Basics of Data Visualization

Vizagrams is more than a diagram drawing package, it is also a tool for data visualization. The main idea is to integrate diagramming with data visualization, hence increasing the kind of visualizations we can express. This integration is made possible mainly due to the way graphical marks are coded.

The way Vizagrams defines graphical marks allows us to create marks such as axes, grids, legends, and so on, which help us assemble a plot. All these elements are used by a more general mark called Plot. The parameters used to create an instance of Plot is what we call a graphic specification.

Vizagrams is a visualization grammar in the sense that it provides a graphic specification. Moreover, the specification follows the specification style of Vega-Lite, with some slight modifications. If you know Vega-Lite, the way Vizagrams defines plots will be very familiar to you. Yet, for those that are not used to visualization grammars, Vizagrams also provides some quick functions to produce plots.

using Vizagrams
using DataFrames
using Statistics
using Random

1. First Data Visualization

We start with a very simple example. We generate some random values, and then create a plot using the function, you guessed it, plot. This function is just a quick way to specify a plot without having to write the whole specification.

Random.seed!(4)
plt = plot(x=rand(10),y=rand(10))
draw(plt,height=300)

We have a plot. More interestingly, the plt object is a diagram, thus, we can manipulate and compose with other marks.

d1 = plt↑S(:fontWeight => :bold)TextMark(; text="This is my plot", anchor=:e, fontsize=14)
d2 =
    R(π / 5) * (
        plt↑(
            S(:fontWeight => :bold) *
            TextMark(; text="This is my plot rotated", anchor=:e, fontsize=14)
        )
    )

d = d1 → (T(50, 0), d2)
draw(d)

2. Graphic Specification

For more control over a data visualization, one can use a graphic specification. A specification is a description of the plot, which contains information such as how to map the data to the visual variables, which type of graphical marks one must use, how to scale the data, and so on. As we have stated, Vizagrams follows the specification style from Vega-Lite.

First, let us import a dataframe to be visualized.

using VegaDatasets
df = DataFrame(dataset("cars"));

# We need to drop missing values, as Vizagrams does not handle them yet.
df = dropmissing(df);

plt = Plot(
    data=df,
    encodings=(
        x=(field=:Horsepower,),
        y=(field=:Miles_per_Gallon,),
        color=(field=:Origin,)
    ),
    graphic=Circle(r=5)
)
draw(plt)

Note that in the example above, Vizagrams tries to infer the proper scaling and axes for the given plot based on the underlying data. For example, Vizgrams is guessing that both :Horsepower and :Miles_per_Gallon are quantitative data, while :Origin is nominal.

Users can provide more parameters to the graphic specifcation in order to get a graphic closer to ones preferences. For example, besides data, encoding and graphic, one can use parameters such as title, figsize and config.

Below we increment the previous plot with more parameters:

p = Plot(
    title="Plot",
    figsize=(500,300),
    config=(
        xgrid=(flag=true,),
        ygrid=(flag=true,)
        ),
    data=df,
    encodings=(
        x=(field=:Horsepower,datatype=:q),
        y=(field=:Miles_per_Gallon,),
        color=(field=:Acceleration,scale_range=(:red,:blue)),
        size = (field = :Cylinders,datatype=:o, scale_range=(1,3)),
    ),
    graphic=S(:stroke=>:black)Circle(r=3)
)

draw(p, height=300)

Let us explore a bit of our specification:

p = Plot(
    title="Plot",
    figsize=(300,300),
    config=(
        xgrid=(flag=true,),
        ygrid=(flag=true,)
        ),
    data=df,
    encodings=(
        x=(field=:Horsepower,datatype=:q),
        y=(field=:Miles_per_Gallon,),
        color=(field=:Acceleration,scale_range=(:red,:blue)),
        size = (field = :Cylinders,datatype=:o, scale_range=(1,3)),
    ),
    graphic=S(:stroke=>:black)Circle(r=3)
)

The main fields we have here are title, figsize, config, data, encodings, and graphic.

title is title above the plot;
figsize is pixel size for frame containing the graphic;
config is

for extra attributes for manipulating aesthetic aspects of the plot, such as grids, axes style, and so on'

data is picking the dataset;
encodings is mapping the visual variables (positions x y, and

color) to the data columns (x goes to:Horsepower,ygoes to:MilesperGallonandcolorgoes to:Field`);

graphic is defininig how to draw the graphic inside the plotting context.

All these parameters are somewhat the same as in Vega-Lite, with the exception of graphic.

In Vega-Lite and other visualization grammars (e.g. ggplot), instead of this field graphic they have a field mark or geom (stands for geometry). Their role is very similar to what we have called graphic in the sense that they describe the sort of "graphical mark" to be drawn in the plotting frame. Yet, as will be shown further, graphic contains what is called graphic expression, which are functions that map a set of data to a diagram.

3. More Examples

We end this tutorial presenting some more examples. We do not dive into the extra capabilities of the graphic parameter in this tutorial.

gdf = combine(groupby(df,[:Year,:Origin]),:Horsepower=>mean,:Miles_per_Gallon=>mean)

plt = Plot(
    config=(xaxis=(ticktextangle=π/2,),),
    data=gdf,
    encodings=(
        x=(field=:Year,),
        y=(field=:Horsepower_mean,),
        color=(field=:Origin,),
    ),
    graphic= S(:strokeWidth=>4)Line()
)
draw(plt)

If we want to draw the points over the line, we can just add the circles.

plt = Plot(
    config=(xaxis=(ticktextangle=π/2,),),
    data=gdf,
    encodings=(
        x=(field=:Year,),
        y=(field=:Horsepower_mean,),
        color=(field=:Origin,),
    ),
    graphic= S(:strokeWidth=>4)Line() + S(:opacity=>1)Square(l=10)
)
draw(plt)

Vizagrams also has default behavior for marks such as bars.

gdf = combine(groupby(df,[:Year,:Origin]),:Horsepower=>mean,:Miles_per_Gallon=>mean)

plt = Plot(
    config=(xaxis=(ticktextangle=π/2,),),
    data=gdf,
    encodings=(
        x=(field=:Year,),
        y=(field=:Horsepower_mean,),
    ),
    graphic= S(:fill=>:steelblue)Bar(w=20)
)
draw(plt)

gdf = combine(groupby(df,[:Year,:Origin]),:Horsepower=>mean,:Miles_per_Gallon=>mean)

plt = Plot(
    config=(xaxis=(ticktextangle=π/2,),),
    data=gdf,
    encodings=(
        x=(field=:Year,),
        y=(field=:Horsepower_mean,scale_domain=(0,400)),
        color=(field=:Origin,),
    ),
    graphic= S(:stroke=>:black)Bar(w=20)
)
draw(plt)

disasters = DataFrame(dataset("disasters"))
disasters = dropmissing(disasters);
disasters = filter(Cols(:Entity)=>x->x!="All natural disasters", disasters);

plt = Plot(
    title="Disasters",
    config = (
        grid=NilD(),
        ),
    figsize=(500,400),
    data=disasters,
    encodings=(
        x = (field = :Year, datatype = :q, scale_domain=(1900,2017)),
        y = (field = :Entity,datatype=:n, scale_domain=unique(sort(disasters.Entity, rev=true))),
        color = (field = :Entity,legend=(flag=false,)),
        size = (field = :Deaths,datatype=:q, scale_domain =(0,maximum(disasters.Deaths)), scale_range=(2,30))
    ),
    graphic = S(:stroke=>:black,:opacity=>0.8)Circle()
)

draw(plt, height=500)