

In [11]:

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

import seaborn
%matplotlib inline

Setting nice axes labels in matplotlib

In this post I want to collect some ideas I had on setting nice labels in matplotlib. In particular, for scientific papers we usually want a label like "time [s]". Then, if the data is very large we may put the exponent next to the units like "time [$10^{-6}$s]", better still is to use an SI prefix e.g "time [$\mu$s]".

The defaults

Matplotlib already has useful routines to format the ticks, but it usually puts the exponent somewhere near to the top of the axis. Here is a typical example using the defaults



In [10]:

    
x = np.linspace(0, 10, 1000)
y = 1e10 * np.sin(x)

fig, ax = plt.subplots()
ax.plot(x, y)
plt.show()

Improving on the defaults

For a scientific publication we tend to use $\times10^{10}$ or $10^{10}$, this can easily be achieved with the ScalarFormatter. There is an example on the docs, but here is a simpler one:



In [25]:

    
x = np.linspace(0, 10, 1000)
y = 1e8 + 1e6 * np.sin(x)

fig, ax = plt.subplots()
ax.yaxis.set_major_formatter(mpl.ticker.ScalarFormatter(useMathText=True, useOffset=False))
ax.plot(x, y)
plt.show()

Note there are two options set here, the first useMathText forces the scientific notation, while the useOffset stops the offset being used which makes most plots illegible in my opinion

Customing the results

At some point I ran into an issue with the location of where the exponent was positioned, it was overlapping with other subplots. I did the usual trick and went to StackOverflow and was helpfully redirected to this post which by chance was posted the same day. Building on the ideas posted there I have developed the following:



In [2]:

    
def update_label(old_label, exponent_text):
    if exponent_text == "":
        return old_label
    
    try:
        units = old_label[old_label.index("[") + 1:old_label.rindex("]")]
    except ValueError:
        units = ""
    label = old_label.replace("[{}]".format(units), "")
    
    exponent_text = exponent_text.replace("\\times", "")
    
    return "{} [{} {}]".format(label, exponent_text, units)
    
def format_label_string_with_exponent(ax, axis='both'):  
    """ Format the label string with the exponent from the ScalarFormatter """
    ax.ticklabel_format(axis=axis, style='sci')

    axes_instances = []
    if axis in ['x', 'both']:
        axes_instances.append(ax.xaxis)
    if axis in ['y', 'both']:
        axes_instances.append(ax.yaxis)
    
    for ax in axes_instances:
        ax.major.formatter._useMathText = True
        plt.draw() # Update the text
        exponent_text = ax.get_offset_text().get_text()
        label = ax.get_label().get_text()
        ax.offsetText.set_visible(False)
        ax.set_label_text(update_label(label, exponent_text))

Which can be used like this



In [35]:

    
fig, ax = plt.subplots()

x = np.linspace(0, 1e8, 1000)
y = 1e10 * np.sin(1e-7*x)

ax.plot(x, y)
ax.set_ylabel("amplitude")
ax.set_xlabel("time [s]")
format_label_string_with_exponent(ax, axis='both')
plt.show()

As you will see, the label is edited and given the exponent, if units exist they are appended to that. While this isn't the most flexible solution, it is quite a standard notation.



In [41]:

    
x = np.linspace(0, 10, 1000)
y = 1e10 * np.sin(x)

def plot(xlabel="time [s]", ylabel="amplitude [Watts]", axis="both"):
    fig, ax = plt.subplots()
    
    ax.plot(x, y)
    ax.set_ylabel(xlabel)
    ax.set_xlabel(ylabel)
    format_label_string_with_exponent(ax, axis=axis)
    plt.show()



In [48]:

    
from IPython.html.widgets import interact

interact(plot, axis=["x", "y", "both"])

Setting nice axes labels in matplotlib

The defaults

Improving on the defaults

Customing the results

An interactive widget to check it works in all the right ways