ValueError: cannot use inplace with CategoricalIndex

Question

I am using pandas 0.18.

This fails

 cat_fields[f[0]].add_categories(s,inplace=True)

However the docs say

inplace : boolean (default: False)

Whether or not to add the categories inplace or return a copy of this categorical with added categories.

Am I missing something ?

I am creating a superset of categories/columns across many dataframes to eventually be able to concatenate them.

My error:

ValueError: cannot use inplace with CategoricalIndex

Answer 1

I think you need assign to original column, because Series.add_categories has inplace parameter and it works nice.

But in CategoricalIndex.add_categories has also inplace parameter, but it fails. I think it is bug.

cat_fields[f[0]] = cat_fields[f[0]].add_categories(s)

or:

cat_fields[f[0]] = cat_fields[f[0]].cat.add_categories(s)

Sample:

cat_fields = pd.DataFrame({'A':[1,2,3]}, index=['a','d','f'])


cat_fields.index = pd.CategoricalIndex(cat_fields.index)
cat_fields.A = pd.Categorical(cat_fields.A)
print (cat_fields)
   A
a  1
d  2
f  3

s = ['b','c']

cat_fields.A.cat.add_categories(s,inplace=True)

print (cat_fields.A)
Name: A, dtype: category
Categories (5, object): [1, 2, 3, b, c]


cat_fields.index.add_categories(s,inplace=True)
print (cat_fields)

ValueError: cannot use inplace with CategoricalIndex