BUG: Fix mixed datetime dtype inference #33749
Conversation
jreback
added
Compat
|
is the docbuild failure real or a fluke? |
Nevermind, was a typo in whatsnew that caused it |
pandas/core/indexes/datetimes.py
Outdated
| other = DatetimeIndex(other) | ||
| try: | ||
| other = DatetimeIndex(other) | ||
| except OutOfBoundsDatetime: |
There was a problem hiding this comment.
what is hitting this? why are we not bubbling this up?
There was a problem hiding this comment.
Cases where we have a datetime.date that's out of bounds for DatetimeIndex (e.g. https://github.com/pandas-dev/pandas/pull/33749/files#diff-0c5fadf83f91d24a931a051d82801917R569) are raising here instead of returning -1 which I believe is a bug on master (that's what @jschendel and I were talking about above). I don't think we'd want this to actually raise since it doesn't seem to be an essential step.
| ) | ||
| def test_infer_dtype_date_order_invariant(self, values): | ||
| # https://github.com/pandas-dev/pandas/issues/33741 | ||
| result = lib.infer_dtype(values, skipna=True) |
There was a problem hiding this comment.
add NaT and parameterize on skipna as well
|
needs a rebase as well |
pandas/core/indexes/base.py
Outdated
| @@ -4662,7 +4662,11 @@ def _maybe_promote(self, other: "Index"): | |||
| """ | |||
|
|
|||
| if self.inferred_type == "date" and isinstance(other, ABCDatetimeIndex): | |||
| return type(other)(self), other | |||
| try: | |||
| maybe_casted = type(other)(self) | |||
There was a problem hiding this comment.
write this as
try:
return type(other)(self)
except OutOfBoundsDatetime:
pass
return Index(self), other
as it needs to copy
pandas/_libs/lib.pyx
Outdated
| return "datetime" | ||
| if is_date_array(values, skipna=skipna): |
|
@jreback ping |
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff