2.34.0

Read Behaviour changes below before upgrading.

TL;DR

New
log_passes()	logs each assertion that passes inside the block, with how far the value was from its bound (#57)

Passed when it should not have
assert_that(2).is_close_to(1, rel=0.5), and every rel= where the value is the larger number	passed, the tolerance scaled with the larger of the two (#56)
assert_that(1e308).is_close_to(-1e308, rel=1.9)	passed, the difference and the allowance both overflowed to infinity

log_passes() logs the assertions that pass

latency, status = 0.42, 200
with log_passes():
    assert_that(latency).is_less_than(0.5)
    assert_that(status).is_equal_to(200)
    assert_that(98.7).is_close_to(100, 2)

Before

ImportError: cannot import name 'log_passes' from 'assertpy2'

Now

[test_api.py:5]: is_less_than(0.5) passed for <0.42>, the value is about 0.08 below the number given
[test_api.py:6]: is_equal_to(200) passed for <200>
[test_api.py:7]: is_close_to(100, 2) passed for <98.7>, the two are about 1.3 apart, about 0.7 less than the tolerance

• Logger assertpy2.passes, level INFO. In pytest: -o log_cli=true --log-cli-level=INFO
• The distance is given for thresholds: is_less_than(), is_greater_than(), is_between(), is_close_to() and their kin
• on_pass= hands each pass to your code as an AssertionOutcome, the distance in hint
• Only real passes: no line for soft or warn failures or under check(). A poll logs the attempt that passed
• Outside a block: one extra check per extension call and poll attempt, about 2 to 3%

Guide: Logging the assertions that pass, Handing each pass to your own code

rel= is a share of the expected value

assert_that(2).is_close_to(1, rel=0.5)  # 1
assert_that(1e308).is_close_to(-1e308, rel=1.9)  # 2
assert_that(2).is_not_close_to(1, rel=0.5)  # 3

Before

1: passed
2: passed
3: AssertionFailure: Expected <2> to not be close to <1> within relative tolerance <0.5>, but was.

Now

1: AssertionFailure: Expected <2> to be close to <1> within relative tolerance <0.5>, but was not.
2: AssertionFailure: Expected <1e+308> to be close to <-1e+308> within relative tolerance <1.9>, but was not.
3: passed

• rel=0.5 against 1 allows 0.5 either side, as in pytest.approx() and numpy.testing.assert_allclose()
• 2.33 took the share of the larger number, as math.isclose() does, so rel=1 passed any two numbers of one sign
• Same for is_equal_to(), snapshot(), matches_inline(), match.close_to() and match.equal_to()

Guide: Numbers

Behaviour changes

rel= on is_close_to(), is_not_close_to(), is_equal_to(), snapshot(), matches_inline(), match.close_to() and match.equal_to()	is a share of the expected value. It was a share of the larger of the two, so a value above the expected one now gets a smaller allowance

Everything else

• build: ty 0.0.95, ruff 0.17.3 and pytest 9.2.0
• ci: every suite runs on four workers by file, and coverage is measured only where it is reported
• test: the rich diff's child pytest runs in the suite's own interpreter