Making a Plotly Plot

I’m a bit nervous about investing time into an interactive plotting framework after getting burned by Google Motion Charts.¹ But, plotly seems to work even offline, which I think means once I’ve installed it, it doesn’t depend on a service or code hosted by the plotly company. That makes me feel a little more confident. I’d like to build some animations in it, but that means learning how it works, so here I go!

library(tidyverse)
library(plotly)
library(palmerpenguins)

Basic scatter.

Following the book and the docs, it looks like if I were to take the “layers” analogy to building a plot, the most basic layer function is going to be plotly::add_trace(). Data gets mapped to plot aesthetics with function notation.

plot_ly(
  data = penguins,
  x = ~bill_length_mm,
  y = ~bill_depth_mm,
  color = ~species
) |> 
  add_trace(
    type = "scatter",
    mode = "markers",
    size = 4
  )

Some thoughts:

• I think the type argument defines the kind of “space” the plot is placed in? Putting in an unsupported type returns a pretty diverse set of options that’s leaving me a bit confused about the exact work this argument does.

• I think mode is how you go about defining the plotted geometry, with "markers" being points.

• It’s nice how the points default to ColorBrewer Dark2 with a slight transparency for overplotting.

Theming

It looks like the approach to theming is to just set everything by hand in plotly::layout(). This took a little bit of messing around with to find what all of the various parameters are called in plotly. My ggplot2::theme() translations are:

ggplot2	plotly
plot.background	paper_bgcolor
panel.background	plot_bgcolor
panel.grid.major.[x,y]	[xy]axis

Another thing to note is that to get a transparent layer, you need to give it a hex code with 00 transparency at the end, rather than an NA or NULL value.

I have a few of these colors defined in my blog .Rprofile.

plot_bg

[1] "#122F4A"

major

[1] "#4C5D75"

minor

[1] "#31455E"

Also, to get the right font family, you have to reference fonts you’ve imported in the html page, rather than fonts imported into R with showtext.

Here it is in plotly.

plot_ly() |> 
  add_trace(
    data = penguins,
    x = ~bill_length_mm,
    y = ~bill_depth_mm,
    color = ~species,
    type = "scatter",
    mode = "markers"
  ) |> 
  layout(
    plot_bgcolor = "#ffffff00",
    paper_bgcolor = plot_bg,
    font = list(
      family = "Fira Sans",
      color = "#fff"
    ), 
    xaxis = list(
      gridcolor = minor
    ),
    yaxis = list(
      gridcolor = minor
    )    
  )

ggplotly

plotly also has a the ability to convert ggplot2 plots into plotly plots, at least somewhat. Here’s how it does by default.

penguin_plot <- 
  ggplot(
    data = penguins, 
    aes(x = bill_length_mm, 
        y = bill_depth_mm, 
        color = species
      )
  ) +
  geom_point() +
  scale_color_brewer(palette = "Dark2")

ggplotly(penguin_plot)

So, it looks like the panel.background = element_blank() I set in my blog theme doesn’t translate over in the conversion. Which is honestly a good illustration of why its probably worth learning a little bit about how the actual plotly system works, even if you’re going to mostly be interacting with it through plotly::ggplotly() like I am

ggplotly(penguin_plot) |> 
  layout(
    plot_bgcolor = "#ffffff00"
  )

Footnotes

1. Really, it’s the deprecation of Flash, but Google never updated how the motion charts work.