How to Easily Fix Japanese Character Issue in Matplotlib

Background

Have you ever had this problem in Matplotlib where the Japanese characters are not properly displayed?

I'm talking about this ugly problem:

Above graph is using Japanese characters for its title, x and y axes labels. As the default font used in Python does not support Japanese characters, they are not printed properly.

And for each character where the program fails to print, a nasty warning message is displayed:

This is a programmer's nightmare.

By the way, above program is created with below code.

import numpy as np
import matplotlib.pyplot as plt

rand_array = np.random.randint(10, size=20)

plt.figure(figsize=(10, 6))
plt.plot(rand_array)
plt.title("晴耕雨読")
plt.ylabel("前人未踏")
plt.xlabel("率先垂範")
plt.show()

Let's go ahead and solve this issue.

Issue

Since the deafult font fails to print Japanese characters, we need to change the deafult font to a font that supports Japanese characters.

Let's check what the default font is:

import matplotlib

# Get font family
print(matplotlib.rcParams["font.family"])
# >> ['sans-serif']

# Get fonts within the family
matplotlib.rcParams["font.sans-serif"]
# >> ['DejaVu Sans', 'Bitstream Vera Sans', 'Computer Modern Sans Serif', 'Lucida Grande', 'Verdana', 'Geneva', 'Lucid', 'Arial', 'Helvetica', 'Avant Garde', 'sans-serif']

Unfortunately the current font does not support Japanese characters, however the solution is very simple.

Diagnosis

While matplotlib is using the sans-serif font family as default, there are many more fonts that can be used, among them the ones which support Japanese.

Let's go ahead and list them:

import matplotlib
fonts = set([f.name for f in matplotlib.font_manager.fontManager.ttflist])
len(fonts), fonts
# >> (303, {'.Aqua Kana', '.Keyboard', '.New York', '.SF Arabic', '.SF Compact', '.SF Compact Rounded', ...

Note that Im using a 2021 model MacBook Pro and these are the available fonts. You might get a different set of fonts depending on the amchine you are using.

Now, let's see which fonts actually do support Japanese in Matplotlib:

import matplotlib
# Matplitlib gives warning when a character cannot be plotted with a font
# We need to escalate it to error so that we can catch it
import warnings
warnings.filterwarnings("error")

# Get all fonts in the OS
fonts = set([f.name for f in matplotlib.font_manager.fontManager.ttflist])

# Initialize a list where we'll keep Japanese compatible fonts
jp_compatible_fonts = []

for i, font in enumerate(fonts):
 # Try plotting a text that says: XXフォントが日本語をサポートしているので、大好き！
 # Fonts that don't support Japanese will give error
 try:
  fig  = plt.figure(figsize=(0.5,0.5)) 

  plt.text(0, 0, f"{font}フォントが日本語をサポートしているので、大好き！", fontname=font)
  plt.axis('off')
  plt.draw()
  plt.pause(0.001)
  
  # If successfully plotted, add it to the list
  jp_compatible_fonts.append(font)

 except UserWarning:
  pass
 
 # Close the figure so that incompatible font does not influence the next plot
 plt.close(fig)

jp_compatible_fonts
# >> ['Apple SD Gothic Neo', 'Hiragino Mincho ProN', 'Arial Unicode MS', 'Heiti TC', 'Hiragino Maru Gothic Pro', 'Hiragino Sans GB', 'Hiragino Sans', 'PingFang HK', '.Aqua Kana']

In my laptop, there were 9 fonts which were able to plot Japanese with no problem:

• Apple SD Gothic Neo
• Hiragino Mincho ProN
• Arial Unicode MS
• Heiti TC
• Hiragino Maru Gothic Pro
• Hiragino Sans GB
• Hiragino Sans
• PingFang HK
• .Aqua Kana

Let's go ahead and take a look at how these fonts actually plot Japanese:

import matplotlib
fig = plt.figure()
for i, font in enumerate(jp_compatible_fonts):
 plt.text(
  0,
  i,
  f"{font}フォントが日本語をサポートしているので、大好き！",
  fontname=font,
 )
    
plt.axis("off")
plt.ylim(0, len(jp_compatible_fonts))
plt.show()

Below is the result:

Solution

So far, we have found that although the default font in Matplotlib does not plot Japanese characters, there are 9 other fonts which can plot Japanese with no problem. Now let's go back to our original chart.

Before charting though, let's change the deafult font to one of the 9 fonts that we have identified below.

I personally like Hiragino Sans so I will change the font to it.

import numpy as np
import matplotlib.pyplot as plt

### CHANGE THE FONT FROM DEFAULT TO HIRAGINO SANS
plt.rcParams['font.family'] = "Hiragino Sans"

rand_array = np.random.randint(10, size=20)

plt.figure(figsize=(10, 6))
plt.plot(rand_array)
plt.title("晴耕雨読")
plt.ylabel("前人未踏")
plt.xlabel("率先垂範")
plt.show()

Result is below:

I love how Japanese characters are displayed beautifully, and we don't get a nasty warning anymore.

Conclusion

In this blog post, we have solved the Japanese character display issue in Matplotlib. It is a beautiful solution where just adding 1 more line of code gets rid of the issue.

import matplotlib.pyplot as plt
plt.rcParams['font.family'] = "Hiragino Sans"

The point is to stop using the default font, which does not support Japanese, to a font that actually does. For example, Hiragino Sans, which was readily available on my machine.

Hope you enjoyed this blog post. Let me know if you have any questions or comments below.

Until further notice, happy hacking!