`ggplot2::stat_summary` with `ggplot2::geom_crossbar` inducing error in `ggplotly`

[akunihiro]

---

I am running into an issue after recently updating plotly (and possibly ggplot2). Using ggplot2::geom_crossbar in ggplot2::stat_summary and visualizing with ggplotly results in the following error:

Error in `[[<-.data.frame`(`*tmp*`, nm, value = numeric(0)) : 
  replacement has 0 rows, data has 3

Here is a reprex:

library(ggplotly)
library(plotly)
library(tidyverse)
library(ggbeeswarm)

data("iris")

iris_plot <- iris %>%
  ggplot(aes(x = Species, y = Sepal.Length)) +
  
  # plot quasirandom distribution
  geom_quasirandom(shape = 21,
                   # size = 2,
                   size = 4,
                   alpha = 1,
                   stroke = 0.1,
                   varwidth = TRUE) +
  
  # add median bar
  stat_summary(fun = function(x) log2(median(2^x)),
               geom = "crossbar",
               linetype = "dashed",
               color = "black") +
  
  # add mean bar
  stat_summary(fun = function(x) log2(mean(2^x)),
               geom = "crossbar",
               linetype = "solid",
               color = "black") +
  
  # log2 scale
  scale_y_continuous(trans = "log2")

# plotting directly works
iris_plot

# plot with ggplotly results in error
ggplotly(iris_plot) 

Changing the geom argument in ggplot2::stat_summary to "pointrange" allows the plot to be visualized with ggplotly

[cpsievert]

Here's a more minimal reprex:

library(plotly)

p <- ggplot(iris, aes(x = Species, y = Sepal.Length)) +
    stat_summary(fun = median, geom = "crossbar")

ggplotly(p)