REF: _is_compatible_with_other -> _can_union_without_object_cast

[jbrockmendel]

• closes #xxxx
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Changes the behavior in one test where ATM we behave differently for Series/ndarray than we do for Index arg

[jorisvandenbossche]

Changes the behavior in one test where ATM we behave differently for Series/ndarray than we do for Index arg

Can you show an example of that behaviour change?

[jbrockmendel]

Can you show an example of that behaviour change?

https://github.com/pandas-dev/pandas/pull/36384/files#diff-a3619540517f93ee7c657a439b986962L50

everything = tm.makeDateIndex(10)
first = everything[:5]
second = everything[5:]

expected_old = first.astype("O").union(pd.Index(second.values, dtype="O")).astype("O")
expected_new = first.union(second)

case = list(second.values)  # <-- same for np.array(second.values)
result = first.union(case)
tm.assert_index_equal(result, expected)

ATM _is_compatible_with_other checks specifically for Index subclasses as opposed to dtype-based checks, and so we end up converting to object when we have anything that isn't an Index.

[jreback]

can rebase

[jbrockmendel]

Rebased +green

[jorisvandenbossche]

ATM _is_compatible_with_other checks specifically for Index subclasses as opposed to dtype-based checks, and so we end up converting to object when we have anything that isn't an Index.

Thanks, then this is totally a reasonable change ;)
Can you add a whatsnew note about it? I would call it a bug fix.

[jreback]

very nice!