Figure¶

The whole figure. The figure keeps
track of all the child Axes, a smattering of
‘special’ artists (titles, figure legends, etc), and the canvas.
(Don’t worry too much about the canvas, it is crucial as it is the
object that actually does the drawing to get you your plot, but as the
user it is more-or-less invisible to you). A figure can have any
number of Axes, but to be useful should have
at least one.

The easiest way to create a new figure is with pyplot:

fig = plt.figure()  # an empty figure with no axes
fig.suptitle('No axes on this figure')  # Add a title so we know which it is

fig, ax_lst = plt.subplots(2, 2)  # a figure with a 2x2 grid of Axes

Axes¶

This is what you think of as ‘a plot’, it is the region of the image
with the data space. A given figure
can contain many Axes, but a given Axes
object can only be in one Figure. The
Axes contains two (or three in the case of 3D)
Axis objects (be aware of the difference
between Axes and Axis) which take care of the data limits (the
data limits can also be controlled via set via the
set_xlim() and
set_ylim() Axes methods). Each
Axes has a title (set via
set_title()), an x-label (set via
set_xlabel()), and a y-label set via
set_ylabel()).

The Axes class and its member functions are the primary entry
point to working with the OO interface.

Axis¶

These are the number-line-like objects. They take
care of setting the graph limits and generating the ticks (the marks
on the axis) and ticklabels (strings labeling the ticks). The
location of the ticks is determined by a
Locator object and the ticklabel strings
are formatted by a Formatter. The
combination of the correct Locator and Formatter gives
very fine control over the tick locations and labels.

Artist¶

Basically everything you can see on the figure is an artist (even the
Figure, Axes, and Axis objects). This
includes Text objects, Line2D objects,
collection objects, Patch objects … (you get the
idea). When the figure is rendered, all of the artists are drawn to
the canvas. Most Artists are tied to an Axes; such an Artist
cannot be shared by multiple Axes, or moved from one to another.

What is a backend?¶

A lot of documentation on the website and in the mailing lists refers
to the “backend” and many new users are confused by this term.
matplotlib targets many different use cases and output formats. Some
people use matplotlib interactively from the python shell and have
plotting windows pop up when they type commands. Some people run
Jupyter notebooks and draw inline plots for
quick data analysis. Others embed matplotlib into graphical user
interfaces like wxpython or pygtk to build rich applications. Some
people use matplotlib in batch scripts to generate postscript images
from numerical simulations, and still others run web application
servers to dynamically serve up graphs.

To support all of these use cases, matplotlib can target different
outputs, and each of these capabilities is called a backend; the
“frontend” is the user facing code, i.e., the plotting code, whereas the
“backend” does all the hard work behind-the-scenes to make the figure.
There are two types of backends: user interface backends (for use in
pygtk, wxpython, tkinter, qt4, or macosx; also referred to as
“interactive backends”) and hardcopy backends to make image files
(PNG, SVG, PDF, PS; also referred to as “non-interactive backends”).

There are four ways to configure your backend. If they conflict each other,
the method mentioned last in the following list will be used, e.g. calling
use() will override the setting in your matplotlibrc.

1. The backend parameter in your matplotlibrc file (see
   Customizing Matplotlib with style sheets and rcParams):

   backend : WXAgg   # use wxpython with antigrain (agg) rendering
   

2. Setting the MPLBACKEND environment variable, either for your
   current shell or for a single script. On Unix:

   > export MPLBACKEND=module://my_backend
   > python simple_plot.py
   
   > MPLBACKEND="module://my_backend" python simple_plot.py
   

   On Windows, only the former is possible:

   > set MPLBACKEND=module://my_backend
   > python simple_plot.py
   

   Setting this environment variable will override the backend parameter
   in any matplotlibrc, even if there is a matplotlibrc in your
   current working directory. Therefore setting MPLBACKEND
   globally, e.g. in your .bashrc or .profile, is discouraged as it
   might lead to counter-intuitive behavior.

3. If your script depends on a specific backend you can use the
   use() function:

   import matplotlib
   matplotlib.use('PS')   # generate postscript output by default
   

   If you use the use() function, this must be done before
   importing matplotlib.pyplot. Calling use() after
   pyplot has been imported will have no effect. Using
   use() will require changes in your code if users want to
   use a different backend. Therefore, you should avoid explicitly calling
   use() unless absolutely necessary.

