multiple piecharts, one legend-CodePudding

I've created a piechart for 4 different dataframes (the four investment portfolio's in my household).

fig=plt.figure(figsize=(20,20), dpi= 100, facecolor='w', edgecolor='k')
fig, axs = plt.subplots(2, 2)
axs[0, 0].pie(m.newvalue, labels = m.name, labeldistance=None)
axs[0, 0].set_title('Axis [0, 0]')
axs[0, 1].pie(j.newvalue, labels = j.name, labeldistance=None)
axs[0, 1].set_title('Axis [0, 1]')
axs[1, 0].pie(h.newvalue, labels = h.name, labeldistance=None)
axs[1, 0].set_title('Axis [1, 0]')
axs[1, 1].pie(x.newvalue, labels = x.name, labeldistance=None)
axs[1, 1].set_title('Axis [1, 1]')

for ax in axs.flat:
    ax.set()

axs[0,0].legend(bbox_to_anchor=(0, 1.5))

The problem is that the legend only shows the composition of the first dataframe. But it should represent the components for the other 3 piecharts as well. Most of the components are the same, but every df has a unique asset as well. The green in [1,0] is something else than the green in [0,0].

How do I do this?

CodePudding user response：

Here is an approach:

make a set containing all the names
make a dictionary to map each unique name to a color
use the dictionary to map the names in each pie to the corresponding color
use the dictionary to create the appropriate legend

import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

m = pd.DataFrame({'newvalue': np.random.randint(10, 20, 4), 'name': [*'ABDE']})
j = pd.DataFrame({'newvalue': np.random.randint(10, 20, 4), 'name': [*'ACEF']})
h = pd.DataFrame({'newvalue': np.random.randint(10, 20, 4), 'name': [*'BDEG']})
x = pd.DataFrame({'newvalue': np.random.randint(10, 20, 4), 'name': [*'ABCE']})

# let colors be a list of unique colors, at least one for each name
colors = plt.get_cmap('tab10').colors
# make a set of all the names
all_names = {*m.name, *j.name, *h.name, *x.name}
# map each of the unique names to a color
name_to_color = {name: color for name, color in zip(all_names, colors)}

fig, axs = plt.subplots(2, 2, figsize=(9, 12), dpi=100, facecolor='w', edgecolor='k')
axs[0, 0].pie(m.newvalue, labels=m.name, labeldistance=None, colors=m.name.map(name_to_color))
axs[0, 0].set_title('Axis [0, 0]')
axs[0, 1].pie(j.newvalue, labels=j.name, labeldistance=None, colors=j.name.map(name_to_color))
axs[0, 1].set_title('Axis [0, 1]')
axs[1, 0].pie(h.newvalue, labels=h.name, labeldistance=None, colors=h.name.map(name_to_color))
axs[1, 0].set_title('Axis [1, 0]')
axs[1, 1].pie(x.newvalue, labels=x.name, labeldistance=None, colors=x.name.map(name_to_color))
axs[1, 1].set_title('Axis [1, 1]')

handles = [plt.Rectangle((0, 0), 0, 0, color=name_to_color[name], label=name) for name in name_to_color]
axs[0, 0].legend(handles=handles, bbox_to_anchor=(0.2, 1.1), loc='lower left')

plt.tight_layout()
plt.show()

PS: plt.figure() should not be used together with plt.subplots(), as plt.subplots() also creates a new figure, leading to a dummy empty plot. The parameters for plt.figure() can directly go to plt.subplots().

When creating a legend with bbox_to_anchor it is important to also set the loc. The loc is the point on the legend where the anchor is placed. Default, loc='best', which depends on the subplot contents, and can be different e.g. when the subplot is a bit smaller or larger.