BUG: Inconsistent behavior in Index.difference #35231
Conversation
simonjayhawkins
added
Index
There was a problem hiding this comment.
Generally this looks good to me.
But can you check how well Float64Index equals is covered in the tests?
For example, you added an assert that floats and strings are not equal, but if we preserve behaviour, we should ensure that float / ints and float / object ints keep evaluating to True:
In [18]: pd.Index([1, 2]).equals(pd.Index([1., 2.]))
Out[18]: True
In [20]: pd.Index([1, 2]).equals(pd.Index([1., 2.], dtype=object))
Out[20]: True
(both of those will work fine as well with the new implementaiton, I think (since array_equivalent right now doesn't check the exact dtype for those, only equality), but just want to ensure it is tested)
|
Thanks @jorisvandenbossche for reviewing this. have extended an existing test |
pandas/tests/indexes/test_numeric.py
Outdated
|
|
||
| i2 = Float64Index([1.0, np.nan]) | ||
| assert i.equals(i2) | ||
| i4 = Float64Index([1.0, np.nan]) |
|
|
||
| def test_float64_index_equals(): | ||
| # https://github.com/pandas-dev/pandas/issues/35217 | ||
| float_index = pd.Index([1.0, 2, 3]) |
|
|
||
| def test_float64_index_difference(): | ||
| # https://github.com/pandas-dev/pandas/issues/35217 | ||
| float_index = pd.Index([1.0, 2, 3]) |
|
@simonjayhawkins are you able to update here? |
i've parameterised the additional tests added in response to #35231 (review) not sure parameterise makes sense for the issue regression tests (parameterise on have pushed changes but not yet appearing here. |
The changes are now showing, but ci not running. |
|
azure and travis triggered, but not github actions. will push another empty commit later if github actions doesn't run. |
|
lgtm. can merge on green. |
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff