Place box plots on the margins of a faceted plot with only plotly::ggplotly compatible geoms

Question

I am making a function that creates a plotly raincloud plot for any given dataset through the ggplotly function. Most of it was very straightforward but the boxplot overlay is not working.

My current best try is this:

raincloud <- ggplot2::ggplot(data = iris) +
  ggplot2::geom_density(mapping = ggplot2::aes(x = Sepal.Length,
                                               fill = Species,
                                               colour = NA),
                        alpha = .85) + 
  ggplot2::geom_rug(mapping = ggplot2::aes(x = Sepal.Length, 
                                       colour = Species),
                size = 1, 
                alpha = .6) +
  ggplot2::geom_boxplot(mapping = ggplot2::aes(x = Sepal.Length,
                                               y = -.1,
                                               fill = Species,
                                               colour = Species),
                        width = 0.2,
                        alpha = .3) +
  ggplot2::facet_grid(rows = 'Species') +
  ggplot2::theme_minimal()


raincloud

plotly::ggplotly(raincloud)

This approach has two main issues:

1. The boxplot is artificially placed inside the plot area, not on the margin. And because I set it at fixed y and fixed width, it can either overlap with the density or look minuscule below it, depending on the height of the distribution. This is why I concluded that I need to place it in the plot margin, over the geom_rug.

2. When generating the ggplotly, the box plot just disappears and leaves only a dotted line, as if it was another geom_rug. I know plotly has boxplot support, but perhaps the combination of horizontal + faceted is too much to ask for its translator.

I insist on the emphasis of needing it to be plotly-compatible, so custom geoms from ggdist, ggExtra or the like are not an option since they are not supported by ggplotly's translator.

Answer 1

I now fully implemented this in my package chronicle. I had to manually draw all lines and rectangles of the boxplot using geom_segment() and geom_rect()

library(chronicle)

make_raincloud(dt = iris,
         value = 'Sepal.Length',
         groups = 'Species')