Plotly legendgroup for subplots so a single legend controls all charts

Question

I'm using plotly in r to generate a number of subplots. A toy example is shown below.

library(shiny)
library(dplyr)
library(plotly)

## Toy Example
ui <- fluidPage(
  h3("Diamonds"),
  plotlyOutput("plot", height = 600)
)

server <- function(input, output, session) {

  # reduce down the dataset to make the example simpler
  dat <- diamonds %>% 
    filter(clarity %in% c("I1", "IF")) %>%
    mutate(clarity = factor(clarity, levels = c("I1", "IF")))

  output$plot <- renderPlotly({

    # Generates the chart for a single clarity
    byClarity <- function(df){

      Clarity <- df$clarity[1];

      plot_ly(df, x = ~carat, y = ~price, color = ~cut, name = ~clarity) %>%
        add_trace(
          type="bar"
          ## Also tried adding this with no success
          # legendgroup = ~cut
        ) %>%
        layout(
          barmode = "stack"
        )
    }

    dat %>% 
      split(.$clarity) %>% 
      lapply(byClarity) %>%
      subplot(nrows = NROW(.), shareX = TRUE, which_layout = "merge")
  })
} 

shinyApp(ui, server)

I would like to make the legends such that clicking on a 'Cut' on the legend will show/hide that 'Cut' from both charts instead of just the chart associated with that legend.

DefaultVsIdeal

I looked at legendgroup but can't figure out how to associate it with cut instead of clarity (clarity is the grouping I'm using to make the subplots).

I also need the solution to work with raw plot_ly and not ggplotly as there are other plot_ly functionalities I need that aren't available in ggplotly.

Any help would be appreciated. I am using plotly_4.5.2, dplyr_0.5.0, and shiny_0.14.

Did you try `Better practice under plotly 4.0 and above.` in this answer: `https://stackoverflow.com/a/41122127/2296728` — Urvah Shabbir, Aug 24 '17 at 03:52

Roman · Answer 1 · 2016-10-18T14:18:36.783

1

Ok, here is a solution using ggplot2:

library(ggplot2)
library(dplyr)
library(plotly)
dat <- diamonds %>% 
  filter(clarity %in% c("I1", "IF")) %>%
  mutate(clarity = factor(clarity, levels = c("I1", "IF")))
# Function for nice labels
k_label <- function(x) {
 c(0, paste0((x)/1000,"K")[-1])
}
# ggplot
p <- ggplot(dat,aes(x=carat, y=price, fill=cut)) + 
           geom_bar(stat="identity") + 
           facet_wrap(~clarity,nrow=2, scales = "free_y") +
           scale_y_continuous(labels = k_label) + 
           theme_minimal() + ylab("") + xlab("") +
           theme(legend.title=element_blank(),
                 panel.grid.major.x=element_blank())
# a plotly
ggplotly(p)

edited Oct 18 '16 at 14:18

answered Oct 18 '16 at 12:08

Roman

17,008
3
36
49

1

Thanks! Unfortunately, I'm looking for a solution that works with raw `plot_ly` and not `ggplotly` as there are other `plot_ly` functionalities I need that aren't available in `ggplotly`... I'll update the question to make that clearer. – adilapapaya Oct 19 '16 at 00:41
1

@adilapapaya I still thinking it is not possible (so far) at least in `R`. Could be helpful to ckeck whether it is possible in `python` or `matlab`. There are some differences between the softwares. – Roman Oct 19 '16 at 07:17
@adilapapaya I'm encountering this problem too, were you ever able to come up with a solution? – Danny Jan 23 '17 at 13:47
@Danny unfortunately not. :-/ – adilapapaya Jan 24 '17 at 18:33
@Danny, if you are still looking for a solution, perhaps my answer can help you out – TrainedMusician Mar 20 '23 at 16:06

shams · Answer 2 · 2016-11-18T16:01:38.920

1

Try adding legendgroup = ~cut to both traces and setting showlegend = F for one of them. Then in layout set showlegend = T

Like this:

plot_ly(df, x = ~carat, y = ~price, color = ~cut, name = ~clarity, legendgroup = ~cut, showlegend = T) %>%
    add_trace( type="bar", legendgroup = ~cut, showlegend = F) %>%
    layout(
      barmode = "stack",showlegend = T
    )

edited Nov 18 '16 at 16:01

answered Nov 18 '16 at 15:30

shams

11
3

score 1 · Answer 3 · answered Mar 20 '23 at 16:05

I have a working example with pure plot_ly code. However, it is not my proudest work:

library(ggplot2)
library(dplyr)
library(plotly)
dat <- diamonds %>% 
    filter(clarity %in% c("I1", "IF")) %>%
    mutate(clarity = factor(clarity, levels = c("I1", "IF")))

firstPlot <- plot_ly(data = dat, x = ~carat, y = ~depth, legendgroup = ~cut, color = ~cut)
secondPlot <- plot_ly(data = dat, x = ~carat, y = ~price, legendgroup = ~cut, color = ~cut)

together <- subplot(firstPlot, secondPlot)
together

Which yields this, but the legend still has duplicates.. This can be resolved with the showLegend = FALSE parameter, but not in the traditional manner..

for (layer in 1L:(length(together$x$data) / 2L)) {
    together$x$data[[layer]]$showlegend <- FALSE
}
together

And this gives the final plots that are still linked together, while only displaying one of the values. If you plan on using more than two subplots, the for loop (which can be an lapply as well) can be altered to select all layers except the first by doing length(unique(dat$cut)):length(together$x$data) which is more robust in the long run.

Note: this works on most plot_ly graphs but not in all (e.g. 'box'), so if you want to apply this but still need help, let me know!

Plotly legendgroup for subplots so a single legend controls all charts

3 Answers3

Linked