Creating legends in ggplot2 to describe discrete axis labels

Question

I want to make a legend in ggplot2 which is not defined by any sort of aes() mapping, or color at all for that matter. I am plotting box and whisker plots for several categories of a variable, but label the categories with numbers 1:8. I simply want to make a legend using ggplot2 that maps each label on the discrete x axis (using scale_x_discrete() ) to the full name of the category. Using the actual name on the axis tick would crowd the plot, especially because I am using facet_wrap() . I just want to use the numbers then have a number-to-scenario legend on the side.

Consider this example:

library(tidyverse)

iris_mod <- iris %>%
  mutate(category = as.integer(as.factor(Species)),
         big_length = ifelse(Sepal.Length > 6.2,1,0))

iris_ex <- ggplot(data = iris_mod, aes(x = Sepal.Width, y = as.factor(category))) +
  geom_boxplot(alpha = 1, outlier.shape = NA, fill = 'steelblue') +  
  theme_bw() +
  scale_y_discrete(name = "Scenario Label") +
  scale_x_continuous(name = "Sepal Width", limits = c(-8,12)) +
  geom_vline(xintercept = 0, lty = 2) +
  coord_flip() +
  facet_wrap( ~ big_length, scales = c('free_x'))

iris_ex

And here is the output. Imagine this with 8 boxplots for four facets, so that there are 8 discrete labels on the bottom axis. I do not want to color 8 plot with different colors. I just want to have numeric labels. I also do not want to put the labels under the plot because the names are somewhat long and would prefer putting them in a legend layer.

How can I do this? The expected output would be a typical legend layer behaving to the normal rules that can be modified in theme() , but rather than having colors in squares, it would correspond to the number. something like

'Legend name'

1 - Flowers without\n Fertilizer

2 - Flowers with\n Fertilizer

3 - Flowers with\n Coarse Soil

Answer 1

Legends in ggplot2 are made to replicate the aesthetic mapping in the plot, so this seems like an off-label use of the legend. I'd suggest faking it with a separate plot like so:

library(patchwork)
(legend <- ggplot(
  data.frame(row = 1:4, label = c('Legend name', 
                                  '1 - Flowers without\n     Fertilizer', 
                                  '2 - Flowers with\n     Fertilizer1', 
                                  '3 - Flowers with\n     Coarse Soil')), 
  aes(1, -row, label = label)) +
  geom_text(hjust = 0, lineheight = 0.8) +
  coord_cartesian(xlim = c(1, 1.5), clip = "off") +
  theme_void())

iris_ex + legend + plot_layout(widths = c(4,1))