Legend changes when converting ggplot2 plot to plotly while using multiple "aes labels" [duplicate]

Question

I created a plot using ggplot2 with two different aesthetics for colour and shape. When I convert the ggplot2 plot to plotly via plotly::ggplotly() the different aes seem to create problems.

library(ggplot2) # Version ‘3.4.4’
library(plotly) # Version ‘4.10.4’
library(dplyr)


test_data <- head(iris) |>
  mutate(ID = rank(Sepal.Length),
         category1 = sample(c("A", "B"), 6, replace = TRUE),
         category2 = sample(c("C", "D"), 6, replace = TRUE))
#ggplot
gg_plot <- ggplot(data = test_data) + 
  geom_line(mapping = aes(x = Sepal.Length,
                          y = Sepal.Width,
                          color = category1)) + 
  geom_point(mapping = aes(x = Sepal.Length, y = ID, shape = category2))
gg_plot

#plotly
plotly_plot <- plotly::ggplotly(gg_plot) 
plotly_plot

ggplot2 plot:

plotly plot:

The same or similar problems where already reported a few years ago:

https://github.com/plotly/plotly.R/issues/572

Convert ggplot2 graph to plotly - legend, labels and values

However, the workarounds either do not work for my case or seem very awkward to implement within a dynamic function.

Does anyone know if there is a somewhat easy way to convert a ggplot2 plot with multiple aesthetics to a plotly plot while keeping the legends?

Edit: I phrased my question wrong. I want to keep the legends and the displayed category values (e.g. "C" instead of "(C,1)").

lailaps · Accepted Answer · 2025-10-08 16:40:39Z

1

plotly::ggplotly falsely groups legends and omits their titles. Best you just add the traces in "bare" R {plotly} as described here.

But if you don't want that, you can try this hacky approach:

Splitting the legend title strsplit(plotly_plot$x$layout$legend$title$text, "<br />")[[1]] conveniently gives us all legend groups
iterate over each trace and get the legend group title. For this I use the hovertext plotly_plot$x$data[[i]]$text[1] and see which category is inside
if the group title is the same as before, omit it, if not set it as legendgrouptitle of that trace

library(ggplot2)
library(plotly)
library(dplyr)

gg <- head(iris, 12) |> 
  mutate(
    ID = rank(Sepal.Length),
    category1 = sample(c("A", "B"), 12, replace = TRUE),
    category2 = sample(c("C", "D"), 12, replace = TRUE),
    category3 = sample(c("E", "F", "G"), 12, replace = TRUE),
    category4 = sample(c("H", "I"), 12, replace = TRUE)
  ) |>
  ggplot(aes(x = Sepal.Length)) + 
  geom_line(aes(y = Sepal.Width, color = category1)) + 
  geom_point(aes(y = ID, shape = category2)) +
  geom_point(aes(y = Petal.Length, size = category3), alpha = 0.6) +
  geom_line(aes(y = Petal.Width, linetype = category4))


plotly_plot <- plotly::ggplotly(gg)

categories <- strsplit(plotly_plot$x$layout$legend$title$text, "<br />")[[1]]
group_title <- ""

for (i in seq_along(plotly_plot$x$data)) {
  if (!is.null(plotly_plot$x$data[[i]]$name)) {
     new_title <- categories[sapply(categories, \(x) grepl(x, plotly_plot$x$data[[i]]$text[1]))]
     group_title <- ifelse(new_title == group_title, "", new_title)
     plotly_plot$x$data[[i]]$legendgrouptitle = list(text = group_title)
  }
}

plotly_plot$x$layout$legend$title$text = "" # remove legend title
plotly_plot

This adds a title to each legend group, similar to your ggplot2.

Note: Since this is a hack, it could fail for some legends, I only tested some simple cases.

It could also break, if you remove the category from the hovertext should you pass custom hover texts to ggplotly
I removed the legend title, you might remove that line should you want to keep it

edited Oct 8 at 16:40

answered Oct 8 at 16:10

lailaps

11.3k1 gold badge5 silver badges25 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Erik Oct 9 at 7:23

Thank your for the detailed answer! Sadly i phrased the question wrong, I did not know that the term legend does not apply to the displayed category values which are my main concern. Your answer still gave me a workaround that maybe works: If you loop through the data names they can be changed.

for(i in 1:length(plotly_plot$x$data)){    plotly_plot$x$data[[i]]$name <- stringr::str_extract(plotly_plot$x$data[[i]]$name, "(?<=\\()[^,]*(?=,)")  }

lailaps Oct 9 at 7:48

Yes! It was my goal to show that one can loop through the traces and make changes to the plotly object afterwards :) - glad you found a solution

Collectives™ on Stack Overflow

Legend changes when converting ggplot2 plot to plotly while using multiple "aes labels" [duplicate]

1 Answer 1

2 Comments

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Linked

Related