[Matplotlib] Do not tilt the axis label

Introduction

• Recently, I see many time series graphs, but some axis labels are tilted.
• Maybe you didn't have time to play with the look? Or maybe I couldn't find a workaround.
• I myself used to tilt the axis label or compromise with a horizontal graph.

background

• In "Data Visualization Design # 4", I saw an article on Twitter saying ** Axis label is not tilted **.
• In that article, there was an introduction of research that tilting the axis label increases the time required for character recognition.
• Then, I investigated how to prevent the axis label from tilting.

Sample data

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import matplotlib.ticker as ticker

#Sample data
data = pd.DataFrame({'MAU':np.random.randint(5000, 8500,(10,))},
                     index=pd.date_range('2019/12/25', freq='d', periods=10))

Graph as usual

fig = plt.figure(figsize=(4,3), dpi=144)
ax = fig.add_subplot(111)

ax.bar(data.index, data['MAU'])
ax.set_xticklabels(data.index, rotation=45)

plt.tight_layout()
plt.savefig('test01.svg')
plt.show()

• Certainly, I am tilting my head according to the tilted label.

Was there such a convenience!

#Looks great with Locator and formatter
fig = plt.figure(figsize=(4,3), dpi=144)
ax = fig.add_subplot(111)

locator = mdates.AutoDateLocator()
formatter = mdates.ConciseDateFormatter(locator)

ax.xaxis.set_major_locator(locator)
ax.xaxis.set_major_formatter(formatter)

ax.bar(data.index, data['MAU'])

plt.savefig('test02.svg')
plt.show()

• If the appearance of the axis label in the graph above is OK, add the following 4 lines! Only this!

locator = mdates.AutoDateLocator()
formatter = mdates.ConciseDateFormatter(locator)

ax.xaxis.set_major_locator(locator)
ax.xaxis.set_major_formatter(formatter)

I want to customize it a little more!

• It is enough that the axis label of the graph above contains all the year, month, and day information, but if you want to cross the year, you need to subtract it in your head.
• For example, I want year information at the beginning of the axis label! What do you do?

Create axis labels yourself

fmt = []
yy = ''
for x in data.index:
    years = x.year
    days = x.day
    if yy != x.year:
        fmt.append(str(days)+'\n'+str(years))
    else:
        fmt.append(str(days))
    yy = x.year

fig = plt.figure(figsize=(4,3), dpi=144)
ax = fig.add_subplot(111)

ax.set_xticklabels(fmt)
ax.bar(data.index, data['MAU'])

plt.savefig('test03.png')
plt.show()

• I think it is suitable when there are few single graphs or axis labels.

Try using FixedFormatter with the matplotlib formatter.

fig = plt.figure(figsize=(4,3), dpi=144)
ax = fig.add_subplot(111)

# positions = [0,1,2,3,4,5,6,7,8,9]
labels = ['Dec\n2019','26','27','28','29','30','31','Jan\n2020','2','3']
# ax.xaxis.set_major_locator(ticker.FixedLocator(positions))
ax.xaxis.set_major_formatter(ticker.FixedFormatter(labels))

ax.bar(data.index, data['MAU'])

plt.savefig('test04.png')
plt.show()

• To use, just create an axis label in list format and set it in formatter.
• When Locator is set, the axis label is not displayed, so it is commented out.
• This is the same in that it is created independently, but I think it is suitable for multiple graphs.

Try using FuncFormatter with the matplotlib formatter.

#matplotlib uses the Gregorian calendar
fig = plt.figure(figsize=(4,3), dpi=144)
ax = fig.add_subplot(111)

YY = ''
@ticker.FuncFormatter
def major_fmt(x, pos):
    global YY
    y = mdates.num2date(int(x)).year
    d = mdates.num2date(int(x)).day
    if YY != y:
        t = '{}\n{}'.format(d, y)
    else:
        t = '{}'.format(d)
    YY = y
    return t

ax.xaxis.set_major_formatter(major_fmt)

ax.bar(data.index, data['MAU'])

plt.savefig('test05.png')
plt.show()

• Create axis labels using user-defined functions.
• The function is automatically executed from @ ticker.FuncFormatter.
• Until I wrote this article, matplotlib didn't know that the Gregorian calendar was the norm. : sweat_smile:
• So it needs to be converted with num2date (int (x)).

About locator and Formatter

• How Locator arranges axis labels
• What kind of display format does Formatter display axis labels?
• If you read the official manual below, ** Axis Labeling Power ** will improve to each level.

[Matplotlib] Official Website

• Tick locators
• Tick formatters

Summary

• ** Customize the "axis label" not tilted ** using Formatter of Matplotlib.
• 1. Create an axis label by yourself. 2) Use automatic settings. 3) Use Fixed Formatter. 4) Use FuncFormatter.
• Writing this article made me feel that ** the design power of data visualization ** has improved a little.