Using Julia as a R user

Authors: - Caroline Cognot - Florian Teste

0: Data to use for the tutorial

We can use the PalmerPenguins dataset (designed for Julia) OR Use the package RDatasets to use our favorite iris dataset.

using Pkg
# Pkg.add("DataFrames")
# Pkg.add("DataFramesMeta")
# Pkg.add("PalmerPenguins")


using DataFrames
using DataFramesMeta

using PalmerPenguins


penguins = dropmissing(DataFrame(PalmerPenguins.load()))

# Pkg.add("RDatasets")

using RDatasets
iris = dataset("datasets","iris")

Plots

list of available packages

List of packages for plots :

• [Plots]
• [StatsPlots] - enriches the Plots package
• [Makie]
• [TidierPlots]
• [AlgebraOfGraphics] uses Makie
• [GadFly]

Other packages are available, see (https://discourse.julialang.org/t/comparison-of-plotting-packages/99860/2)

We tried Makie,TidierPlots,AlgebraOfGraphics and GadFly.

---

Important note : in the following document, every time we used a function, it was called using PackageName.functionName. This is necessary because we are using packages importing functions of the same name. For example plot is shared between most packages.

Functions with names that are not shared can be called by their name only.

1. TidierPlots

https://tidierorg.github.io/TidierPlots.jl/latest/

The goal here is to be able to do the same sort of graphics as when using R ggplot2, that we all love.

Install the package using ], then add TidierPlots. It takes a long time.

The package supports ggplot(), and :

• some Geoms
• some themes, with the default being the ggplot2 theme
• scale_colour_manual and discrete
• facetting with wrap and grid
• scale, labs,lims

TidierPlots uses AlgebraOfGraphics, which uses Makie.

Example using a Julia dataframe :  you need to run the PalmerPenguins.load() in the console when doing it for the first time and press y when it asks if you want the data downloaded.

Issues during the work : it stopped working. First we looked if the version was good.

• ] status TidierPlots.
• ] update TidierPlots : its not better
• remove it : rm TidierPlots
• install from github :

There are many dependances, we update all packages before. - using Pkg - Pkg.update()

• Pkg.add(“TidierPlots”)

using TidierPlots

g=TidierPlots.ggplot(data = penguins) + geom_bar(@aes(x = species)) + labs(x = “Species”); g

Combining plots

Just like the patchwork package, we can combine plots, with + and | for horizontal and / for vertical.

(g +g)

g/(g+g)

g/g +g

((g + g + g) | g) / (g / g)

Warning ; the aes syntax exactly as ggplot2 does not exist. It has many equivalents

• @aes(x=x,y=y)
• @es(x=x,y=y)
• aes(x=:x,y=:y)
• aes(x=“x”,y=“y”)

g1=ggplot(penguins, @aes(x = bill_length_mm, y = bill_depth_mm, color = species)) + geom_point() g2=ggplot(penguins, @aes(x=:bill_length_mm, y=:bill_depth_mm, color=:species)) + geom_point() g3=ggplot(penguins, aes(x = “bill_length_mm”, y = “bill_depth_mm”, color = “species”)) + geom_point() g4=ggplot(penguins, @es(x = bill_length_mm, y = bill_depth_mm, color = species)) + geom_point()

g1/g2+g3/g4

The different syntaxes gave the same results ! The rest of the code is exactly the same as in ggplot2.

Issues

• Package still not stable. It downgrades others when using add, and the GitHub version has unresolved issues.
• It worked at some point (20/08/2024 afternoon) and then in the same evening stopped.

2: Makie

https://docs.makie.org/v0.21/ https://github.com/MakieOrg/Makie.jl

• Date of today : 22/08/2024
• Last release : 19/08/2024
• Last change to the GitHub repository : 20/08/2024

Backend packages : GLMakie, WGLMakie, CairoMakie, RPRMakie.

2.1 Installation : choose a backend and install it.

• GLMakie (OpenGL based, interactive)
• CairoMakie (Cairo based, static vector graphics)
• WGLMakie (WebGL based, displays plots in the browser)
• RPRMakie (Experimental ray-tracing using RadeonProRender)

Then install it using Julia’s package manager Pkg:

using Pkg
# Pkg.add("CairoMakie")

There’s no need to install Makie.jl separately, it is re-exported by each backend package.

2.2 Basic plotting

lines(x,y) scatter(x,y)

using CairoMakie

Makie looks more like base R plots. We did not have the time to finish this part.

3: Gadfly

https://gadflyjl.org/stable/

Gadfly can be installed using in the REPL julia> ]add Gadfly

Gadfly is based on R ggplot2 and supports DataFrames. As opposed to ggplot2, it does not only support DataFrames, you can also use vectors as arguments.

