9

In my server.R file, I have a reactive called myNet which generates a visNetwork. In my ui.R, I have multiple tab panels that ideally will have different input widgets that effect the visNetwork.

Is it possible to re-use the same binding?

Currently, when I try to run the code similar to below, I receive an error: Uncaught Duplicate binding for ID vis.

Snippet of server.R

  myNet <- reactive({
    nodes <- df_nodes
    edges <- df_edges        
    visNetwork(nodes, edges, height = '800px')
  })

  output$vis <- renderVisNetwork(
    myNet()
  )

Snippet of ui.R

  ...

  tabPanel("First Panel",
    sidebarLayout(
      sidebarPanel(
        sliderInput("input1", "Title 1", 
                    min=1, max=10, value=1),
        sliderInput("input2", "Title 2",
                    min=1, max=10, value=1),
        sliderInput("input3", "Title 3",
                    min=1, max=10, value=1)
      ),
      mainPanel(
        visNetworkOutput("vis", height = '800px') # *** ISSUE HERE***
      )
    )
  ),
  tabPanel("Second Panel",
    sidebarLayout(
      sidebarPanel(
        sliderInput("input4", "Title 4", 
                    min=1, max=10, value=1),
        sliderInput("input5", "Title 5",
                    min=1, max=10, value=1),
      ),
      mainPanel(
        visNetworkOutput("vis", height = '800px') # *** ISSUE HERE***       
      )
    )
  ), ...
JasonAizkalns
  • 20,243
  • 8
  • 57
  • 116
  • 2
    Nope, sorry. IDs have to be unique. – Matthew Plourde Jun 16 '15 at 15:11
  • @MatthewPlourde What do you think would be the best way to implement this type of behavior? That is, the output/plot is affected by different controls that are on different "panels" -- are conditional panels my only outlet? – JasonAizkalns Jun 16 '15 at 15:13
  • 4
    Abstract the code in `renderVisNetwork` (or whatever your render function is called) into a `reactive` expression, then call that reactive expression in your two render functions. – Matthew Plourde Jun 16 '15 at 15:15
  • @MatthewPlourde Hmm...just feels really un-DRY. Everything is already abstracted into `myNet()` -- so basically I have to create `output$vis_1 <- renderVisNetwork(myNet())` and `output$vis_2 <- renderVisNetwork(myNet())` and so on... – JasonAizkalns Jun 16 '15 at 15:27
  • Exactly. That's as dry as you're going to get. It sounds like there might be a better place for your output than within your tabs, ie, some other place in your interface – Matthew Plourde Jun 16 '15 at 15:28
  • Thanks +1 - maybe I'll play with `conditionalPanel` to accomplish the same thing without re-rendering. May also move this [CodeReview](http://codereview.stackexchange.com/) to get more input. – JasonAizkalns Jun 16 '15 at 15:30
  • I would also suggest opening a github issue or contacting Joe Cheng, since I've seen a lot of people asking this exact question in the past few weeks, so maybe it is time for shiny to start thinking about a potential solution – DeanAttali Jun 16 '15 at 20:19

1 Answers1

10

I ended up opening this GitHub issue. The consensus appears to be that the following is the best method:

output$vis_1 <- output$vis_2 <- renderVisNetwork(myNet())

Alternatively, as alluded to in the comments, you could use a not-so-DRY approach:

output$vis_1 <- renderVisNetwork(myNet())
output$vis_2 <- renderVisNetwork(myNet())

Joe Cheng appropriately made the following comment on the GitHub issue:

"Allowing multiple outputs to share the same ID would have a lot of gnarly side effects, it makes my head hurt to even think about what it would mean. For example, two Leaflet maps with the same ID--but they both have lots of state in the browser, and communicate that state back to the client. How can you make sense of that when they share the same ID?"

All said and told, I was able to ahieve the same behavior using the dynamic UI approach with a switch that toggles which controls are displayed to the user via uiOutput() (in the ui.R file) and output$ui <- renderUI({...}) (in the server.R file).

JasonAizkalns
  • 20,243
  • 8
  • 57
  • 116