Confusing explanation for `pytest.approx(subset) == superset` on mappings

[bluetech]

Consider this:

import pytest

def test_it():
    assert pytest.approx({"a": 1, "b": 2}) == {"a": 1, "b": 2, "c": 3}

Currently results in this output:

    def test_it():
>       assert pytest.approx({"a": 1, "b": 2}) == {"a": 1, "b": 2, "c": 3}
E       AssertionError: assert approx({'a': ... 2 ± 2.0e-06}) == {'a': 1, 'b': 2, 'c': 3}
E         
E         comparison failed. Mismatched elements: 0 / 2:
E         Max absolute difference: -inf
E         Max relative difference: -inf
E         Index | Obtained | Expected

The "comparison failed" messages doesn't actually show anything wrong (no mismatches), which is confusing.

I think it should give a message like ApproxSequence does when the lengths don't match:

E         Impossible to compare mappings with different sizes.
E         Lengths: 2 and 3

[Pierre-Sassoulas]

I generally add an assert len(x) == len(y) when that happens, so the proposed behavior would save some time for me.