Closed
Description
In the context of #23192 (and #23604 / #23606), I want to use pandas.core.dtypes.cast.maybe_upcast_putmask, because it solves exactly the problem I need it to solve.
Unfortunately, it does not work as advertised (and I already found the culprit).
The docstring says:
def maybe_upcast_putmask(result, mask, other):
"""
A safe version of putmask that potentially upcasts the result
Parameters
----------
result : ndarray
The destination array. This will be mutated in-place if no upcasting is
necessary.
mask : boolean ndarray
other : ndarray or scalar
The source array or value
in other words, it expects result and other to be ndarrays. Curiously enough, in some branches, it only works for Series and produces wrong results for ndarray, e.g.
>>> import pandas as pd
>>> import numpy as np
>>> s = pd.Series([10, 11, 12])
>>> t = pd.Series([np.nan, 61, np.nan])
>>> from pandas.core.dtypes.cast import maybe_upcast_putmask
>>> result, _ = maybe_upcast_putmask(s, np.array([False, True, False]), t)
>>> result # correct
0 10
1 61
2 12
dtype: int64
>>> result, _ = maybe_upcast_putmask(s.values, np.array([False, True, False]), t.values)
>>> result # incorrect
array([10., nan, 12.])
This is because the code does
try:
[...]
new_result = result.values.copy()
[...]
return [...]
except:
# do something else
which actually expects a Series (since .values won't ever work on an ndarray).