reverse the order of CategoricalDtype in Pandas

Question

I have an ordered categorical variable in my dataframe like the following:

CategoricalDtype(categories=['UNDER $1 000', '$1 000 TO 2 999', '$3 000 TO 3 999', 
                              '$90000 - $109999', '$110000 OR OVER', 'REFUSED'], 
                               ordered=True)

For CategoricalIndex in a dataframe I know I can do the following:

df.sort_index(ascending=False, inplace=True)

I tried the method I find here for the CategoricalDtype object:

from pandas import Categorical
Categorical.sort(ascending=False)

but it doesn't work and returned:

AttributeError: type object 'Categorical' has no attribute 'sort'.

Thus, I wonder if there is an easy way to reverse the order for CategoricalDtype with Pandas.

Answer 1

If you are only looking for a CategoricalIndex object with the reversed list of categories, you might consider creating a new object with the reversed list of categories from your original instance.

if x is your existing CategoricalDtype object:

x_reversed = CategoricalDtype(categories=reversed(x.categories), ordered=True)

Answer 2

I just discovered another way to do it.

Since CategoricalDtype in pandas has an attribute cat.categories, we can call it from a variable right away and reserve its order directly by using reversed() or [::-1]. If a pandas Series is categorical, pandas also offers lots of methods like cat.set_categories. Thus, for a variable named var in the dataframe, we can do the following:

order = df.var.cat.categories[::-1]
df.var.cat.set_categories(new_categories=order, ordered=True, inplace=True)

By using this method, we don't have to create a CategoricalDtype object or importing anything. It also saves us from using .astype() to replace the old order in the categorical variable as well.