Consider this neat little plot a friend of mine made for teaching purposes. It uses dice rolls to visualize how the distribution converges normally.
My attempt to make his code reproducible by being explicit about unicode characters resulted in the following code:
library(ggplot2)
w6_2 <- sample(c(1:6), 2000, replace = T) + sample(c(1:6), 2000, replace = T)
# see stringi::stri_escape_unicode
d1 <- "\u2680"
d2 <- "\u2681"
d3 <- "\u2682"
d4 <- "\u2683"
d5 <- "\u2684"
d6 <- "\u2685"
dice <- c(
"2" = paste0(d1, d1),
"3" = paste0(d1, d2, "\n", d2, d1),
"4" = paste0(d2, d2, "\n", d1, d3, "\n", d3, d1),
"5" = paste0(d1, d4, "\n", d2, d3, "\n", d3, d2, "\n", d4, d1),
"6" = paste0(d1, d5, "\n", d2, d4, "\n", d3, d3, "\n", d4, d2, "\n", d5, d1),
"7" = paste0(d1, d6, "\n", d2, d5, "\n", d3, d4, "\n", d4, d3, "\n", d5, d2, "\n", d6, d1),
"8" = paste0(d2, d6, "\n", d3, d5, "\n", d4, d4, "\n", d5, d3, "\n", d6, d2),
"9" = paste0(d3, d6, "\n", d4, d5, "\n", d5, d4, "\n", d6, d3),
"10" = paste0(d4, d6, "\n", d5, d5, "\n", d6, d4),
"11" = paste0(d5, d6, "\n", d6, d5),
"12" = paste0(d6, d6)
)
ggplot(NULL, aes(x = w6_2)) +
geom_histogram(bins = 11, color = "white", fill = "#1aadff", alpha = .5) +
scale_x_continuous(breaks = 2:12, labels = dice) +
labs(title = "Distribution of Dice Rolls",
subtitle = "After 2000 Rolls with 2 Dice",
x = "Combinations (row-wise)", y = "Count") +
theme_minimal()
...Which works fine on his Ubuntu 17.10 machine, yet not on my macOS Sierra machine. On my machine, it looks like this, with the tell-tale boxes indicating missing glyphs. After much research I ended up being able to reproduce the correct plot by manually specifying this free font by adding the following line:
ggplot(NULL, aes(x = w6_2)) +
geom_histogram(bins = 11, color = "white", fill = "#1aadff", alpha = .5) +
scale_x_continuous(breaks = 2:12, labels = dice) +
labs(title = "Distribution of Dice Rolls",
subtitle = "After 2000 Rolls with 2 Dice",
x = "Combinations (row-wise)", y = "Count") +
theme_minimal() +
theme(axis.text.x = element_text(size = rel(2), family = "FreeSerif", lineheight = .5))
Which now works on my macOS machine and on his Ubuntu machine, and possibly requires you to install/load the font via extrafont beforehand.
So, while I'm happy to have a reproducible plot now, my question is this: Why does it work on Linux out of the box, even with various different fonts? According to my research, the fonts that work on Linux with this plot don't actually have glyphs for these dice symbols. I also tried the code on a fresh <rstudio.cloud> session without any extrafont additions and the code above just works fine without specifying the custom font.
Apparently there's a lot of differences concerning fonts across different platforms, but I'm not familiar enough with the R/ggplot2 internals to wrap my head around this minor inconsistency here.
Any insights would be appreciated, as I'm writing code for teaching purposes and platform inconsistencies are good to be aware of.