Daniel Liden

Blog / About Me / Photos / LLM Fine Tuning / Notes /

Basic Plotting in Julia

In this short post, I show one of the many ways of using Julia within emacs org mode, and will describe some of the basic plotting functionality in Julia.

Getting Started with Julia in Org Mode: jupyter-julia.

While julia-snail is my favorite Julia development environment, it's support for org-mode is, at present, quite limited. For example, it seems to ignore the :file argument, making it difficult to save figures to specific locations. I find that emacs-jupyter provides the most featureful and reliable way to work with Julia in org-mode. Read the emacs-jupyter documentation for instructions on how to use emacs-jupyter with org-mode.

Making some Plots

Let's make some plots. We'll use the Plots.jl package and explore a few different plotting styles. First, we load the packages we'll be using. I tend to plot statistical distributions fairly often, so I'll load Distributions and Statplots in addition to Plots.

using Plots
using Distributions
using StatsPlots

Building Plots Incrementally

With these packages loaded, we'll start simply and plot a standard normal distribution.

plot(Normal(), fill=(0,0.5,:red))
Figure 1: A basic plot

The Plots.jl package makes it easy to update plots after creation. A Julia convention is that methods ending in ! modify their arguments in place. In this case, we can call plot!() to incrementally add to our plot.

plot!(title="Standard Normal Distribution", xlabel="x", ylabel="p(x)") # Add Labels
plot!(leg=false) # Remove the Legend
Figure 2: Removing the legend and adding labels

Let's make one final set of changes and update some of the font sizes for better readability.

plot!(tickfont=font(18, "courier"),
font(18, "Computer Modern"))
Figure 3: Changing some Fonts

Plotting Real Data

That's enough of that. Usually we're plotting real data, not standard standard distributions. Let's get some. First we'll pull from the RDatasets package, which is an excellent source of go-to data science and statistics examples such as mtcars and iris. We'll use the venerable mtcars to show how to work with data in a basic way.

using RDatasets, DataFrames
mtcars = dataset("datasets", "mtcars")

10 rows × 12 columns (omitted printing of 3 columns)

1Mazda RX421.06160.01103.92.6216.460
2Mazda RX4 Wag21.06160.01103.92.87517.020
3Datsun 71022.84108.0933.852.3218.611
4Hornet 4 Drive21.46258.01103.083.21519.441
5Hornet Sportabout18.78360.01753.153.4417.020
7Duster 36014.38360.02453.213.5715.840
8Merc 240D24.44146.7623.693.1920.01
9Merc 23022.84140.8953.923.1522.91
10Merc 28019.26167.61233.923.4418.31

We can use the @df macro and access columns by prepending them with :.

@df mtcars scatter(:MPG, :HP, group=:Cyl, background=:black, msize=8, keytitle="N Cylinders")
title!("Horsepower by MPG and N Cylinders")
Figure 4: plotting data with Plots.jl

Modifying Plot Visual Components

It's not always straightforward to figure out how to modify visual components of a plot. Most of the relevant information lies in the Attributes section of the Plots.jl documentation (and its subsections).

Different Plot Styles

One of the benefits of the Plots.jl package is its support for many different plotting backends. The default is GR which, according to the documentation, is

The default backend. Very fast with lots of plot types. Still actively developed and improving daily.

and it offers speed; 2D and 3D plots, and standalone or inline plotting.

Here we'll repeat the plot above with the UnicodePlots backend. We first need to install the package with add UnicodePlots in the package manager. Note that you can do this right from the Jupyter repl in emacs-jupyter by pressing ] in the repl.

We then specify that we'd like to use this backend with the unicodeplots() function.

# scatterplot
@df mtcars scatter(:MPG, :HP, title="HP vs MPG", xlabel="Mpg", ylabel="HP")
                        HP vs MPG                   
343.49 |        ⚬                               | y1
       |                                        |   
       |                                        |   
       |         ⚬                              |   
       |     ⚬ ⚬                                |   
       |        ⚬                               |   
       | ⚬                                      |   
    HP |                                        |   
       |        ⚬ ⚬ ⚬ ⚬⚬⚬                       |   
       |        ⚬⚬                              |   
       |                                        |   
       |             ⚬ ⚬  ⚬              ⚬      |   
       |                  ⚬  ⚬    ⚬             |   
       |                            ⚬       ⚬ ⚬ |   
 43.51 |                       ⚬         ⚬      |   
        9.695             Mpg             34.605    

I haven't yet figured out how to get the colors to appear as they are supposed to in org mode, so for now, I'm simply showing a basic scatterplot with no grouping. I'll update if and when I figure that out. It should look like this:

Figure 5: screenshot of unicodeplot from vterm


There's a lot more to get into with plotting in Julia and with using Julia in emacs. This post serves as a small jumping-off point—just enough to get started, with a few pointers to further resources, and some questions to start pursuing. I'll write more on this topic as I learn more!

Date: 2022-07-19 Tue 00:00

Emacs 29.3 (Org mode 9.6.15)