BUG: Series.fillna is altering None values of unspecified columns

[RSchweiger]

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of pandas (1.0.5)

series = pd.Series({'1':1,'2':None,'3':np.nan, '4':''})

will create a series which looks like:

1       1
2    None
3     NaN
4        
dtype: object

When using the fillna method on this series on index 3 it changes also the entry in index 2!

series.fillna({'3':0})

This will result in a output as follows:

1      1
2    NaN
3      0
4       
dtype: object

Problem:
The value at index 2 changed from None to NaN.

Expected Behaviour:
The value at index 2 should never be touched when using fillna and specifying the indices you want to alter.

[mzeitlin11]

Hi @RSchweiger, thanks for the clear report! I can reproduce this on current master, investigations welcome to fix.

[jreback]

I don't buy this is a bug at all: #40509 (review)

[ghost]

@jreback @RSchweiger @mzeitlin11 I saw that this is still reproducible on Master, labeled as a 'Bug', and Unassigned. I was wondering if I could work on this?

[mzeitlin11]

Go ahead! Commentary in #40509 might be helpful.

[ghost]

take

[simonjayhawkins]

removed milestone