2  Pictogram

I recently discovered the waffle package for making waffle plots. Emily and I refuse to teach students on our Applied Data Skills course how to make pie charts, and these are a sweet alternative.

Setup

#devtools::install_github("hrbrmstr/waffle")
library(tidyverse)  # for data wrangling
library(waffle)     # for pictograms
library(showtext)   # for fonts

2.1 Fonts

I’m never sure what I’m doing with fonts, but this procedure seems to work to install the FontAwesome fonts that you need for pictograms and also any other specialist fonts you’d like.

Code

waffle::install_fa_fonts()
fontdir <- "/Library/Frameworks/R.framework/Versions/4.1/Resources/library/waffle/fonts/"
font_add(family = waffle::fa5_brand,
         regular = paste0(fontdir, "fa-brands-400.ttf"))
font_add(family = waffle::fa5_solid,
         regular = paste0(fontdir, "fa-solid-900.ttf"))

# https://www.dafont.com/market-deco.font
font_add(family = "Market Deco",
         regular = "fonts/Market_Deco.ttf")
font_add_google("Sen")
showtext_auto()

2.2 Pictogram data

Pictograms are generally used for counts, and pets are my go-to example for unordered categorical data (even though ferrets are obviously the best).

I found a table of pet ownership in the USA. They had an excel file download option for the data, but the file was corrupted, so I used datapasta to copy it below.

Code

# Number of U.S. Households That Own a Pet, by Type of Animal 
us_pets <- tibble::tribble(
               ~Pet, ~Number,
              "Dog",      69,
              "Cat",    45.3,
  "Freshwater fish",    11.8,
             "Bird",     9.9,
     "Small animal",     6.2,
          "Reptile",     5.7,
            "Horse",     3.5,
   "Saltwater fish",     2.9
  ) %>%
  mutate(Pet = factor(Pet, Pet)) # keep this order

2.3 Icons

Then I need to find matching free icons at fontawesome.

Make sure you’re searching the v5 icons; v6 icons won’t work and will result in an impenetrable error-message from scale_label_pictogram().

Code

icons <- c("dog", "cat", "fish-fins", "crow", "frog", "dragon", "horse", "fish")
ggplot(us_pets, aes(label = Pet, values = Number)) +
  geom_pictogram() +
  scale_label_pictogram(values = icons)

## Error in vapply(if (n > length(values)) rep(values[[1]], n) else values, : values must be length 1,
##  but FUN(X[[8]]) result is length 0

Yes, I know I used fish twice, and not many people owned winged dragons, but they were the only reptile!

Code

icons <- c("dog", "cat", "fish", "crow", "frog", "dragon", "horse", "fish")
ggplot(us_pets, aes(label = Pet, values = Number)) +
  geom_pictogram() +
  scale_label_pictogram(values = icons)

2.4 Colour

Now, let’s add some colour. I’m also going to make the icon size a little smaller.

Code

ggplot(us_pets, aes(label = Pet, values = Number, color = Pet)) +
  geom_pictogram(size = 8) +
  scale_label_pictogram(values = icons)

I like the ggplot default colours, except for the green fish. I’d rather it be yellow. I found this function to re-create the ggplot default colour scheme.

Code

ggplotColours <- function(n, h = c(0, 360) + 15){
  h[2] <- h[2] - 360/n
  hcl(h = (seq(h[1], h[2], length = n)), c = 100, l = 65)
}
pet_colours <- ggplotColours(8)
pet_colours[3] <- "#F5C748"

Code

ggplot(us_pets, aes(label = Pet, values = Number, color = Pet)) +
  geom_pictogram(size = 8) +
  scale_label_pictogram(values = icons) +
  scale_colour_manual(values = pet_colours)

2.5 Clean it up

Add theme_enhance_waffle() to get rid of extraneous background, and add labels to aid interpretation and for the data provenance.

Code

ggplot(us_pets, aes(label = Pet, values = Number, color = Pet)) +
  geom_pictogram(size = 8) +
  scale_label_pictogram(values = icons) +
  scale_colour_manual(values = pet_colours) +
  theme_enhance_waffle() +
  labs(title = "Pet Ownership in America",
       subtitle = "Each icon represents 1 million houeholds",
       caption = "Data from the American Pet Products Association (APPA) 2021-2022 National Pet Owners Survey")

2.6 Customise

Finally, use theme() to really customise the look.

Code

ggplot(us_pets, aes(label = Pet, values = Number, color = Pet)) +
  geom_pictogram(size = 10) +
  scale_label_pictogram(values = icons) +
  scale_colour_manual(values = pet_colours) +
  theme_enhance_waffle() +
  labs(title = "Pet Ownership in America",
       subtitle = "Each icon represents 1 million houeholds",
       caption = "Data from the American Pet Products Association (APPA) 2021-2022 National Pet Owners Survey") +
  theme(axis.ticks = element_blank(),
        plot.background = element_rect(fill = "black", color = "black"),
        plot.margin = unit(c(.5, .5, .5, .0), 'cm'),
        plot.title = element_text(family = "Market Deco", hjust = 0.5, size = 40),
        plot.subtitle = element_text(size = 25, hjust = 0.5),
        plot.caption.position = "plot",
        panel.background = element_rect(fill = "black", color = "transparent"),
        legend.position = "bottom",
        legend.background = element_blank(),
        legend.title = element_blank(),
        legend.key = element_blank(),
        legend.text = element_text(size = 15, vjust = 1),
        legend.key.width = unit(0.55, "in"),
        legend.box.margin = margin(-.3,0,0,0, "inches"),
        text = element_text(family = "Sen", color = "white"))

Number of US households (in millions) with each of 8 common pet types.

2.7 Save Pictogram

However, I had trouble saving the plot using ggsave(). The icons ended up all question marks. I eventually fixed that by doing something with the fonts (maybe adding them with showtext?), but the icons were still tiny, and the font was blurry at a dpi that makes the icons look right.

This turned out to be because ggsave() uses ragg by default (HT Steve Haroz), which isn’t playing well with fontawesome icons, so I had to set the device back to png. However, this wasn’t a problem when I updated my R and RStudio and started using Quarto (not sure which fixed it).

Code

ggsave("images/day2.png", width = 8, height = 8, device = png)

knitr::include_graphics("images/day2.png")