Changing Project Defaults

Upshot

I’ve moved a bunch of R defaults that I want for each post from .Rprofile into _defaults.R, and now run source(here::here("_defaults.R")) in each post where I want them. That looks like more work, but it actually makes things run a bit faster with the way Quarto runs R and freezes outputs.

Initial Defaults

Around when I was setting up this blog project, I decided that I wanted some consistent theming for the figures so that they would fit in nicely into the rest of the blog, but I didn’t want to have to include a megablock of code in every post that looked like this:

library(ggplot2)
library(khroma)
library(ggdark)
library(showtext)
library(colorspace)

# get Fira Sans from google
font_add_google(name = "Fira Sans", family = "Fira Sans")
showtext_auto()
body_bg <- "#222222"
plot_bg <- darken("#375a7f", 0.50)

major <- lighten(
  plot_bg,
  amount = 0.25
)

minor <- lighten(
  plot_bg,
  amount = 0.125
)

strip_bg <- lighten(plot_bg, 0.5)

theme_set(dark_theme_gray(base_size = 12) + 
            theme(text = element_text(family = "Fira Sans"),
                  plot.background = element_rect(fill = plot_bg),
                  panel.background = element_blank(),
                  panel.grid.major = element_line(color = major, linewidth = 0.2),
                  panel.grid.minor = element_line(color = minor, linewidth = 0.2),
                  legend.key = element_blank(),
                  strip.background = element_rect(fill = strip_bg),
                  strip.text = element_text(color = body_bg),
                  axis.ticks = element_blank(),
                  legend.background = element_blank()))

options(
  ggplot2.discrete.colour = khroma::scale_color_bright,
  ggplot2.discrete.fill = khroma::scale_fill_bright,
  ggplot2.continuous.colour = khroma::scale_color_batlow,
  ggplot2.continuous.fill = khroma::scale_fill_batlow
)

All that means I can just do some minimal ggplot2 code in each post and it’ll look something like this:

data(penguins, package = "palmerpenguins")

ggplot(
  penguins, 
  aes(
    x = bill_length_mm,
    y = bill_depth_mm,
    color = species
  )
)+
  geom_point()

So, I stuck that big block of code into the .Rprofile for the blog project so that every time I opened the project, R would automatically source it. Nice, right?

The heaviness of .Rprofile

I started realizing this wasn’t optimal every time I re-rendered the blog for a new post. I have my quarto set to “freeze” each post after it’s rendered, meaning it won’t re-run all of the R code in a post unless I make a change to it, instead using the output of the previous time it ran. That’s a time saver, cause even with many very simple posts with code, it just takes a while to run everything.

The issue was, even with freeze: true, Quarto would still source .Rprofile on every post. Which means that big block of code, including the call to showtext::font_add_google() would run for every post when I re-rendered the blog. And that was starting to get tedious!

Moving to _defaults.R

So, I moved all of the customization code from .Rprofile into _defaults.R file. I forget where I saw a _defaults.R first, but I think it was in some repository maintained by Hadley Wickham. The downside is that it’s not as automatic as .Rprofile, in that I need to source it at the start of every post. That would be annoying if I was going to write the path out by hand, but it’s a little easier with here::here().

source(
  here::here("_defaults.R")
)

The major upside, though, is that sourcing code gets frozen along with all of the other code chunks in a post! So when I re-render the whole blog, Quarto won’t re-run all of the code in _defaults.R unless the code has changed in a post. Overall, it feels worth it!.