BUG: pd.sort_values() not working on categorical column

[BelaNeuendorf]

BUG: pd.sort_values() not working on categorical column

[ahmedembeddedx]

import pandas as pd

# Create DataFrame
df = pd.DataFrame([[1, 2, 'March'],[5, 6, 'Dec'],[3, 4, 'April']], columns=['a','b','month'])

# Define custom_dict
custom_dict = {'March': 0, 'April': 1, 'Dec': 3}

# Convert 'month' column to Categorical with custom order
df['month'] = pd.Categorical(df['month'], categories=custom_dict.keys())

# Sort DataFrame based on the custom order
df.sort_values(by='month', key=lambda x: x.map(custom_dict))

[rtlee9]

To be more specific, sort_values doesn't respect the key parameter for categorical series.

Sorting does work as expected if the categorical is ordered using the ordered param:

import pandas as pd

df = pd.DataFrame([[1, 2, 'March'],[5, 6, 'Dec'],[3, 4, 'April']], columns=['a','b','month'])
df['month_cat_ordered'] = pd.Categorical(df.month, ['March', 'April', 'Dec'], ordered=True)
df.sort_values('month_cat_ordered')

	a	b	month	month_cat_ordered
0	1	2	March	March
2	3	4	April	April
1	5	6	Dec	Dec

Mapping an unordered categorical as in the original example maps the old categories to the new categories but doesn't reorder the new categories. This is possible to implement this reordering in Categorical.map but would come at a performance cost and it isn't immediately clear it's worth it.