plotly ignores scale_fill_gradient2

Thank you for writing and maintaining plotly!

When making a scatterplot, I like to specify pch = 21 to print an outline around points. This helps distinguish points when they are close together. When I use this plotting scheme, ggplot2 gives the expected output, but plotly ignores the aesthetics supplied by scale_fill_gradient2() and gives the error number of items to replace is not a multiple of replacement length. I believe this issue is similar to https://github.com/plotly/plotly.R/issues/1149.

# R v4.0.5
library(plotly) # v 4.9.4.1
library(ggplot2) # v 3.3.3

p <- ggplot(data = wdat, mapping = aes(x = Age, y = as.numeric(ABV))) +
  geom_jitter(mapping = aes(fill = Score), size = 2.5, pch = 21) +
  scale_fill_gradient2(low = "#0008ff",
                       midpoint = mean(wdat$Score),
                       mid = "purple",
                       high = "#ff0000") +
  scale_y_continuous(name = "% ABV",
                     limits = c(40, 70),
                     breaks = seq(from = 40, 70, 5)) +
  theme_classic()

p + ggtitle("ggplot2")

ggplotly(p + ggtitle("plotly"))

If I switch my mapping aesthetic to color with pch = 19 and specify a gradient with scale_color_gradient2(), then ggplot2 and plotly give the same output without error. However, the desired outlines are lost.

p <- ggplot(data = wdat, mapping = aes(x = Age, y = as.numeric(ABV))) +
  geom_jitter(mapping = aes(color = Score), size = 2.5, pch = 19) +
  scale_color_gradient2(low = "#0008ff",
                       midpoint = mean(wdat$Score),
                       mid = "purple",
                       high = "#ff0000") +
  scale_y_continuous(name = "% ABV",
                     limits = c(40, 70),
                     breaks = seq(from = 40, 70, 5)) +
  theme_classic()

p + ggtitle("ggplot2")

ggplotly(p + ggtitle("plotly"))

Here is my scheme for downloading the dataset, in case you want to troubleshoot with the same data:

library(dplyr) # v 1.0.5
library(gsheet) # v 0.4.5

url <- "https://docs.google.com/spreadsheets/d/1geSFmGKWhoSzoSblQQK6fDzWSujY-MQ9w2b3VNbTxZk"

wdat <- gsheet::construct_download_url(url = url) %>%
  gsheet::gsheet2tbl(url = .) %>%
  mutate(ABV = gsub("%","",ABV) %>%
           as.numeric() %>%
           round(., 3) %>%
           format(., nsmall = 1)) %>%
  na.omit()

Run the supplied R example and compare the ggplot2 and ggplotly outputs for pch = 21 with scale_fill_gradient2(). Start at the ggplotly conversion path and trace how fill and point outlines are handled; done means the example no longer errors, preserves the gradient, and retains the outlines.