Skip to main content
  1. Posts/

Working with Fonts in Matplotlib

··744 words·4 mins·
Table of Contents

This post continues my previous post on various tips and tricks collected during the process of using Matplotlib.

Working with math font

Default math font is ugly?

If you try to play with the built-in LaTeX-style math support in the recent version of Matplotlib, e.g., version 2.0. You may have noticed that the default math font is ugly compared to the nicer look of math symbols produced by LaTeX. See the figure above for an example.

It is because the default math font has changed since Matplotlib version 2.0. If you want to use the LaTeX-style font. You need to change the default math font to computer modern. There are two ways to achieve to this.

Change the math font temporarily

You can change math font in your plotting script so that it only takes effect in your script. To do this, add the following statement to your script,

import matplotlib as mpl
mpl.rcParams['mathtext.fontset'] = 'cm'

Change the math font globally

You can also change the math font options in your matplotlibrc files. This will affect all your scripts afterwards. This file is usually located in $HOME under the directory ~/.config/matplotlib. But its location may vary. Use the following command to find where rc file is on your system,

import matplotlib as mpl

Then find the line

#mathtext.fontset : dejavusans

Uncomment it and change it to

mathtext.fontset : cm

Now, the math font will look much better. All your script will use the new settings.

Mix math text with regular text

If we want to use line break in math expression, it will be useless since everything inside raw text is interpretted verbatim. In order to combine math expression with regular text, we can concatenate them together.

import matplotlib.pyplot as plt

fig, ax = plt.subplots()

ax.text(0.5, 0.5, r"$\mu=1$" + "\n" + r"$\sigma=2$",

The produced image is

Type 3 font used by Matplotlib PDF backend

If you use the default Matplotlib setting to produce a PDF file and then include it in your paper, chances are that you will encounter errors when you test the PDF for compliance with the IEEE conference standard. The error shows that Type 3 font has been used in your paper. That is because Matplotlib use Type 3 font to produce PDF file by default.

The disadvantage of Type 3 font is that it can not be edited by a PDF editor such as Adobe Acrobat Pro. The “advantage” is that the produced PDF is small in size.

You can change this behavior in your script to use Type 42 font instead. This setting will work locally:

import matplotlib as mpl
mpl.rcParams['pdf.fonttype'] = 42

If you want to change this settings globally1, you can edit your matplotlibrc file like the example of changing math font.

Find the line,

#pdf.fonttype       : 3       # Output Type 3 (Type3) or Type 42 (TrueType)

Uncomment it and change it to

pdf.fonttype       : 42        # Output Type 3 (Type3) or Type 42 (TrueType)

Reducing the PDF size when Type-42 fonts are used

I have come to find that if you use Chinese font in your image and use the Type 42 font, the produced PDF size will be much larger. If you do not want to edit the produce pdf file, I think using Type 3 font may be fine.

Some publisher explicitly require that Type 3 font is not included. Then we need to reduce the size of PDF in which Type 42 fonts are used. According to solutions here, we can convert the PDF file to Ghostscript file and convert back to PDF.

pdf2ps file.pdf
ps2pdf -dPDFSETTINGS=/prepress file-optimized.pdf

The option -dPDFSETTINGS defines the quality of the produced PDF. Possible options and explanations are listed below:

-dPDFSETTINGS=/screen(screen-view-only quality, 72 dpi images)
-dPDFSETTINGS=/ebook(low quality, 150 dpi images)
-dPDFSETTINGS=/printer(high quality, 300 dpi images)
-dPDFSETTINGS=/prepress(high quality, color preserving, 300 dpi imgs)
-dPDFSETTINGS=/default(almost identical to /screen)

The conversion process will reduce the PDF size considerably. In my case, the PDF size went from 5.3M to just 29k. The quality of the optimized PDF is acceptable. So this is an viable solution.

P.S.: The pdffonts shipped by most Linux system can be used to detect which fonts are embedded in your PDF files. Here is another example to use pdffonts command.


  1. It means that you do not need to set the fonttype in every source file. ↩︎


How to Plot Unicode Characters with Matplotlib
··461 words·3 mins
A Guide on Using Unicode Characters in Matplotlib
··664 words·4 mins
解决 Matplotlib 使用中文乱码问题
·168 words·1 min