How to link plotly traces for legend and colour selection?

Question

Problem

I am migrating a number of ggplot/ggvis plots to plotly in a shiny application. There is an issue I've encountered regarding the linking of traces. I want to be able to show/hide traces by group on the legend, which is shared between related data frames.

Minimal working example

# load libraries
library(dplyr)
library(plotly)
library(viridis)

# contrived data to represent actual data points
df1 <- data.frame(x = rnorm(100),
                  y = rnorm(100),
                  group = rep(c("G1", "G2", "G3", "G4"), 25))

# contrived data to represent theoretical relationship
df2 <- data.frame(x = c(rep(-2, 4), rep(2, 4)),
                  y = c(seq(1.9, 1, -0.3), seq(-1, -1.9, -0.3)),
                  group = rep(c("G1", "G2", "G3", "G4"), 2))

# create plot with scatter and line traces
df1 %>%
  plot_ly(x = x,
          y = y,
          color = group,
          colors = viridis(n_distinct(group)),
          mode = "markers") %>%
  add_trace(x = x,
            y = y,
            color = group,
            colors = viridis(n_distinct(group)),
            mode = "lines",
            data = df2)

Attempts so far

My online searches and especially reading the plotly documentation have not taken me far.

I can add showlegend = FALSE to the second trace. That does go part way to addressing the challenge, however, I still want to show/hide that trace based on the group value.

Possible solutions

Based on the architecture of plotly, it seems that if I could put the scatter and line onto one trace per group then I would get the desired behaviour. However, it seems that a trace may only have a single "mode", which is why I've taken the approach that I have.

If I continue down the path I've started, I think I need to somehow trap the "on click" event for the legend and show/hide the group traces... but I'm not really sure where to begin with that.

Related / Secondary

In my MWE, I have set the colors argument to viridis. While that is unimportant to the problem, I have not found a way to ensure the colour selection is instead linked to the group (i.e. if the trace for group on df1 is blue, I want to make the same group blue on the trace for df2. If this is non-trivial and warrants a second question (I searched and found no match... possibly because it is trivial and I'm missing something simple), then I'll ask this part separately.

There's a `legendgroup` attribute that could let you do what you want, but only if you plot each group separately (8 traces, here), which is sort of a pain and not very R-like. — alistaire, Apr 05 '16 at 04:26
@alistaire, thank you for pointing me to the `legendgroup` attribute. I'm going to have a play with that to see if I get any traction, however, the number of groups is reactive with shiny so I'm not sure that will be able to provide a solution. — Kevin Arseneau, Apr 05 '16 at 05:19
Since the code for each trace probably won't be too different, you could probably write it programmatically so it repeats as necessary for given inputs. — alistaire, Apr 05 '16 at 05:33
Also, `ggplotly` unifies groups by default, if you don't mind the style. Check out `ggplotly(ggplot(iris, aes(x = Petal.Length, y = Petal.Width, colour = Species)) + geom_point() + geom_line())` — alistaire, Apr 05 '16 at 05:34

score 4 · Accepted Answer · answered Aug 31 '17 at 03:47

Revisiting history for posterity

Several changes to ggplot2 and plotly in the intervening time since this question was first asked. In the current version (4.7.1) there is an argument legendgroup that attends to solving the problem.

Lazy code

Please forgive the minimum effort placed on elegant coding, however, this extension to the MWE demonstrates the ability to show/hide traces by group.

df1_G1 <- df1 %>% filter(group == "G1")
df2_G1 <- df2 %>% filter(group == "G1")
df1_G2 <- df1 %>% filter(group == "G2")
df2_G2 <- df2 %>% filter(group == "G2")
df1_G3 <- df1 %>% filter(group == "G3")
df2_G3 <- df2 %>% filter(group == "G3")
df1_G4 <- df1 %>% filter(group == "G4")
df2_G4 <- df2 %>% filter(group == "G4")

plot_ly(type = "scatter", mode = "markers") %>%
  add_trace(df1_G1, x = df1_G1$x, y = df1_G1$y, color = I("red"),
            legendgroup = "G1", name = "G1 - scatter") %>%
  add_trace(df2_G1, x = df2_G1$x, y = df2_G1$y, color = I("red"),
            legendgroup = "G1", name = "G1 - line", mode = "lines") %>%
  add_trace(df1_G2, x = df1_G2$x, y = df1_G2$y, color = I("green"), 
            legendgroup = "G2", name = "G2 - scatter") %>%
  add_trace(df2_G2, x = df2_G2$x, y = df2_G2$y, color = I("green"),
            legendgroup = "G2", name = "G2 - line", mode = "lines") %>%
  add_trace(df1_G3, x = df1_G3$x, y = df1_G3$y, color = I("blue"),
            legendgroup = "G3", name = "G3 - scatter") %>%
  add_trace(df2_G3, x = df2_G3$x, y = df2_G3$y, color = I("blue"),
            legendgroup = "G3", name = "G3 - line", mode = "lines") %>%
  add_trace(df1_G4, x = df1_G4$x, y = df1_G4$y, color = I("orange"), 
            legendgroup = "G4", name = "G4 - scatter") %>%
  add_trace(df2_G4, x = df2_G4$x, y = df2_G4$y, color = I("orange"), 
            legendgroup = "G4", name = "G4 - line", mode = "lines")

Sample output

Showing the first group G1 deselected. Also note that the colours have been made to match between the scatter and line traces for the same group.