ggplotly: geom_histogram(aes(fill = ...)) + facet_*() results in tooltip NAs

I noticed a bug when trying to use ggplotly on a geom_histogram with fill in the aes and facet_*. I was able to recreate it on a simple example using mtcars. (I am using ggplot2 v 3.0.0 and plotly v 4.8.0.) I looked through the issues and it seems similar to #693.

This is the plot I am trying to use ggplotly on:

library(ggplot2)
library(dplyr)
library(plotly)
tp_na <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb, fill = cyl), binwidth = 1) + 
  facet_grid(.~vs)
tp_na

Using ggplotly, the values of the tooltip for the x-axis variable are all NA:

ggplotly(tp_na)

(There is one exception to the NA in this reprex. The first red bar in the left facet has the correct value. The rest are NA. But, the same is not true for the plot in which I first encountered this problem.)

My workaround is:

tp <- mtcars %>% mutate(cyl = as.factor(cyl), carb= as.factor(carb)) %>%
  ggplot() + 
  geom_bar(aes(x = carb, fill = cyl))  +
  facet_grid(.~vs)
ggplotly(tp)

But, I would prefer to use geom_histogram(binwitdth = 1) for because the x-axis values for my desired plot are days (0, 1, 2, ..., 21), and I like the histogram for that better than the bar chart.

I tried a few other plots to see if what could possibly be causing the error:

# remove binwidth arg. still see NAs in tooltip
tp_na2 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb, fill = cyl)) + 
  facet_grid(.~vs)
ggplotly(tp_na2)
# using a more continuous variable. still see NAs.
tp_na3 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = qsec, fill = cyl)) + 
  facet_wrap(~vs)
ggplotly(tp_na3)
# add a text aes() to force x label to show. In tooltip, there are two entries for x, both correct
tp2 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb, fill = cyl, text = paste("Carb:", carb)), binwidth = 1) + 
  facet_grid(.~vs)
ggplotly(tp2)
# no facets, it works as expected
tp3 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb, fill = cyl), binwidth = 1) 
ggplotly(tp3)
# no fill, it works as expected
tp4 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb), binwidth = 1) + 
  facet_grid(.~vs)
ggplotly(tp4)

# adding just carb as the text aes results in multiple NAs in the tooltip
tp5 <- mtcars %>% mutate(cyl = as.factor(cyl)) %>%
  ggplot() + 
  geom_histogram(aes(x = carb, fill = cly, text = carb), binwidth = 1) + 
  facet_grid(.~vs)
ggplotly(tp5)

As you can see in the examples, facet_grid vs. facet_wrap doesn't seem to make a difference. There must be something about the fill + facet combo that goes wonky. I find it especially interesting that adding the text inside aes seems to work, but then I have double labels. I hope this is helpful for others. 🤷‍♀️

Beginne damit, das bereitgestellte mtcars-reprex mit ggplot2, plotly, geom_histogram, fill und facet_grid oder facet_wrap auszuführen, und vergleiche es dann mit dem aufgeführten Workaround sowie mit Fällen ohne Facets oder fill. Als erledigt gilt die Aufgabe, wenn die Tooltips des Histogramms die x-Achsenwerte anstelle von NAs anzeigen und dabei das Faceting- und fill-Verhalten erhalten bleibt.