The PyPlot module for Julia

This module provides a Julia interface to the Matplotlib plotting library from Python, and specifically to the matplotlib.pyplot module.

PyPlot uses the Julia PyCall package to call Matplotlib directly from Julia with little or no overhead (arrays are passed without making a copy). (See also PythonPlot.jl for a version of PyPlot.jl using the alternative PythonCall.jl package.)

This package takes advantage of Julia's multimedia I/O API to display plots in any Julia graphical backend, including as inline graphics in IJulia. Alternatively, you can use a Python-based graphical Matplotlib backend to support interactive plot zooming etcetera.

(This PyPlot package replaces an earlier package of the same name by Junfeng Li, which used PyPlot over a ZeroMQ socket with IPython.)

Installation

You will need to have the Python Matplotlib library installed on your machine in order to use PyPlot. You can either do inline plotting with IJulia, which doesn't require a GUI backend, or use the Qt, wx, or GTK+ backends of Matplotlib as described below.

Once Matplotlib is installed, then you can just use Pkg.add("PyPlot") in Julia to install PyPlot and its dependencies.

Automated Matplotlib installation

If you set up PyCall to use the Conda.jl package to install a private (not in the system PATH) Julia Python distribution (via Miniconda), then PyPlot will automatically install Matplotlib as needed.

If you are installing PyCall and PyPlot for the first time, just do ENV["PYTHON"]="" before running Pkg.add("PyPlot"). Otherwise, you can reconfigure PyCall to use Conda via:

ENV["PYTHON"]=""
Pkg.build("PyCall")

The next time you import PyPlot, it will tell Conda to install Matplotlib.

OS X

On MacOS, you should either install XQuartz for MacOS 10.9 or later or install the Anaconda Python distribution in order to get a fully functional PyPlot.

MacOS 10.9 comes with Python and Matplotlib, but this version of Matplotlib defaults to with the Cocoa GUI backend, which is not supported by PyPlot. It also has a Tk backend, which is supported, but the Tk backend does not work unless you install XQuartz.

Alternatively, you can install the Anaconda Python distribution (which also includes ipython and other IJulia dependencies).

Otherwise, you can use the Homebrew package manager:

brew install python gcc freetype pyqt
brew link --force freetype
export PATH="/usr/local/bin:$PATH"
export PYTHONPATH="/usr/local/lib/python2.7:$PYTHONPATH"
pip install numpy scipy matplotlib

(You may want to add the two export commands to your ~/.profile file so that they are automatically executed whenever you start a shell.)

Basic usage

Once Matplotlib and PyPlot are installed, and you are using a graphics-capable Julia environment such as IJulia, you can simply type using PyPlot and begin calling functions in the matplotlib.pyplot API. For example:

using PyPlot
# use x = linspace(0,2*pi,1000) in Julia 0.6
x = range(0; stop=2*pi, length=1000); y = sin.(3 * x + 4 * cos.(2 * x));
plot(x, y, color="red", linewidth=2.0, linestyle="--")
title("A sinusoidally modulated sinusoid")

In general, all of the arguments, including keyword arguments, are exactly the same as in Python. (With minor translations, of course, e.g. Julia uses true and nothing instead of Python's True and None.)

The full matplotlib.pyplot API is far too extensive to describe here; see the matplotlib.pyplot documentation for more information. The Matplotlib version number is returned by PyPlot.version.

Exported functions

Only the currently documented matplotlib.pyplot API is exported. To use other functions in the module, you can also call matplotlib.pyplot.foo(...) as plt.foo(...). For example, plt.plot(x, y) also works. (And the raw PyObject for the matplotlib modules is also accessible as PyPlot.matplotlib.)

Matplotlib is somewhat inconsistent about capitalization: it has contour3D but bar3d, etcetera. PyPlot renames all such functions to use a capital D (e.g. it has hist2D, bar3D, and so on).

