Edit on GitHub

Gadfly.jl

Gadfly is a system for plotting and visualization written in Julia. It is based largely on Hadley Wickhams's ggplot2 for R and Leland Wilkinson's book The Grammar of Graphics. It was Daniel C. Jones' brainchild and is now maintained by the community. Please consider citing it if you use it in your work.

Package features

  • Renders publication quality graphics to SVG, PNG, Postscript, and PDF
  • Intuitive and consistent plotting interface
  • Works with Jupyter notebooks via IJulia out of the box
  • Tight integration with DataFrames.jl
  • Interactivity like panning, zooming, toggling powered by Snap.svg
  • Supports a large number of common plot types

Installation

The latest release of Gadfly can be installed from the Julia REPL prompt with

julia> ]add Gadfly

The closing square bracket switches to the package manager interface and the add commands installs Gadfly and any missing dependencies. To return to the Julia REPL hit the delete key.

From there, the simplest of plots can be rendered to your default internet browser with

julia> using Gadfly
julia> plot(y=[1,2,3])

Now that you have it installed, check out the Tutorial for a tour of basic plotting and the various manual pages for more advanced usages.

Compilation

Julia is just-in-time (JIT) compiled, which means that the first time you run a block of code it will be slow.

One strategy for dealing with the tens of seconds it can take to display your first plot is to rarely close your REPL session. Revise.jl is useful in this case as it establishes a mechanism which automatically reloads code after it has been modified, thereby reducing the need to restart.

Alternatively, one can avoid the first-time-to-plot penalty altogether by ahead-of-time (AOT) compiling Gadfly into the Julia system image using PackageCompiler.jl.

For example, after making a directory for creating sysimage

mkdir $HOME/JuliaGadflySysImage
cd $HOME/JuliaGadflySysImage

one can compile Gadfly.jl and create the sysimage as follows:

(@v.1.4) pkg> add PackageCompiler
julia> using PackageCompiler
(@v.1.4) pkg> activate .
(JuliaGadflySysImage) pkg> add Gadfly
julia> create_sysimage(:Gadfly; sysimage_path="GadFlySysimage.so")
julia> exit()

At the end of the resulting copious output will be the command to launch this custom version of julia: something like julia --sysimage $HOME/JuliaGadflySysImage/GadFlySysimage.so. Make it convenient by putting an alias in your .bashrc: alias julia-gadfly="julia --sysimage ...".

Note that multiple packages can be built into a new system image at the same time by adding additional arguments: create_sysimage([:Gadfly, :MyOtherFavoritePackage], ...). Conversely, you don't have to precompile everything you need though, as ]add ... still works.

Now that using Gadfly takes just a split second, there's no reason not to do so automatically in your $HOME/.julia/config/startup.jl file.

A few caveats:

  • Updating to the latest versions of compiled packages requires a recompile. ]uping only works for those that haven't been built into the system image.

  • Plots won't be automatically displayed in your default browser unless you tweak Base.Multimedia.displays to return the GadflyDisplay to the last entry. To do so, add atreplinit(x->pushdisplay(Gadfly.GadflyDisplay())) to your startup.jl, or pushdisplay manually.

  • JULIA_PROJECT is entirely disregarded– you'll have to manually ]activate .... see https://github.com/JuliaLang/PackageCompiler.jl/issues/228.