Use empty list as fill value for Series.fillna

Question

I need to change the NaN values the column 'FocusColumn' to empty lists.

---

I tried to do it like this:

df['FocusColumn'].fillna([],inplace=True)

But it throws me the following error:

337         if validate_scalar_dict_value and isinstance(value, (list, tuple)):
338             raise TypeError( 
339                 '"value" parameter must be a scalar or dict, but 
340                 f'you passed a "{type(value).__name__}"'
341             )
TypeError: "value" parameter must be a scalar or dict, but you passed a "list"

What is the correct way to do it?

Answer 1

Looks like you can work around this by supplying a dict to Series.fillna.

>>> df
   FocusColumn
0          NaN
1          1.0
2          NaN
>>> df['FocusColumn'].fillna({i: [] for i in df.index})
0    []
1     1
2    []
Name: FocusColumn, dtype: object

Notes:

1. Surprisingly, this only works for Series.fillna.
2. For large Series with few missing values, this might create an unreasonable amount of throwaway empty lists.
3. Tested with pandas 1.0.5.