With a typical installation of matplotlib, such as from a
binary installer or a linux distribution package, a good default
backend will already be set, allowing both interactive work and
plotting from scripts, with output to the screen and/or to
a file, so at least initially you will not need to use any of the
methods given above.

If, however, you want to write graphical user interfaces, or a web
application server (Matplotlib in a web application server), or need a better
understanding of what is going on, read on. To make things a little
more customizable for graphical user interfaces, matplotlib separates
the concept of the renderer (the thing that actually does the drawing)
from the canvas (the place where the drawing goes). The canonical
renderer for user interfaces is Agg which uses the Anti-Grain
Geometry C++ library to make a raster (pixel) image of the figure.
All of the user interfaces except macosx can be used with
agg rendering, e.g., WXAgg, GTK3Agg, QT4Agg, QT5Agg,
TkAgg. In addition, some of the user interfaces support other rendering
engines. For example, with GTK+ 3, you can also select Cairo rendering
(backend GTK3Cairo).

For the rendering engines, one can also distinguish between vector or raster renderers. Vector
graphics languages issue drawing commands like “draw a line from this
point to this point” and hence are scale free, and raster backends
generate a pixel representation of the line whose accuracy depends on a
DPI setting.

Here is a summary of the matplotlib renderers (there is an eponymous
backend for each; these are non-interactive backends, capable of
writing to a file):

And here are the user interfaces and renderer combinations supported;
these are interactive backends, capable of displaying to the screen
and of using appropriate renderers from the table above to write to
a file:

ipympl¶

The Jupyter widget ecosystem is moving too fast to support directly in
Matplotlib. To install ipympl

pip install ipympl
jupyter nbextension enable --py --sys-prefix ipympl

or

conda install ipympl -c conda-forge

See jupyter-matplotlib
for more details.

GTK and Cairo¶

GTK3 backends (both GTK3Agg and GTK3Cairo) depend on Cairo
(pycairo>=1.11.0 or cairocffi).

How do I select PyQt4 or PySide?¶

The QT_API environment variable can be set to either pyqt or pyside
to use PyQt4 or PySide, respectively.

Since the default value for the bindings to be used is PyQt4,
matplotlib first tries to import it, if the import fails, it tries to
import PySide.

Interactive example¶

From an ordinary python prompt, or after invoking ipython with no options,
try this:

import matplotlib.pyplot as plt
plt.ion()
plt.plot([1.6, 2.7])

Assuming you are running version 1.0.1 or higher, and you have
an interactive backend installed and selected by default, you should
see a plot, and your terminal prompt should also be active; you
can type additional commands such as:

plt.title("interactive test")
plt.xlabel("index")

and you will see the plot being updated after each line. Since version 1.5,
modifying the plot by other means should also automatically
update the display on most backends. Get a reference to the Axes instance,
and call a method of that instance:

ax = plt.gca()
ax.plot([3.1, 2.2])

If you are using certain backends (like macosx), or an older version
of matplotlib, you may not see the new line added to the plot immediately.
In this case, you need to explicitly call draw()
in order to update the plot:

plt.draw()

Non-interactive example¶

Start a fresh session as in the previous example, but now
turn interactive mode off:

import matplotlib.pyplot as plt
plt.ioff()
plt.plot([1.6, 2.7])

Nothing happened–or at least nothing has shown up on the
screen (unless you are using macosx backend, which is
anomalous). To make the plot appear, you need to do this:

plt.show()

Now you see the plot, but your terminal command line is
unresponsive; the show() command blocks the input
of additional commands until you manually kill the plot
window.

What good is this–being forced to use a blocking function?
Suppose you need a script that plots the contents of a file
to the screen. You want to look at that plot, and then end
the script. Without some blocking command such as show(), the
script would flash up the plot and then end immediately,
leaving nothing on the screen.

In addition, non-interactive mode delays all drawing until
show() is called; this is more efficient than redrawing
the plot each time a line in the script adds a new feature.

Prior to version 1.0, show() generally could not be called
more than once in a single script (although sometimes one
could get away with it); for version 1.0.1 and above, this
restriction is lifted, so one can write a script like this:

import numpy as np
import matplotlib.pyplot as plt

plt.ioff()
for i in range(3):
    plt.plot(np.random.rand(10))
    plt.show()

which makes three plots, one at a time. I.e. the second plot will show up,
once the first plot is closed.