The GitHub page is there : https://github.com/GiovineItalia/Gadfly.jl/tree/master

• Last release : 2021
• Last change to the GitHub repository : 5 months ago (so, early 2024)

# Pkg.add("Gadfly")

using Gadfly

3.1 Basic plots

The basic plotting function is also called plot. The syntax when using a DataFrame is as follows :

plot(data::AbstractDataFrame, elements::Element…; mapping…)

It is not exactly the same as the ggplot2 syntax, but close.

Every information goes in the same plot() function call.

hstack(plots) stacks the plots side by side, vstack stacks vertically.

3.1.1 Geom(s) :

• The default geometry is Geom.point.
• others : https://gadflyjl.org/stable/gallery/geometries/ (with examples), or https://gadflyjl.org/stable/lib/geometries/ for all possibilities
• You can add as much Geom as you want separated by “,”

p1 = Gadfly.plot(penguins, x=:bill_length_mm, y=:bill_depth_mm,color=:species, Geom.point);
p2 = Gadfly.plot(penguins, x=:bill_length_mm, y=:bill_depth_mm,color=:species, Geom.point,Geom.line);
hstack(p1,p2)

Saving a plot is done using the draw function.

Notes :

• default has no background.

img = SVG(“penguin_plot.svg”, 14cm, 8cm) draw(img, p1)

img = SVG(“penguin_plot.png”, 14cm, 8cm) draw(img, p2)

When using arrays instead of DataFrames, instead, use the data directly. However, the automatic axes do not exist and we have to add them manually using Guide.

3.1.2 Guides:

• For adding informations on the plot.
• https://gadflyjl.org/stable/gallery/guides/ for examples, https://gadflyjl.org/stable/lib/guides/ for an exhaustive list

Example : same graph usign first arrays, then the DataFrame directly.

bill_length=penguins.bill_length_mm
bill_depth=penguins.bill_depth_mm
species = penguins.species
p1 = Gadfly.plot(x=bill_length,y=bill_depth,color=species, Geom.point,
Guide.xlabel("bill length"),Guide.ylabel("bill depth"),
Guide.colorkey(title="Species",labels=["my favorite specie","others","another one"]),
Gadfly.Theme(key_position=:inside),Guide.title("I love penguins"));
p2 = Gadfly.plot(penguins,x=:bill_length_mm,y=:bill_depth_mm,color=:species, Geom.point,
Guide.colorkey(title="Species",labels=["my favorite specie","others","another one"]),
Gadfly.Theme(key_position=:inside),Guide.title("I love penguins more"));
hstack(p1,p2)

3.1.3 Layers

Layers allow several instructions in the same plot() call.

Example of layers using the penguins dataframe :

penguinm = @subset(penguins, :sex .== "male")
penguinf = @subset(penguins, :sex .== "female")

p3 = Gadfly.plot(penguinm, x=:bill_length_mm, y=:bill_depth_mm,color=["male"],shape=:species, Geom.point,
    layer(penguinf,x=:bill_length_mm,y=:bill_depth_mm,color=["female"],shape=:species,Geom.point));
    p3

Color, and other aesthetics, can also be mapped by using arrays with group labels or functional types e.g. [“group label”] or [colorant”red”]. [“Group labels”] are added to the key. [colorant”red”] are not added to the key, and not present in the legend of the plot.

#example from Gadfly tutorial :
y1 = [0.1, 0.26, NaN, 0.5, 0.4, NaN, 0.48, 0.58, 0.83]
Gadfly.plot(x=1:9, y=y1, Geom.line, Geom.point,
        color=["Item 1"], linestyle=[:dash], size=[3pt],
    layer(x=1:10, y=rand(10), Geom.line, Geom.point,
        color=["Item 2"], size=[5pt], shape=[Shape.square]),
    layer(x=1:10, y=rand(10), color=[colorant"hotpink"],
        linestyle=[[8pt, 3pt, 2pt, 3pt]], Geom.line))

3.1.4 Scales

https://gadflyjl.org/stable/tutorial/#Continuous-Scales https://gadflyjl.org/stable/tutorial/#Discrete-Scales

Scale can be supplied with Scale.myscale for continuous and discrete scales.

3.2 Compositing

https://gadflyjl.org/stable/man/compositing/#Compositing

With ggplot2, we also like using faceting and grids to represent different data.

3.2.1 Grids using Geom.subplot_grid

For grids, we have to use “xgroup” and/or “ygroup” and “Geom.subplot_grid(Geom.mygeomIwant)” instead of just Geom.mygeomIwant. This does the work of +facet_grid().

