在图解图例中分隔符号和颜色

Question

I want to achieve the same result as this ggplot code with plotly:

mtcars %>% add_rownames('car') %>% 
ggplot(aes(x = mpg,
         y = disp,
         color = as.factor(gear),
         shape = as.factor(cyl))) +
geom_point()

which results in:

My plotly code is:

library(dplyr)
mtcars %>% add_rownames('car') %>% 
plot_ly(x = ~mpg, 
      y = ~disp,
      text = ~car,
      color = ~as.factor(gear),
      symbol = ~as.factor(cyl),
      mode = 'markers')

which enumerates all possible combinations of colors and shapes in the legend.

Is there a way to have a similar legend to the ggplot?

Answer 1

UPDATE : To overcome some of the issues mentioned for my previous solution (see below) and to increase the usability of the legend, one can simply add the column name to the legend description and then assign the legendgroups to each category.

mtcars %>% rownames_to_column('car') %>%
    plot_ly() %>%
    #Plot symbols for cyl
    add_trace(type = "scatter",
              x = ~mpg, 
              y = ~disp,
              text = ~car,
              symbol = ~paste0(cyl," cyl."),
              mode = 'markers',
              marker = list(color = "grey", size = 15)) %>%
    #Overlay color for gears
    add_trace(type = "scatter",
              x = ~mpg, 
              y = ~disp,
              text = ~car,
              color = ~paste0(gear, " gears"),
              mode = 'markers')

---

This is the previous solution, which is visually closer to the ggplot2 equivalent:

Based on the answer of dww in this thread , we can manually create the groups for cylinders and gears . Subsequently, with the answer of Artem Sokolov this thread , we can add the legend titles as annotations.

mtcars %>% rownames_to_column('car') %>% 
  plot_ly() %>%
  #Plot symbols for cyl
  add_trace(type = "scatter",
            x = ~mpg, 
            y = ~disp,
            text = ~car,
            symbol = ~as.factor(cyl),
            mode = 'markers',  
            legendgroup="cyl",
            marker = list(color = "grey", size = 15)) %>%
  #Overlay color for gears
  add_trace(type = "scatter",
            x = ~mpg, 
            y = ~disp,
            text = ~car,
            color = ~as.factor(gear),
            mode = 'markers', 
            legendgroup="gear") %>%
  #Add Legend Titles (manual)
  add_annotations( text="Cylinders:", xref="paper", yref="paper",
                   x=1.02, xanchor="left",
                   y=0.9, yanchor="bottom",    # Same y as legend below
                   legendtitle=TRUE, showarrow=FALSE ) %>%

  add_annotations( text="Gears:", xref="paper", yref="paper",
                   x=1.02, xanchor="left",
                   y=0.7, yanchor="bottom",    # Y depends on the height of the plot
                   legendtitle=TRUE, showarrow=FALSE ) %>%

  #Increase distance between groups in Legend
  layout(legend=list(tracegroupgap =30, y=0.9, yanchor="top")) 

Unsolved issues:

• Groups have to be created manually
• Groups are just overlayed (color over shape). This means that only the whole group can be dis-/activated in the legend (eg, it is not possible to only show only the entries with 4 cylinders)
• The position of the second legend title (annotation) depends on the height of the plot!

---