Shiny: Dynamic Number of Output Elements/Plots

Question

I want to make a reactive display, that displays a different number of plots depending on which value of the input selector is chosen. In the case of the mtcars dataset, let's say I want to let the user choose beetween cutting by Nr. of Gears or Nr. of Carburatos for the plots to be produced.

Looking at unique(mtcars$gear) we see it has 4 3 5 so 3 possible values, while unique(mtcars$carb) has 4 1 2 3 6 8 so 6 possible values. I therefore want to produce 6 separate plots when Nr. of Carburators is selected and only 3 plots when Nr. of Gears is selected. I've played with conditionalPanel but it invariably blows up after I switch between selectors once or twice. Help?

Shiny UI:

library(shiny)
library(googleVis)

shinyUI(bootstrapPage(
    selectInput(inputId = "choosevar",
              label = "Choose Cut Variable:",
              choices = c("Nr. of Gears"="gear",
                          "Nr. of Carburators"="carb")),
    htmlOutput('mydisplay')  ##Obviously I'll want more than one of these... 
#   conditionalPanel(...)
  ))

Shiny Server:

shinyServer(function(input, output) {
   #Toy output example for one out of 3 unique gear values:
    output$mydisplay <- renderGvis({
    gvisColumnChart(    
    mtcars[mtcars$gear==4,], xvar='hp', yvar='mpg' 
    )
  })  
})

Polyclitus

Answer 1

Inspired from this, you could do:

ui.R

shinyUI(pageWithSidebar(            
        headerPanel("Dynamic number of plots"),            
        sidebarPanel(
                selectInput(inputId = "choosevar",
                            label = "Choose Cut Variable:",
                            choices = c("Nr. of Gears"="gear", "Nr. of Carburators"="carb"))
        ),            
        mainPanel(
                # This is the dynamic UI for the plots
                uiOutput("plots")
        )
))

server.R

library(googleVis)
shinyServer(function(input, output) {
        #dynamically create the right number of htmlOutput
        output$plots <- renderUI({
                plot_output_list <- lapply(unique(mtcars[,input$choosevar]), function(i) {
                        plotname <- paste0("plot", i)
                        htmlOutput(plotname)
                })

                tagList(plot_output_list)
        }) 

        # Call renderPlot for each one. Plots are only actually generated when they
        # are visible on the web page. 


        for (i in 1:max(unique(mtcars[,"gear"]),unique(mtcars[,"carb"]))) {
                local({
                        my_i <- i
                        plotname <- paste0("plot", my_i)

                        output[[plotname]] <- renderGvis({
                                data <- mtcars[mtcars[,input$choosevar]==my_i,]
                                if(dim(data)[1]>0){
                                gvisColumnChart(    
                                        data, xvar='hp', yvar='mpg' 
                                )}
                                else NULL
                        })  
                })
        }

})

It basically creates htmlOutput plots dynamically and binds the googleVis plots when there is data in the subset.

Keratin

Wow, the tagList is what I was missing all along. Thanks a bunch! – Polyclitus

I found the top link very helpful, +1 – Perl

Answer 2

Have you tried making a total of 9 (6+3) conditionalPanels? If yes, have you tried 3 naked output panels that have conditionals inside them to switch between the plots, and 3 additional conditionalPanels for the non-overlapped plots?

Another way might be to make a single output panel with an internal conditional, and then stack your 3 or 6 plots into a single plot ala

if(cond1) {
   par(mfrow=c(3,1))
   plot1
   plot2
   plot3
} else {
   par(mfrow=c(3,2))
   plot1
   plot2
   plot3
   plot4
   plot5
   plot6
}

Radiotelegraphy