Marginal histogram of third dimension in ggplot2? Or histogram for legend?

If we use ggplot2, and ggExtra, on a scatterplot, the ggMarginal will add marginal histograms, but only for X and Y dimensions.

I want to add a histogram onto the color legend. Say we have 3 dimensions, x,y,z. The scatterplot displayes x,y as location, and then z as color. We then have a legend concerning the color of points. Is there a way to add a histogram to the legend, so that the viewer can see relative frequency beyond just guessing from visual interpretation of color?

Starting point:

x = c(rnorm(100, 2, 2), rnorm(100, -1, 3))
y = c(rnorm(100, -1, 3), rnorm(100, 2, 1))
z = c(rnorm(50, -2, 2), rnorm(100, 1, 1), rnorm(50, 3, 2))

d = data.frame(x, y, z)
library(ggplot2)
ggplot(d, aes(x = x, y = y, color = z)) + geom_point() + scale_color_viridis_c()

I want to add a histogram to the legend for z. Any idea how?

Solution

Maybe something like this, using patchwork to give a rotated geom_histogram an arbitrary 1/6th of the width of the available space?

library(patchwork)
library(ggplot2)
library(scales)

x = c(rnorm(100, 2, 2), rnorm(100, -1, 3))
y = c(rnorm(100, -1, 3), rnorm(100, 2, 1))
z = c(rnorm(50, -2, 2), rnorm(100, 1, 1), rnorm(50, 3, 2))
d = data.frame(x, y, z)

p1 <- ggplot(d, aes(x = x, y = y, color = z)) + 
  geom_point(show.legend = F) + 
  scale_color_viridis_c()

p2 <- ggplot(d, aes(y = z, fill = after_stat(y))) + 
  geom_histogram(show.legend = F, bins = 21) + 
  scale_fill_viridis_c() +
  scale_y_continuous(breaks = pretty_breaks(n = 8)) +
  labs(y = "Histogram Legend for z", x = NULL) +
  theme(
    panel.background = element_blank(),
    axis.text.x = element_blank(),
    axis.ticks.x = element_blank(),
    )

p1 + p2 + plot_layout(widths = c(5, 1))

^{Created on 2024-04-28 with reprex v2.1.0}