ENH: infer_objects copy kwd #50096
ENH: infer_objects copy kwd #50096
Conversation
mroeschke
added
Dtype Conversions
|
Thanks @jbrockmendel |
| def convert(self: T, copy: bool) -> T: | ||
| def _convert(arr): | ||
| if is_object_dtype(arr.dtype): | ||
| # extract PandasArray for tests that patch PandasArray._typ | ||
| arr = np.asarray(arr) | ||
| return lib.maybe_convert_objects( | ||
| result = lib.maybe_convert_objects( | ||
| arr, | ||
| convert_datetime=True, | ||
| convert_timedelta=True, | ||
| convert_period=True, | ||
| ) | ||
| if result is arr and copy: | ||
| return arr.copy() | ||
| return result | ||
| else: | ||
| return arr.copy() | ||
| return arr.copy() if copy else arr |
There was a problem hiding this comment.
@jbrockmendel sorry I was just about to ask - is there a test that hits this?
if not, then is it because the plan is to "kill arraymanager" anyway?
There was a problem hiding this comment.
is there a test that hits this?
TestInferObjects.test_copy should hit this in the CI build specific to ArrayManager
the plan is to "#48880 (comment)" anyway
I'd like this, but have no plans to push for it.
|
@jbrockmendel for 1.5 we had a discussion related to new |
|
Two big motivations. One was that we removed automatic inference in a few places, telling users to use infer_objects instead. Wanted to allow that operation to be cheap where possible. Second is that there was a bug in here in Block[Subclass].convert that was inconsistent about making copies. Fixing that bug required making more copies, and it's worth softening that blow. |
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.