You must also explicitly qualify some functions built-in Julia functions. In particular, PyPlot.xcorr, PyPlot.axes, and PyPlot.isinteractive must be used to access matplotlib.pyplot.xcorr etcetera.

If you wish to access all of the PyPlot functions exclusively through plt.somefunction(...), as is conventional in Python, you can do import PyPlot; const plt = PyPlot instead of using PyPlot.

Figure objects

You can get the current figure as a Figure object (a wrapper around matplotlib.pyplot.Figure) by calling gcf().

The Figure type supports Julia's multimedia I/O API, so you can use display(fig) to show a fig::PyFigure and show(io, mime, fig) (or writemime in Julia 0.4) to write it to a given mime type string (e.g. "image/png" or "application/pdf") that is supported by the Matplotlib backend.

Non-interactive plotting

If you use PyPlot from an interactive Julia prompt, such as the Julia command-line prompt or an IJulia notebook, then plots appear immediately after a plotting function (plot etc.) is evaluated.

However, if you use PyPlot from a Julia script that is run non-interactively (e.g. julia myscript.jl), then Matplotlib is executed in non-interactive mode: a plot window is not opened until you run show() (equivalent to plt.show() in the Python examples).

Interactive versus Julia graphics

PyPlot can use any Julia graphics backend capable of displaying PNG, SVG, or PDF images, such as the IJulia environment. To use a different backend, simply call pushdisplay with the desired Display; see the Julia multimedia display API for more detail.

On the other hand, you may wish to use one of the Python Matplotlib backends to open an interactive window for each plot (for interactive zooming, panning, etcetera). You can do this at any time by running:

pygui(true)

to turn on the Python-based GUI (if possible) for subsequent plots, while pygui(false) will return to the Julia backend. Even when a Python GUI is running, you can display the current figure with the Julia backend by running display(gcf()).

If no Julia graphics backend is available when PyPlot is imported, then pygui(true) is the default.

Choosing a Python GUI toolkit

Only the Tk, wxWidgets, GTK+ (version 2 or 3), and Qt (version 4 or 5; via the PyQt5, PyQt4 or PySide), Python GUI backends are supported by PyPlot. (Obviously, you must have installed one of these toolkits for Python first.) By default, PyPlot picks one of these when it starts up (based on what you have installed), but you can force a specific toolkit to be chosen by importing the PyCall module and using its pygui function to set a Python backend before importing PyPlot:

using PyCall
pygui(gui)
using PyPlot

where gui can currently be one of :tk, :gtk3, :gtk, :qt5, :qt4, :qt, or :wx. You can also set a default via the Matplotlib rcParams['backend'] parameter in your matplotlibrc file.

Color maps

The PyPlot module also exports some functions and types based on the matplotlib.colors and matplotlib.cm modules to simplify management of color maps (which are used to assign values to colors in various plot types). In particular:

• ColorMap: a wrapper around the matplotlib.colors.Colormap type. The following constructors are provided:

  • ColorMap{T<:Colorant}(name::String, c::AbstractVector{T}, n=256, gamma=1.0) constructs an n-component colormap by linearly interpolating the colors in the array c of Colorants (from the ColorTypes.jl package). If you want a name to be constructed automatically, call ColorMap(c, n=256, gamma=1.0) instead. Alternatively, instead of passing an array of colors, you can pass a 3- or 4-column matrix of RGB or RGBA components, respectively (similar to ListedColorMap in Matplotlib).

  • Even more general color maps may be defined by passing arrays of (x,y0,y1) tuples for the red, green, blue, and (optionally) alpha components, as defined by the matplotlib.colors.LinearSegmentedColormap constructor, via: ColorMap{T<:Real}(name::String, r::AbstractVector{(T,T,T)}, g::AbstractVector{(T,T,T)}, b::AbstractVector{(T,T,T)}, n=256, gamma=1.0) or ColorMap{T<:Real}(name::String, r::AbstractVector{(T,T,T)}, g::AbstractVector{(T,T,T)}, b::AbstractVector{(T,T,T)}, alpha::AbstractVector{(T,T,T)}, n=256, gamma=1.0)

  • ColorMap(name::String) returns an existing (registered) colormap, equivalent to matplotlib.cm.get_cmap(name).

