We already raised for a mismatching freq for a normal Datetime/TimedeltaIndex in assert_frame_equal, and #65194 updated assert_index_equal to have a similar behaviour (behind a deprecation warning first).
But a potential side effect of that PR (didn't verify) is that this now raises for a MultiIndex level:
dates = pd.date_range("2012-01-01", periods=3)
midx1 = pd.MultiIndex.from_arrays([dates, [1, 2, 3]])
midx2 = pd.MultiIndex.from_arrays([dates._with_freq(None), [1, 2, 3]])
df1 = pd.DataFrame({"a": [1, 2, 3]}, index=midx1)
df2 = pd.DataFrame({"a": [1, 2, 3]}, index=midx2)On main:
# this now raises a warning -> expected
>>> pd.testing.assert_index_equal(midx1, midx2)
Pandas4Warning: assert_index_equal will check the 'freq' attribute by default in a future version; these freqs <Day> and None do not match. Pass check_freq=True or check_freq=False to silence this warning
# this directly errors
>>> pd.testing.assert_frame_equal(df1, df2)
...
Attribute "freq" are different
[left]: <Day>
[right]: None
while with released pandas (3.0.5) both cases pass:
>>> pd.testing.assert_index_equal(midx1, midx2)
>>> pd.testing.assert_frame_equal(df1, df2)
cc @jbrockmendel
We already raised for a mismatching freq for a normal Datetime/TimedeltaIndex in
assert_frame_equal, and #65194 updatedassert_index_equalto have a similar behaviour (behind a deprecation warning first).But a potential side effect of that PR (didn't verify) is that this now raises for a MultiIndex level:
On main:
while with released pandas (3.0.5) both cases pass:
cc @jbrockmendel