matplotlib-tufte -- styles from Tufte's "Visual Display Of Quantitative Information" for matplotlib

matplotlib-tufte styles figures to make them easier on the eyes, and keep only essential ink.

Based on : Tufte, Edward. The Visual Display of Quantitative Information. 2nd ed., Graphics Press, 2007.

Installation

python -m pip install "matplotlib_tufte @ git+https://github.com/ninivert/matplotlib_tufte.git"

Usage

setup tries to load the New Computer Modern font (comes packaged with matplotlib_tufte) and the tufte.mplstyle stylesheet.

Changes made by the custom style :

• better readability using a serif font
• New Computer Modern (used to be : Latin Modern Roman) font
  • applied to all text (ticks, labels, legends, titles, etc.)
  • in math mode, gives LaTeX-like math results, without needing a LaTeX installation
• smaller figures (increase data density)
  • figsize is defaulted to (4, 3)
  • tight layout applied automatically
• thinner axes to put emphasis on data ink
• remove chartjunk and non-data ink
  • remove top and right spines
  • remove legend fancybox and frame
• insert black as the first color of the cycler

Documentation for breathe

Add some space between the axes and the spines

Parameters
----------
ax : matplotlib.axes.Axes
which : str or AxisWhich, optional
        which axis to apply the spacing on, by default AxisWhich.BOTH
pad_frac_start : float, optional
        space to add to the start of the axis, as a fraction of the data span, by default 0.04
pad_frac_end : float, optional
        space to add to the end of the axis, as a fraction of the data span, by default 0.04

Documentation for data_lim

Sets the axis to use the limits of the data

Parameters
----------
ax : matplotlib.axes.Axes
which : str or AxisWhich, optional
        which axis to apply the spacing on, by default AxisWhich.BOTH

In other cases, you might just need to get your hands dirty, using methods such as

• ax.spines.bottom.set_visible(False) to disable the bottom line
• ax.set_xticks(...) and ax.set_xticklabels(...) to customize ticks
• ax.set_xlim(...) and ax.set_xticklabels(...) to have nice integer limits. Use this in combination with breathe if you were used to matplotlib adding padding to the limits

Examples

See also the demo notebook

from matplotlib_tufte import *
setup()
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np

x = np.linspace(1, 9, 100)
y1 = np.exp(-(x-2)) - np.sin(x-1) + 2
y2 = 5 * 2/x * np.exp(-(np.log(x-0.99)-0.5)**2)

plt.close('all')
fig, ax = plt.subplots()
ax.set_xlabel('$x$')
ax.set_ylabel('$y$')
ax.plot(x, y1, label='data 1')
ax.plot(x, y2, label='data 2')
ax.legend()
plt.show()
fig.savefig('fig1.png')

Letting the axis breathe

breathe(ax)

Setting data limits

data_lim(ax)

Getting your hands dirty

np.random.seed(0)
N = 50
sample1 = np.random.lognormal(0.9, 0.5, N)
sample2 = np.random.lognormal(0.2, 0.7, N)

means = [np.mean(sample1), np.mean(sample2)]
stds = [np.std(sample1) * np.sqrt(N/(N-1)), np.std(sample2) * np.sqrt(N/(N-1))]  # sample stdev
x = [0, 1]

fig, ax = plt.subplots(figsize=(2, 3))
for x_, sample, m, s in zip(x, (sample1, sample2), means, stds):
        ax.plot([x_]*len(sample), sample, linestyle='', marker='_', color='k')
        ax.hlines(np.mean(sample), x_-0.3, x_+0.3, linewidth=0.5, color='k')
        ax.add_patch(mpl.patches.FancyArrowPatch(
                arrowstyle='|-|,widthA=2.0,widthB=2.0',
                linewidth=0.5,
                path=mpl.path.Path([(x_-0.3, m-s), (x_-0.3, m+s)]))
        )
ax.set_xticks(x)
ax.set_xticklabels(['S1', 'S2'])
ax.set_xlim((-1, 2))
ax.set_ylim((0, 8))
ax.set_ylabel('measurement')
breathe(ax, 'y')
ax.spines.bottom.set_visible(False)
plt.show()

It is also possible to use matplotlib_tufte without using setup

# demo/demo_nosetup.py

from matplotlib_tufte import *
import matplotlib.pyplot as plt

# by default, will use gca
# this is however discouraged
plt.plot([0,1,2],[-2,4,5])
despine()
data_lim()
breathe()
plt.savefig('fig_nosetup.png')

Changelog

v0.2: changed default font from "Latin Modern Roman" to "New Computer Modern", which is more readable at smaller sizes.

TODO

• Range frames
• Data ticks
• Axis histograms