  • matplotlib.colors.Colormap objects returned by Python functions are automatically converted to the ColorMap type.

• get_cmap(name::String) or get_cmap(name::String, lut::Integer) call the matplotlib.cm.get_cmap function.

• register_cmap(c::ColorMap) or register_cmap(name::String, c::ColorMap) call the matplotlib.cm.register_cmap function.

• get_cmaps() returns a Vector{ColorMap} of the currently registered colormaps.

Note that, given an SVG-supporting display environment like IJulia, ColorMap and Vector{ColorMap} objects are displayed graphically; try get_cmaps()!

3d Plotting

The PyPlot package also imports functions from Matplotlib's mplot3d toolkit. Unlike Matplotlib, however, you can create 3d plots directly without first creating an Axes3d object, simply by calling one of: bar3D, contour3D, contourf3D, plot3D, plot_surface, plot_trisurf, plot_wireframe, or scatter3D (as well as text2D, text3D), exactly like the correspondingly named methods of Axes3d. We also export the Matlab-like synonyms surf for plot_surface (or plot_trisurf for 1d-array arguments) and mesh for plot_wireframe. For example, you can do:

surf(rand(30,40))

to plot a random 30×40 surface mesh.

You can also explicitly create a subplot with 3d axes via, for example, subplot(111, projection="3d"), exactly as in Matplotlib, but you must first call the using3D() function to ensure that mplot3d is loaded (this happens automatically for plot3D etc.). The Axes3D constructor and the art3D module are also exported.

LaTeX plot labels

Matplotlib allows you to use LaTeX equations in plot labels, titles, and so on simply by enclosing the equations in dollar signs ($ ... $) within the string. However, typing LaTeX equations in Julia string literals is awkward because escaping is necessary to prevent Julia from interpreting the dollar signs and backslashes itself; for example, the LaTeX equation $\alpha + \beta$ would be the literal string "\$\\alpha + \\beta\$" in Julia.

To simplify this, PyPlot uses the LaTeXStrings package to provide a new LaTeXString type that be constructed via L"...." without escaping backslashes or dollar signs. For example, one can simply write L"$\alpha + \beta$" for the abovementioned equation, and thus you can do things like:

title(L"Plot of $\Gamma_3(x)$")

If your string contains only equations, you can omit the dollar signs, e.g. L"\alpha + \beta", and they will be added automatically. As an added benefit, a LaTeXString is automatically displayed as a rendered equation in IJulia. See the LaTeXStrings package for more information.

SVG output in IJulia

By default, plots in IJulia are sent to the notebook as PNG images. Optionally, you can tell PyPlot to display plots in the browser as SVG images, which have the advantage of being resolution-independent (so that they display without pixellation at high-resolutions, for example if you convert an IJulia notebook to PDF), by running:

PyPlot.svg(true)

This is not the default because SVG plots in the browser are much slower to display (especially for complex plots) and may display inaccurately in some browsers with buggy SVG support. The PyPlot.svg() method returns whether SVG display is currently enabled.

Note that this is entirely separate from manually exporting plots to SVG or any other format. Regardless of whether PyPlot uses SVG for browser display, you can export a plot to SVG at any time by using the Matplotlib savefig command, e.g. savefig("plot.svg").

Modifying matplotlib.rcParams

You can mutate the rcParams dictionary that Matplotlib uses for global parameters following this example:

rcParams = PyPlot.PyDict(PyPlot.matplotlib."rcParams")
rcParams["font.size"] = 15

(If you instead used PyPlot.matplotlib.rcParams, PyCall would make a copy of the dictionary so that the Python rcParams wouldn't be modified.)

Author

This module was written by Steven G. Johnson.