How to add hatch to custom legend in matplotlib?

user_5670296 0
Viewed 2112

Edit: FWIW I am fixing the legend in Illustrator, but would prefer to do it automatically because I have about 20 of these plots to make

I followed this very helpful answer for adding a custom legend to a matplotlib violinplot, which doesn't support its own legend. It worked great except when I tried to add hatches.

Here's my code for the labels (I tried adding patch two different ways):

    labels = [ 'Low entropy bin', 'Medium entropy bin', 'High entropy bin' ]
    legend_patches = 3*[matplotlib.patches.Patch( color='#DCDCDC', hatch='//' )]
    for i in legend_patches:
       i.set_hatch( '//' )

The code for hatching the violins themselves works fine:

parts = plt.violinplot( data, showmeans=False, showextrema=True, showmedians=True )

hatch_dict = { 0:'', 1:'///', 2:'xx' }

for t in range(0, 3):
    third = range( 0, len( labels ) )[ t*(int(len(labels)/3)):(((t+1)*int(len(labels)/3))) ]
    for i in third:
        face = parts['bodies'][i]
        face.set_hatch( hatch_dict[t] )

enter image description here

The data the bins relate to (not shown) is already color coded for other categories so I'd really like to show the bins in different hatches.

matplotlib python
1 Answers

You are almost there - just need to be careful with the color argument with patches. There are two sub-components: an edge (edgecolor) and a face (facecolor); with this patch artist setting the color= defines both of these colors. The hatch and the background then come out the same color and you can't see one from the other.

Bottom line: use something like this for your patch constructor:

p = matplotlib.patches.Patch(facecolor='#DCDCDC', hatch=hatch_dict[i])

hatches

Full code for this plot:

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches

# generate some data
n = 50
sigmas = np.array([0.1, 0.05, 0.15])
means  = np.array([0.2, 0.5, 0.75])
data = sigmas * np.random.randn(n, 3) + means

labels = [ 'Low entropy bin', 'Medium entropy bin', 'High entropy bin' ]

parts = plt.violinplot( data, showmeans=False, showextrema=True, showmedians=True)

# set up color and hatching on the violins
hatch_dict = { 0:'', 1:'///', 2:'xx' }
for i, face in enumerate(parts['bodies']):
    face.set_hatch(hatch_dict[i])
    face.set_facecolor('#DCDCDC')

# for completeness update all the lines (you already had this styling applied)
for elem in ['cbars', 'cmedians', 'cmaxes', 'cmins']:
    parts[elem].set_edgecolor('0.5')

# construct proxy artist patches
leg_artists = []
for i in range(len(hatch_dict)):
    p = matplotlib.patches.Patch(facecolor='#DCDCDC', hatch=hatch_dict[i])
    # can also explicitly declare 2nd color like this
    #p = matplotlib.patches.Patch(facecolor='#DCDCDC', hatch=hatch_dict[i], edgecolor='0.5')

    leg_artists.append(p)

# and add them to legend.
ax = plt.gca()
ax.legend(leg_artists, labels, loc='upper left')
user_16439460
Related