When the dataframe has duplicate columns, it seems that fillna function cannot work correctly with dict parameter

Question

I find that after using pd.concat() to concatenate two dataframes with same column name, then df.fillna() will not work correctly with the dict parameter specifying which value to use for each column.

I don't know why? Is something wrong with my understanding?

a1 = pd.DataFrame({'a': [1, 2, 3]})
a2 = pd.DataFrame({'a': [1, 2, 3]})
b = pd.DataFrame({'b': [np.nan, 20, 30]})
c = pd.DataFrame({'c': [40, np.nan, 60]})
x = pd.concat([a1,a2, b, c], axis=1)
print(x)
x = x.fillna({'b':10, 'c': 50})
print(x)

Initial dataframe:

   a  a     b     c
0  1  1   NaN  40.0
1  2  2  20.0   NaN
2  3  3  30.0  60.0

Data is unchanged after df.fillna():

   a  a     b     c
0  1  1   NaN  40.0
1  2  2  20.0   NaN
2  3  3  30.0  60.0

Answer 1

As mentioned in the comments, there's a problem assigning values to a dataframe in the presence of duplicate column names. However, you can use this workaround:

for col,val in {'b':10, 'c': 50}.items():
    new_col = x[col].fillna(val)
    idx = int(x.columns.get_loc(col))
    x = x.drop(col,axis=1)
    x.insert(loc=idx, column=col, value=new_col)

print(x)

result:

   a  a     b     c
0  1  1  10.0  40.0
1  2  2  20.0  50.0
2  3  3  30.0  60.0