len(items) - 1)

Applied to both choices_distribution_unique and the fallback path of choices_distribution.

[fcurella]

Thank you! Could you add a unit test?