Seaborn's pairplot is very slow when Matplotlib.interactive(true)

[mrkn]

On this Ruby script:

# pairplot.rb
require 'matplotlib'
Matplotlib.interactive(true) if ARGV[0] == 'interactive'
Matplotlib.rcParams['backend'] = 'module://ruby.matplotlib.backend_inline'
require 'matplotlib/pyplot'
require 'pandas'
sns = PyCall.import_module('seaborn')
df = Pandas.read_csv('data/raw/iris.csv', header: nil, names: [
  'sepal_length', 'sepal_width', 'petal_length', 'petal_width', 'class'
])
t = Time.now
sns.pairplot(df, hue: 'class')
d = Time.now - t
puts "#{d} sec"
Matplotlib::Pyplot.savefig('pairplot.png')

sns.pairplot is very slow only if Matplotlib.interactive(true) is called.

$ ruby pairplot.rb
1.663664 sec
$ ruby pairplot.rb interactive
225.765168 sec

But on the following Python script:

# pairplot.py
import matplotlib
matplotlib.rcParams['backend'] = 'module://ipykernel.pylab.backend_inline'
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
import time
df = pd.read_csv('data/raw/iris.csv', header=None, names=[
  'sepal_length', 'sepal_width', 'petal_length', 'petal_width', 'class'
])
t = time.time()
sns.pairplot(df, hue='class')
d = time.time() - t
print("{} sec".format(d))
plt.savefig('pairplot.png')

sns.pairplot consumes less than 2 secs even though matplotlib.interactive(True) is called.

$ python pairplot.py
1.7206127643585205 sec
$ python pairplot.py interactive
1.6132988929748535 sec

This may be a bug of the backend of matplotlib.rb.

[adnanaliarshad]

no solution for this?

[mrkn]

Use Matplotlib.interactive(false).