why is_close_to is a tolerance relative to the larger of the value and expected

[dpinol]

relatively_within docstring provides:

Scaled by the larger, as `math.isclose` scales.  `pytest.approx` scales by the expected side alone and
    answers ``100`` against ``98`` within ``0.02`` one way round and not the other.

I understand math.isclose behaviour because there's no different semantic for both arguments, but when one is an expected/reference one, I think pytest behaviour is more intuitive. Otherwise, this assertion passes
assert_that(10).is_close_to(1, rel=1)
where I wanted to express the value to be 100% away from the expected value

[Solganis]

Checked this and I agree, it smells bad

In an assertion the expected value is the reference, so "within 100%" should be measured against it, aspytest.approx, numpy.testing.assert_allclose and AssertJ do

Will change this with next minor version

Right now you can use something like this assert_that(10).is_equal_to(pytest.approx(1, rel=1))