DEPR: Deprecate dtype inference on pandas objects #56244
Conversation
|
There was some weird additional inference going on in cat, which was uncovered by the deprecation, is fixed now but doesn't really need a whatsnew since we merged another fix today as well |
|
cc @mroeschke good to merge here? |
pandas/core/series.py
Outdated
| "constructor will keep the original dtype in the future. " | ||
| "Call ``infer_objects on the result", | ||
| FutureWarning, | ||
| stacklevel=2, |
There was a problem hiding this comment.
Does find_stack_level return the incorrect result here?
There was a problem hiding this comment.
This is nice for internal debugging, but yeah I can use find_stack_level as well
Co-authored-by: Matthew Roeschke <[email protected]>
Co-authored-by: Matthew Roeschke <[email protected]>
Co-authored-by: Matthew Roeschke <[email protected]>
pandas/core/frame.py
Outdated
| "Dtype inference on a pandas object " | ||
| "(Series, Index, ExtensionArray) is deprecated. The Index " | ||
| "constructor will keep the original dtype in the future. " | ||
| "Call `infer_objects` on the result", |
| if original_dtype is None and is_pandas_object and data_dtype == np.object_: | ||
| if self.dtypes.iloc[0] != data_dtype: | ||
| warnings.warn( | ||
| "Dtype inference on a pandas object " |
There was a problem hiding this comment.
"in the DataFrame constructor" somewhere in this sentence? I think the "The Index constructor" below is a copy/paste leftover
|
Should this affect pd.array? |
Yes, but it already preserves the dtype, so we should be good there |
| @@ -460,6 +460,7 @@ Other Deprecations | |||
| - Deprecated behavior of :meth:`Index.insert` with an object-dtype index silently performing type inference on the result, explicitly call ``result.infer_objects(copy=False)`` for the old behavior instead (:issue:`51363`) | |||
| - Deprecated casting non-datetimelike values (mainly strings) in :meth:`Series.isin` and :meth:`Index.isin` with ``datetime64``, ``timedelta64``, and :class:`PeriodDtype` dtypes (:issue:`53111`) | |||
| - Deprecated downcasting behavior in :meth:`Series.where`, :meth:`DataFrame.where`, :meth:`Series.mask`, :meth:`DataFrame.mask`, :meth:`Series.clip`, :meth:`DataFrame.clip`; in a future version these will not infer object-dtype columns to non-object dtype, or all-round floats to integer dtype. Call ``result.infer_objects(copy=False)`` on the result for object inference, or explicitly cast floats to ints. To opt in to the future version, use ``pd.set_option("future.no_silent_downcasting", True)`` (:issue:`53656`) | |||
| - Deprecated dtype inference in :class:`Index`, :class:`Series` and :class:`DataFrame` constructors when giving a pandas input, call ``.infer_objects`` on the input to keep the current behavior (:issue:`56012`) | |||
There was a problem hiding this comment.
on the input -> on the result?
There was a problem hiding this comment.
either way is totally fine
| @@ -20,7 +20,7 @@ def test_between(self): | |||
| tm.assert_series_equal(result, expected) | |||
|
|
|||
| def test_between_datetime_object_dtype(self): | |||
| ser = Series(bdate_range("1/1/2000", periods=20).astype(object)) | |||
| ser = Series(bdate_range("1/1/2000", periods=20).astype(object), dtype=object) | |||
There was a problem hiding this comment.
The .astype(object) here is redundant
| ) | ||
| with tm.assert_produces_warning(FutureWarning, match="Dtype inference"): | ||
| ser = Series( | ||
| period_range("20130101", periods=5, freq="D", name="xxx").astype(object) |
There was a problem hiding this comment.
can you construct the index outside the context so it is obvious where the warning comes from
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.This is very weird to begin with and becomes a real PITA when we infer strings as arrow backed strings