# Pkg.add("Compose")

using Compose
p = Gadfly.plot(penguins, x=:bill_length_mm, y=:bill_depth_mm,xgroup=:species,ygroup=:sex,
 shape=:sex, color=:island, Geom.subplot_grid(Geom.point),alpha=[0.5]);
p

Additional Guides can be placed inside the Geom.subplot_grid :

p4 = Gadfly.plot(penguins, x=:bill_length_mm, y=:bill_depth_mm,xgroup=:species,color=:sex,
 shape=:sex,
  Geom.subplot_grid(Geom.point,Guide.ylabel(orientation=:vertical)),
  alpha=[0.5],Guide.title("I love penguins even more"));
p4

3.2.2 Stacks

stackings used different plots and hstack,vstack or gridstack.

vstack(…) puts all argument on top of each other. hstack(…) puts them side by side. gridstack(…) works with a matrix of plots such as [ p1 p2;p3 p4] combining vstack and pstack also works to define different arrangements. a blank panel can be used using only plot()

vstack(hstack(p1,p2,p3,Gadfly.plot()),hstack(p4))

3.3 Conclusion on this package

The documentation is rich with examples from the RDatasets. There are replacements for common ggplot2 functions. The library lists all available functions, Geom, Guides, Statistics, Coordinates, Scales and Shape available.

4 AlgebraOfGraphics

“Define an algebra of graphics based on a few simple building blocks that can be combined using + and *. Still somewhat experimental, may break often.”

Last release : 21/08/2024 Last GitHub activity : 21/08/2024

This package relies on Makie.

Warning : draw of AlgebraOfGraphics is not compatible with draw of Combine loaded earlier in the notebook.

# import Pkg; Pkg.add("AlgebraOfGraphics")
import AlgebraOfGraphics as AOG
using CairoMakie, PalmerPenguins

penguins = dropmissing(DataFrame(PalmerPenguins.load()))
first(penguins, 6)

AOG.set_aog_theme!()


axis = (width = 225, height = 225)
penguin_frequency = AOG.data(penguins) * AOG.frequency() * AOG.mapping(:species)

AOG.draw(penguin_frequency; axis = axis)

Saving a plot :

fg = AOG.draw(penguin_frequency; axis = axis) save(“figure.png”, fg, px_per_unit = 3) # save high-resolution png

4.1 First basic plotting

AlgebraOfGraphics relies on the * and + operations to combine plotting elements.

The plot object is defined through different blocks :

• data() to declare the data used (DataFrame)
• mapping() to declare the aesthetics (aes in ggplot2 or TidierPlots)
  • if classic plot with x,y : you can use mapping(x=:truc1,truc2) OR mapping (truc1,truc2) but not mapping(x=:truc1,y=:truc2).
• visual()
• analyses The resulting plot is then shown using draw(plt).

penguin_bill = AOG.data(penguins) * AOG.mapping(:bill_length_mm, :bill_depth_mm,color=:species)
AOG.draw(penguin_bill; axis = axis)

To rename x and y axis, add => “name of axis”

penguin_bill = AOG.data(penguins) * AOG.mapping(
    :bill_length_mm  => "bill length (mm)",
    :bill_depth_mm  => "bill depth (mm)",
)
AOG.draw(penguin_bill; axis = axis)

To apply a transformation to the values (mm to cm in this example), add => (t->transformation(t))

penguin_bill = AOG.data(penguins) * AOG.mapping(
    :bill_length_mm => (t -> t / 10) => "bill length (cm)",
    :bill_depth_mm => (t -> t / 10) => "bill depth (cm)",
)
AOG.draw(penguin_bill; axis = axis)

To add information, take your first plot object and add new mappings :

plt = penguin_bill * AOG.mapping(color = :species)
AOG.draw(plt; axis = axis)

4.2 Combining + and *

• the * operator adds information on a plot
• the + operator adds a new layer

We can use factorisation to combine the two operations. In this “Algebra Of Graphics”, mapping() is the “neutral” element.

The linear regression removes the point plots. We have to add it back on the plot, using +

And factorise this :

plt = penguin_bill * (AOG.linear() + AOG.mapping()) * AOG.mapping(color = :species)
AOG.draw(plt; axis = axis)

4.3 mapping()

Examples of mappings :

• color =: colorkey
• col= :variablename; to use when gridding plots horizontally across variablename
• row=: variablename; for gridding vertically
• layout=:variablename

If no data was set, the entries of mapping have to be vectors.

The structure of the mapping is tied to the plotting function used with it (analysis), and the visual attribute. See documentations for more information.

• Pair operator => is used to rename columns, transform columns, map to a custom scale

4.4 visual()

default is the xy scatterplot. command : visual(plottype; attributes…)

Examples : - BarPlot - Heatmap

penguin_bill = AOG.data(penguins) * AOG.mapping(
    :body_mass_g,:species,color=:sex,layout=:island
    
)*AOG.visual(BarPlot)
AOG.draw(penguin_bill; axis = axis)

4.5 analyses

They do more complicated operation. Examples :

• histogram()
• density()
• frequency()
• linear()
• expectation() : compute expected value of last arg, conditionned on previous.
• smooth
• contours, filled_contours

4.6 One last example

penguinsplot = AOG.data(penguins)*
    AOG.mapping(:bill_length_mm,:bill_depth_mm,
    col=:species,row=:sex,marker=:sex,color=:island)
AOG.draw(penguinsplot)