CLN: shift with freq and fill_value, to_pydatetime returning Series #57425
Conversation
pandas/io/sql.py
Outdated
| warnings.filterwarnings("ignore", category=FutureWarning) | ||
| # GH#52459 to_pydatetime will return Index[object] | ||
| d = np.asarray(ser.dt.to_pydatetime(), dtype=object) | ||
| d = ser.dt.to_pydatetime().to_numpy(dtype=object) |
There was a problem hiding this comment.
@phofl what's the way to make this writable later on (d[mask] = None below)?
There was a problem hiding this comment.
Is
ser.dt.to_pydatetime()
ever not backed by a NumPy array? If it's always a NumPy array I would just do
ser.dt.to_pydatetime()._values
Otherwise you can do
d.flags.writeable = True, but this is a bit tricky, because this won't work if d was initially a PyArrow array
There was a problem hiding this comment.
In this branch, ser was backed by ArrowExtensionArray so it's initially a pyarrow array
There was a problem hiding this comment.
But the result of to_pydatetime is always backed by numpy, correct?
|
Going to merge this one. Can follow up if needed |
…andas-dev#57425) * Enforce shift with freq and fill_value * Enforce to_pydatetime change * Change doc example * Use to_numpy * Use _values * Fix doctest
I am opening a PR #57469 to add future warning when calling |
…andas-dev#57425) * Enforce shift with freq and fill_value * Enforce to_pydatetime change * Change doc example * Use to_numpy * Use _values * Fix doctest
xref #52459, #54818