Summary¶

In interactive mode, pyplot functions automatically draw
to the screen.

When plotting interactively, if using
object method calls in addition to pyplot functions, then
call draw() whenever you want to
refresh the plot.

Use non-interactive mode in scripts in which you want to
generate one or more figures and display them before ending
or generating a new set of figures. In that case, use
show() to display the figure(s) and
to block execution until you have manually destroyed them.

Line segment simplification¶

For plots that have line segments (e.g. typical line plots,
outlines of polygons, etc.), rendering performance can be
controlled by the path.simplify and
path.simplify_threshold parameters in your
matplotlibrc file (see
Customizing Matplotlib with style sheets and rcParams for
more information about the matplotlibrc file).
The path.simplify parameter is a boolean indicating whether
or not line segments are simplified at all. The
path.simplify_threshold parameter controls how much line
segments are simplified; higher thresholds result in quicker
rendering.

The following script will first display the data without any
simplification, and then display the same data with simplification.
Try interacting with both of them:

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

# Setup, and create the data to plot
y = np.random.rand(100000)
y[50000:] *= 2
y[np.logspace(1, np.log10(50000), 400).astype(int)] = -1
mpl.rcParams['path.simplify'] = True

mpl.rcParams['path.simplify_threshold'] = 0.0
plt.plot(y)
plt.show()

mpl.rcParams['path.simplify_threshold'] = 1.0
plt.plot(y)
plt.show()

Matplotlib currently defaults to a conservative simplification
threshold of 1/9. If you want to change your default settings
to use a different value, you can change your matplotlibrc
file. Alternatively, you could create a new style for
interactive plotting (with maximal simplification) and another
style for publication quality plotting (with minimal
simplification) and activate them as necessary. See
Customizing Matplotlib with style sheets and rcParams for
instructions on how to perform these actions.

The simplification works by iteratively merging line segments
into a single vector until the next line segment’s perpendicular
distance to the vector (measured in display-coordinate space)
is greater than the path.simplify_threshold parameter.

Marker simplification¶

Markers can also be simplified, albeit less robustly than
line segments. Marker simplification is only available
to Line2D objects (through the
markevery property). Wherever
Line2D construction parameters
are passed through, such as
matplotlib.pyplot.plot() and
matplotlib.axes.Axes.plot(), the markevery
parameter can be used:

plt.plot(x, y, markevery=10)

The markevery argument allows for naive subsampling, or an
attempt at evenly spaced (along the x axis) sampling. See the
Markevery Demo
for more information.

Splitting lines into smaller chunks¶

If you are using the Agg backend (see What is a backend?),
then you can make use of the agg.path.chunksize rc parameter.
This allows you to specify a chunk size, and any lines with
greater than that many vertices will be split into multiple
lines, each of which has no more than agg.path.chunksize
many vertices. (Unless agg.path.chunksize is zero, in
which case there is no chunking.) For some kind of data,
chunking the line up into reasonable sizes can greatly
decrease rendering time.

The following script will first display the data without any
chunk size restriction, and then display the same data with
a chunk size of 10,000. The difference can best be seen when
the figures are large, try maximizing the GUI and then
interacting with them:

import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
mpl.rcParams['path.simplify_threshold'] = 1.0

# Setup, and create the data to plot
y = np.random.rand(100000)
y[50000:] *= 2
y[np.logspace(1,np.log10(50000), 400).astype(int)] = -1
mpl.rcParams['path.simplify'] = True

mpl.rcParams['agg.path.chunksize'] = 0
plt.plot(y)
plt.show()

mpl.rcParams['agg.path.chunksize'] = 10000
plt.plot(y)
plt.show()

Legends¶

The default legend behavior for axes attempts to find the location
that covers the fewest data points (loc='best'). This can be a
very expensive computation if there are lots of data points. In
this case, you may want to provide a specific location.

Using the fast style¶

The fast style can be used to automatically set
simplification and chunking parameters to reasonable
settings to speed up plotting large amounts of data.
It can be used simply by running:

import matplotlib.style as mplstyle
mplstyle.use('fast')

It is very light weight, so it plays nicely with other
styles, just make sure the fast style is applied last
so that other styles do not overwrite the settings:

mplstyle.use(['dark_background', 'ggplot', 'fast'])

Keywords: matplotlib code example, codex, python plot, pyplot
Gallery generated by Sphinx-Gallery