#MonthOfJulia Day 18: Plotting

Julia-Logo-Plotting

There’s a variety of options for plotting in Julia. We’ll focus on those provided by Gadfly, Bokeh and Plotly and.

Gadfly

Gadfly is the flavour of the month for plotting in Julia. It’s based on the Grammar of Graphics, so users of ggplot2 should find it familiar.

gadfly-logo

To start using Gadfly we’ll first need to load the package. To enable generation of PNG, PS, and PDF output we’ll also want the Cairo package.

julia> using Gadfly
julia> using Cairo

You can easily generate plots from data vectors or functions.

julia> plot(x = 1:100, y = cumsum(rand(100) - 0.5), Geom.point, Geom.smooth)
julia> plot(x -> x^3 - 9x, -5, 5)

Gadfly plots are by default rendered onto a new tab in your browser. These plots are mildly interactive: you can zoom and pan across the plot area. You can also save plots directly to files of various formats.

julia> dampedsin = plot([x -> sin(x) / x], 0, 50)
julia> draw(PNG("damped-sin.png", 800px, 400px), dampedsin)

damped-sin

Let’s load up some data from the nlschools dataset in R’s MASS package and look at the relationship between language score test and IQ for pupils broken down according to whether or not they are in a mixed-grade class.

julia> using RDatasets
julia> plot(dataset("MASS", "nlschools"), x="IQ", y="Lang", color="COMB",
            Geom.point, Geom.smooth(method=:lm), Guide.colorkey("Multi-Grade"))

nlschools

Those two examples just scratched the surface. Gadfly can produce histograms, boxplots, ribbon plots, contours and violin plots. There’s detailed documentation with numerous examples on the homepage.

Watch the video below (Daniel Jones at JuliaCon 2014) then read on about Bokeh and Plotly.

Bokeh

Bokeh is a visualisation library for Python. Bokeh, like D3, renders plots as Javascript, which is viewable in a web browser. In addition to the examples on the library homepage, more can be found on the homepage for Julia’s Bokeh package.

The first thing you’ll need to do is install the Bokeh library. If you already have a working Python installation then this is easily done from the command line:

$ pip install bokeh

Next load up the package and generate a simple plot.

julia> using Bokeh
julia> autoopen(true);
julia> x = linspace(0, pi);
julia> y = cos(2 * x);
julia> plot(x, y, title = "Cosine")
Plot("Cosine" with 1 datacolumns)

The plot will be written to a file bokeh_plot.html in the working directory, which will in turn be opened by the browser. Use plotfile() to change the name of the file. The plot is interactive, with functionality to pan and zoom as well as resize the plot window.

bokeh-plot

Plotly

The Plotly package provides a complete interface to plot.ly, an online plotting service with interfaces for Python, R, MATLAB and now Julia. To get an idea of what’s possible with plot.ly, check out their feed. The first step towards making your own awesomeness with be loading the package.

using Plotly

Next you should set up your plot.ly credentials using Plotly.set_credentials_file(). You only need to do this once because the values will be cached.

Data series are stored in Julia dictionaries.

julia> p1 = ["x" => 1:10, "y" => rand(0:20, 10), "type" => "scatter", "mode" => "markers"];
julia> p2 = ["x" => 1:10, "y" => rand(0:20, 10), "type" => "scatter", "mode" => "lines"];
julia> p3 = ["x" => 1:10, "y" => rand(0:20, 10), "type" => "scatter", "mode" => "lines+markers"];
julia> Plotly.plot([p1, p2, p3], ["filename" => "basic-line", "fileopt" => "overwrite"])
Dict{String,Any} with 5 entries:
  "error"    => ""
  "message"  => ""
  "warning"  => ""
  "filename" => "basic-line"
  "url"      => "https://plot.ly/~collierab/17"

You can either open the URL provided in the result dictionary or do it programatically:

julia> Plotly.openurl(ans["url"])

plotly-scatter

By making small jumps through similar hoops it’s possible to create some rather intricate visualisations like the 3D scatter plot below. For details of how that was done, check out my code on github.
plotly-3d-scatter

Obviously plotting and visualisation in Julia are hot topics. Other plotting packages worth checking out are PyPlot, Winston and Gaston. Come back tomorrow when we’ll take a look at using physical units in Julia.




The post #MonthOfJulia Day 18: Plotting appeared first on Exegetic Analytics.