API: downcating on fillna no-ops intentional?

[jbrockmendel]

ser = pd.Series([1, 2, 3], dtype=object)

>>> ser.fillna(4).dtype
dtype('int64')

Better to retain dtype in these cases?

[jreback]

this was the point of the downcast kw

as we normally will cast object if possible

[jorisvandenbossche]

IMO we shouldn't retain or not retain dtype based on whether it was in practice a no-op or not. In which case the discussion is about downcasting in general, like:

In [31]: ser = pd.Series([1, 2, 3, np.nan], dtype=object)

In [32]: ser.fillna(4)
Out[32]: 
0    1
1    2
2    3
3    4
dtype: int64

[jbrockmendel]

closed by #45673