Shiny

What is Shiny ?

help(package = shiny)

Description:        Shiny makes it incredibly easy to build interactive web applications
                    with R. Automatic "reactive" binding between inputs and outputs and 
                    extensive pre-built widgets make it possible to build beautiful, 
                    responsive, and powerful applications with minimal effort.

• Left side (sidebar panel) : inputs

• Right side (main panel) : outputs

• Engine : two R files, namely ui.R and server.R, generate the app, and reactivity is handled by Javascript

• \(\frac12\) presentation, \(\frac12\) training

• Each participant uses its own computer with a recent version of R and peferably RStudio, as well as an appropriate browser: Google Chrome of Firefox

• Intended public: non-proficient R users

• Skills expected after the presentation :

  • Making an easy Shiny app (such as the Gauss app)
  • Understanding reactivity
  • Self-improvement of these skills by studying the Shiny tutorial
    

Warning The examples given in these slides are mainly conceived to illustrate some concepts, and don't necessarily represent best practices !

• The ui.R file generates the html code of the app

• For instance, textInput("title","Title") generates the following html code:

<label for="title">Title</label>
<input id="title" type="text" value=""/>

(try textInput("title","Title") in the R console)

• There are some html style files (css) in the shiny package folder, as well as Javascript libraries (js files)

• Javascript handles the connection between R and the interface (the server and the client), namely reactivity; we will see in the next slides how reactivity works

• Is it true that using Shiny does not require skills in html/Javascript ?

  • Yes, I am the proof that it's possible
  • But such skills are useful for customization

• Type options(shiny.reactlog=TRUE) at the beginning of server.R, or before running the app, and press Ctrl+F3 while the app is running

• It shows the reactive objects constituting the app, and the connections between them

• There's no reactive conductor in the Gauss app, only direct connections:

• The output object GAUSS is connected to each input object; then it is updated each time an input object is modified

• Let us try to isolate an input:

  output$GAUSS <- renderPlot({
    curve(dnorm(x), from=-3, to=3, axes=FALSE, xlab=NA, ylab=NA, 
      main=input$title, col=isolate(input$color), lwd=input$linewidth)
    axis(1)
  })

• We can isolate all input parameters of the plot: the user firstly selects the parameters and generates the plot by pressing Go! (useful when modifying the inputs generates a time-consuming task)

ui.R

  actionButton(inputId="gobutton", label="Go!")

server.R

  output$GAUSS <- renderPlot({
    input$gobutton # create connection with the Go! button
    curve(dnorm(x), from=-3, to=3, axes=FALSE, xlab=NA, ylab=NA, 
      main=isolate(input$title), col=isolate(input$color), lwd=isolate(input$linewidth))
    axis(1)
  })

• Here the content of input$gobutton has no importance: its role is to generate a value that changes at each press; actually input$gobutton initially is 0 and is incremented at each press

• Exercise: modify the app in order that the plot first time appears only once the button has been pressed (see result in next slide)

Let's make an app returning the n-th Fibonacci number and its inverse:

# Calculate nth number in Fibonacci sequence
fib <- function(n) ifelse(n<3, 1, fib(n-1)+fib(n-2))

shinyServer(function(input, output) {
  output$nthValue    <- renderText({ fib(input$n) })
  output$nthValueInv <- renderText({ 1/fib(input$n) })
})

sidebarPanel(
  numericInput("n", "Enter n:", value=6) 
),
mainPanel(
  h3("n-th Fibonacci number:"),
  textOutput("nthValue"),
  br(),
  h3("inverse n-th Fibonacci number:"),
  textOutput("nthValueInv")
)

old server.R :

# Calculate nth number in Fibonacci sequence
fib <- function(n) ifelse(n<3, 1, fib(n-1)+fib(n-2))

shinyServer(function(input, output) {
  output$nthValue    <- renderText({ fib(input$n) })
  output$nthValueInv <- renderText({ 1/fib(input$n) })
})

new server.R :

# Calculate nth number in Fibonacci sequence
fib <- function(n) ifelse(n<3, 1, fib(n-1)+fib(n-2))

shinyServer(function(input, output) {
  getFib <- reactive({ fib(input$n) })
  output$nthValue <- renderText({ getFib() })
  output$nthValueInv <- renderText({ 1 / getFib() })
})

• In the app below, user uploads a data file and select columns of the dataset

• Thus the UI widgets (selectInput() and actionButton()) are dynamic, not static

• The dynamic UI is made as follows:

output$column_selection <- renderUI({
  dat <- getFile()  ## to be explained below
  if(!is.null(dat)){
    list(
      selectInput("column_x", "Column x:", choices=names(dat)),      
      selectInput("column_y", "Column y:", choices=names(dat)),
      actionButton("gobutton", "Plot!")
    )
  }
})

uiOutput("column_selection")

fileInput("userfile", "Upload a CSV file:")

getFile <- reactive({
  if(!is.null(input$userfile)){
    read.csv(input$userfile$datapath)
  }
})

And finally the plot:

output$plot <- renderPlot({
  if(!is.null(input$gobutton)){
    dat <- getFile()
    if(input$gobutton==0) return(NULL)
    xname <- input$column_x
    yname <- input$column_y
    plot(x=dat[,xname], y=dat[,yname], xlab=xname, ylab=yname)
  }
})

plotOutput("plot")

• conditionalPanel() in ui.R: to hide/unhide something in the UI

• observe({}): this is like reactive({}), but does not return anything

• updateXXXInput() in server.R: to change the value of the input (updateNumericInput, updateTextInput, ...)

• reactiveValues() in server.R: similar to a list, but to store reacting values

• the Shiny tutorial on RStudio website (from which I plagiarized the Fibonacci example)

• the Shiny Google group, to get some help

• the shinyIncubator package: unofficial add-ons for Shiny

• the shinyAce package: a cosy text area input/output

Currently doesn't work within these slides

Shiny allows to render the Javascript charts generated by the rCharts package

About these slides

• This HTML5 slide deck has been created with Ramnath Vaidyanathan's slidify package

• He is also the author or the rCharts package, R package to create interactive javascript visualizations from R, possibly jointly with Shiny

Acknowledgements

I am grateful to Ramnath and to the members of the Shiny Google group for their help