General way to break on unique values in ggplot2 continuous scales - r

Question

I often need to create custom breaks for axis or color/fill/size to reflect the actual data point. Typically in my data, the variable is continuous, but the measurement is at discrete points. I think this may apply to many others from what I see on SO. Below is an example of plotting mpg vs. cyl:
mpg %>%
ggplot(aes(cyl, cty)) +
geom_point() +
scale_x_continuous(breaks = unique(mpg$cyl))
But one does not really want to type different "mpg$cyl" for different exploratory data analysis all the time. So I am here to look for a general solution.
p.s. I read that ggplot does not pass the data to the scale functions -- probably just the range for calculation. I filed an issue but have not yet get any response.

Answer 1

Indeed, ggplot2 does not have a general way to do this. For continuous scales, the training method is to update the range of the scale every time a new layer is examined. It makes sense in the 'grammar of graphics' that scales are mostly independent of geometry layers.
You could, in theory, tackle this problem from the bottom up by making a new Range ggproto class that keeps track of unique values. However, ggplot2 does not export their Range classes, which likely means they don't support tinkering with this. Also, its quite the task to setup a new type of scale.
Instead I'm proposing to hack the ggplot_add() method to leak information from the global plot to the scale. First thing to do is to wrap the constructor of a scale, that tags on an extra class to that scale.
library(ggplot2)
scale_x_unique <- function(...) {
sc <- scale_x_continuous(...)
new <- ggproto("ScaleUnique", sc)
new
}
Next, we want to update the ggplot_add method for our ScaleUnique class. The function beneath essentially checks if there are any user-defined breaks and, if there are none, evaluate the scale's aesthetics in the global plot data.
ggplot_add.ScaleUnique <- function(object, plot, object_name) {
# "waiver" class is for undefined arguments
if (inherits(object$breaks, "waiver")) {
# Find common aesthetic between scale and plot mapping
aes <- intersect(object$aesthetics, names(plot$mapping))
# Find out the expression associated with that aesthetic
aes <- plot$mapping[[aes[[1]]]]
# Evaluate the aesthetic
values <- rlang::eval_tidy(aes, plot$data)
# Assign unique values to breaks
object$breaks <- sort(unique(values))
}
plot$scales$add(object)
plot
}
Now you can use it like any other scale
ggplot(mpg, aes(cyl, cty)) +
geom_point() +
scale_x_unique()
Created on 2021-08-11 by the reprex package (v1.0.0)
This of course only works if the aesthetic is defined in the global plot call and the data is available in the global plot. You could in theory traverse all layers and keep updating your unique values, but this becomes cumbersome.