Custom sorting of plot series

import pandas as pd
import numpy as np
from pandas.api.types import CategoricalDtype
from plotnine import (
    ggplot,
    aes,
    geom_bar,
    coord_flip,
    labs,
    scale_x_discrete
)
from plotnine.data import mpg

Bar plot of manufacturer - Default Output

(ggplot(mpg)
 + aes(x='manufacturer')
 + geom_bar(size=20)
 + coord_flip()
 + labs(y='Count', x='Manufacturer', title='Number of Cars by Make')
)

<Figure Size: (640 x 480)>

Bar plot of manufacturer - Ordered by count (Categorical)

By default the discrete values along axis are ordered alphabetically. If we want a specific ordering we use a pandas.Categorical variable with categories ordered to our preference.

# Determine order and create a categorical type
# Note that value_counts() is already sorted
manufacturer_list = mpg['manufacturer'].value_counts().index.tolist()
manufacturer_cat = pd.Categorical(mpg['manufacturer'], categories=manufacturer_list)

# assign to a new column in the DataFrame
mpg = mpg.assign(manufacturer_cat = manufacturer_cat)

(ggplot(mpg)
 + aes(x='manufacturer_cat')
 + geom_bar(size=20)
 + coord_flip()
 + labs(y='Count', x='Manufacturer', title='Number of Cars by Make')
)

<Figure Size: (640 x 480)>

We could also modify the existing manufacturer category to set it as ordered instead of having to create a new CategoricalDtype and apply that to the data.

mpg = mpg.assign(manufacturer_cat =
  mpg['manufacturer'].cat.reorder_categories(manufacturer_list))

Bar plot of manufacturer - Ordered by count (limits)

Another method to quickly reorder a discrete axis without changing the data is to change it's limits

# Determine order and create a categorical type
# Note that value_counts() is already sorted
manufacturer_list = mpg['manufacturer'].value_counts().index.tolist()

(ggplot(mpg)
 + aes(x='manufacturer_cat')
 + geom_bar(size=20)
 + scale_x_discrete(limits=manufacturer_list)
 + coord_flip()
 + labs(y='Count', x='Manufacturer', title='Number of Cars by Make')
)

<Figure Size: (640 x 480)>

You can 'flip' an axis (independent of limits) by reversing the order of the limits.

# Determine order and create a categorical type
# Note that value_counts() is already sorted
manufacturer_list = mpg['manufacturer'].value_counts().index.tolist()[::-1]

(ggplot(mpg)
 + aes(x='manufacturer_cat')
 + geom_bar(size=20)
 + scale_x_discrete(limits=manufacturer_list)
 + coord_flip()
 + labs(y='Count', x='Manufacturer', title='Number of Cars by Make')
)

<Figure Size: (640 x 480)>

Further Reading

• Pandas documentation of how to use categorical data in practice

• Pandas API Reference for categorical

• Pandas documentation of pd.Categorical