Note
Click here to download the full example code
Typesetting With XeLaTeX/LuaLaTeX¶
How to typeset text with the pgf
backend in Matplotlib.
Using the pgf
backend, Matplotlib can export figures as pgf drawing
commands that can be processed with pdflatex, xelatex or lualatex. XeLaTeX and
LuaLaTeX have full Unicode support and can use any font that is installed in
the operating system, making use of advanced typographic features of OpenType,
AAT and Graphite. Pgf pictures created by plt.savefig('figure.pgf')
can be embedded as raw commands in LaTeX documents. Figures can also be
directly compiled and saved to PDF with plt.savefig('figure.pdf')
by
switching the backend
matplotlib.use('pgf')
or by explicitly requesting the use of the pgf
backend
plt.savefig('figure.pdf', backend='pgf')
or by registering it for handling pdf output
from matplotlib.backends.backend_pgf import FigureCanvasPgf
matplotlib.backend_bases.register_backend('pdf', FigureCanvasPgf)
The last method allows you to keep using regular interactive backends and to save xelatex, lualatex or pdflatex compiled PDF files from the graphical user interface.
Matplotlib's pgf support requires a recent LaTeX installation that includes
the TikZ/PGF packages (such as TeXLive), preferably with XeLaTeX or LuaLaTeX
installed. If either pdftocairo or ghostscript is present on your system,
figures can optionally be saved to PNG images as well. The executables
for all applications must be located on your PATH
.
rcParams
that control the behavior of the pgf backend:
Parameter Documentation pgf.preamble Lines to be included in the LaTeX preamble pgf.rcfonts Setup fonts from rc params using the fontspec package pgf.texsystem Either "xelatex" (default), "lualatex" or "pdflatex"
Note
TeX defines a set of special characters, such as:
# $ % & ~ _ ^ \ { }
Generally, these characters must be escaped correctly. For convenience, some characters (_, ^, %) are automatically escaped outside of math environments.
Multi-Page PDF Files¶
The pgf backend also supports multipage pdf files using
PdfPages
from matplotlib.backends.backend_pgf import PdfPages
import matplotlib.pyplot as plt
with PdfPages('multipage.pdf', metadata={'author': 'Me'}) as pdf:
fig1, ax1 = plt.subplots()
ax1.plot([1, 5, 3])
pdf.savefig(fig1)
fig2, ax2 = plt.subplots()
ax2.plot([1, 5, 3])
pdf.savefig(fig2)
Font specification¶
The fonts used for obtaining the size of text elements or when compiling
figures to PDF are usually defined in the rcParams
. You can also use the
LaTeX default Computer Modern fonts by clearing the lists for rcParams["font.serif"]
(default: ['DejaVu Serif', 'Bitstream Vera Serif', 'Computer Modern Roman', 'New Century Schoolbook', 'Century Schoolbook L', 'Utopia', 'ITC Bookman', 'Bookman', 'Nimbus Roman No9 L', 'Times New Roman', 'Times', 'Palatino', 'Charter', 'serif']
),
rcParams["font.sans-serif"]
(default: ['DejaVu Sans', 'Bitstream Vera Sans', 'Computer Modern Sans Serif', 'Lucida Grande', 'Verdana', 'Geneva', 'Lucid', 'Arial', 'Helvetica', 'Avant Garde', 'sans-serif']
) or rcParams["font.monospace"]
(default: ['DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', 'monospace']
). Please note that the glyph
coverage of these fonts is very limited. If you want to keep the Computer
Modern font face but require extended Unicode support, consider installing the
Computer Modern Unicode fonts CMU Serif, CMU Sans Serif, etc.
When saving to .pgf
, the font configuration Matplotlib used for the
layout of the figure is included in the header of the text file.
"""
=========
Pgf Fonts
=========
"""
import matplotlib.pyplot as plt
plt.rcParams.update({
"font.family": "serif",
# Use LaTeX default serif font.
"font.serif": [],
# Use specific cursive fonts.
"font.cursive": ["Comic Neue", "Comic Sans MS"],
})
fig, ax = plt.subplots(figsize=(4.5, 2.5))
ax.plot(range(5))
ax.text(0.5, 3., "serif")
ax.text(0.5, 2., "monospace", family="monospace")
ax.text(2.5, 2., "sans-serif", family="DejaVu Sans") # Use specific sans font.
ax.text(2.5, 1., "comic", family="cursive")
ax.set_xlabel("µ is not $\\mu$")
fig.tight_layout(pad=.5)
Custom preamble¶
Full customization is possible by adding your own commands to the preamble.
