Bar charts · Geoff Ruddock

This notebook contains boilerplate code for producing various types of bar charts, using various python plotting packages incl. pandas, matplotlib, and seaborn.

Generally:

Plotting from pandas involves simply calling df.plot() with some variation of parameters.
Plotting directly from matplotlib gives less polished of an output (out-of-the-box), and is much more verbose, but provides infinite customization.
seaborn is the least verbose and most beautiful, but often expects the input data to be in an unintuitive format.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import bokeh

Package     Version
----------  ---------
python      3.9.12
matplotlib  3.5.1
seaborn     0.11.2
bokeh       2.4.2

1D

data_1d = pd.Series([0.2, 0.3, 0.1, 0.4], index=['A', 'B', 'C', 'D'])
data_1d

A    0.2
B    0.3
C    0.1
D    0.4
dtype: float64

Vertical

pandas

data_1d.plot(kind='bar');

png

matplotlib

plt.bar(x=range(4), height=data_1d);

png

seaborn

seaborn.barplot() expects a DataFrame as input, with x, y args specifying the column names of said data.

data_1d_df = data_1d.rename('pct').reset_index()

_ = sns.barplot(x='index', y='pct', data=data_1d_df)

png

Horizontal

pandas

data_1d.plot(kind='barh');

png

matplotlib

plt.barh(y=range(4), width=data_1d);

png

seaborn

Error bars

matplotlib

Different ways of specifying error bars — Matplotlib 3.5.2 documentation

plt.barh(y=range(4), width=data_1d, xerr=0.05);

png

2D

Bar charts with multiple lines per group.

data_2d = pd.DataFrame({
    'X': [0.2, 0.3, 0.1, 0.4],
    'Y': [0.4, 0.2, 0.3, 0.1]
    }, index=['A', 'B', 'C', 'D'])

data_2d

	X	Y
A	0.2	0.4
B	0.3	0.2
C	0.1	0.3
D	0.4	0.1

Normal

pandas

data_2d.plot(kind='bar');

png

matplotlib

fig, ax = plt.subplots()

labels = data_2d.index.tolist()
x = np.arange(len(labels))
width = 0.35

_ = ax.bar(x-width/2, data_2d['X'], width, label='X')
_ = ax.bar(x+width/2, data_2d['Y'], width, label='Y')

_ = ax.set_xticks(x)
_ = ax.set_xticklabels(labels)

plt.legend();

png

seaborn

to_viz = (
    data_2d
    .stack()
    .rename_axis(('group', 'metric'))
    .rename('val')
    .reset_index()
)

#to_viz

_ = sns.barplot(x='group', y='val', hue='metric', data=to_viz)

png

Stacked

pandas

data_2d.plot(kind='bar', stacked=True);

png

Gotchas with pandas

Put grid behind main elements

The DataFrame.plot() function takes a grid=True argument, but by default it displays above the bars/lines—which looks terrible.

data_1d.plot(kind='bar', grid=True);

png

You can fix this by assigning the return value to ax, then running ax.set_axisbelow(True) afterwards.

ax = data_1d.plot(kind='bar', grid=True) # or ax.grid()
ax.set_axisbelow(True)

png