Use rcParams["pgf.preamble"]
(default: ''
) if you want to configure the math fonts,
using unicode-math
for example, or for loading additional packages. Also,
if you want to do the font configuration yourself instead of using the fonts
specified in the rc parameters, make sure to disable rcParams["pgf.rcfonts"]
(default: True
).
"""
============
Pgf Preamble
============
"""
import matplotlib as mpl
mpl.use("pgf")
import matplotlib.pyplot as plt
plt.rcParams.update({
"font.family": "serif", # use serif/main font for text elements
"text.usetex": True, # use inline math for ticks
"pgf.rcfonts": False, # don't setup fonts from rc parameters
"pgf.preamble": "\n".join([
r"\usepackage{url}", # load additional packages
r"\usepackage{unicode-math}", # unicode math setup
r"\setmainfont{DejaVu Serif}", # serif font via preamble
])
})
fig, ax = plt.subplots(figsize=(4.5, 2.5))
ax.plot(range(5))
ax.set_xlabel("unicode text: я, ψ, €, ü")
ax.set_ylabel(r"\url{https://matplotlib.org}")
ax.legend(["unicode math: $λ=∑_i^∞ μ_i^2$"])
fig.tight_layout(pad=.5)
Choosing the TeX system¶
The TeX system to be used by Matplotlib is chosen by rcParams["pgf.texsystem"]
(default: 'xelatex'
).
Possible values are 'xelatex'
(default), 'lualatex'
and 'pdflatex'
.
Please note that when selecting pdflatex, the fonts and Unicode handling must
be configured in the preamble.
"""
=============
Pgf Texsystem
=============
"""
import matplotlib.pyplot as plt
plt.rcParams.update({
"pgf.texsystem": "pdflatex",
"pgf.preamble": "\n".join([
r"\usepackage[utf8x]{inputenc}",
r"\usepackage[T1]{fontenc}",
r"\usepackage{cmbright}",
]),
})
fig, ax = plt.subplots(figsize=(4.5, 2.5))
ax.plot(range(5))
ax.text(0.5, 3., "serif", family="serif")
ax.text(0.5, 2., "monospace", family="monospace")
ax.text(2.5, 2., "sans-serif", family="sans-serif")
ax.set_xlabel(r"µ is not $\mu$")
fig.tight_layout(pad=.5)
Troubleshooting¶
- Please note that the TeX packages found in some Linux distributions and MiKTeX installations are dramatically outdated. Make sure to update your package catalog and upgrade or install a recent TeX distribution.
- On Windows, the
PATH
environment variable may need to be modified to include the directories containing the latex, dvipng and ghostscript executables. See Environment Variables and Setting environment variables in Windows for details. - A limitation on Windows causes the backend to keep file handles that have been opened by your application open. As a result, it may not be possible to delete the corresponding files until the application closes (see #1324).
- Sometimes the font rendering in figures that are saved to png images is very bad. This happens when the pdftocairo tool is not available and ghostscript is used for the pdf to png conversion.
- Make sure what you are trying to do is possible in a LaTeX document, that your LaTeX syntax is valid and that you are using raw strings if necessary to avoid unintended escape sequences.
rcParams["pgf.preamble"]
(default:''
) provides lots of flexibility, and lots of ways to cause problems. When experiencing problems, try to minimalize or disable the custom preamble.- Configuring an
unicode-math
environment can be a bit tricky. The TeXLive distribution for example provides a set of math fonts which are usually not installed system-wide. XeTeX, unlike LuaLatex, cannot find these fonts by their name, which is why you might have to specify\setmathfont{xits-math.otf}
instead of\setmathfont{XITS Math}
or alternatively make the fonts available to your OS. See this tex.stackexchange.com question for more details. - If the font configuration used by Matplotlib differs from the font setting
in yout LaTeX document, the alignment of text elements in imported figures
may be off. Check the header of your
.pgf
file if you are unsure about the fonts Matplotlib used for the layout. - Vector images and hence
.pgf
files can become bloated if there are a lot of objects in the graph. This can be the case for image processing or very big scatter graphs. In an extreme case this can cause TeX to run out of memory: "TeX capacity exceeded, sorry" You can configure latex to increase the amount of memory available to generate the.pdf
image as discussed on tex.stackexchange.com. Another way would be to "rasterize" parts of the graph causing problems using either therasterized=True
keyword, or.set_rasterized(True)
as per this example. - If you still need help, please see Getting help
Keywords: matplotlib code example, codex, python plot, pyplot Gallery generated by Sphinx-Gallery