Introduction to Shiny

What is Shiny?

• A framework for building interactive web applications in R
• No JavaScript or HTML knowledge required
• Applications run entirely in R, served by a local or remote server
• The browser is the interface; R handles all computation

The three-part structure

Every Shiny application consists of exactly three parts:

1. UI — a description of the interface the user sees
2. Server — R code that processes inputs and produces outputs
3. shinyApp(ui, server) — the call that creates and runs the app

The UI is a function call

ui <- fluidPage(
  sliderInput("n", "Points", min = 10, max = 500, value = 100),
  plotOutput("scatter")
)

• fluidPage creates a responsive HTML page
• Input functions create controls (here, a slider)
• Output functions create placeholders (here, for a plot)
• Input and output functions use a unique ID, usually as their first argument

The server is a function

server <- function(input, output) {
  output$scatter <- renderPlot({
    x <- rnorm(input$n)
    y <- rnorm(input$n)
    plot(x, y)
  })
}

• input$n reads the current value of the slider named "n"
• output$scatter assigns to the placeholder named "scatter"
• renderPlot({...}) wraps the plotting code and makes it reactive

Reactive programming: the core idea

In a script, code runs once, top to bottom.

In Shiny:

• Shiny monitors which inputs each output reads
• When an input changes, every output that depends on it re-executes automatically
• You write what to compute, not when to recompute

The reactive graph

input$n  ──►  renderPlot  ──►  output$scatter

• Each arrow is a reactive dependency
• Change input$n → Shiny invalidates renderPlot → it re-runs

This is the mental model you need for everything that follows.

Input–output pairs

Common input widgets

• sliderInput — draggable numeric slider
• selectInput — dropdown menu
• radioButtons — mutually exclusive options
• checkboxGroupInput — multiple selections
• textInput / numericInput — free text or number entry
• actionButton — triggers computation on click

A first real application

ui <- fluidPage(
  sliderInput("bins", "Bins", min = 5, max = 50, value = 20),
  plotOutput("hist")
)
server <- function(input, output) {
  output$hist <- renderPlot({
    hist(rnorm(500), breaks = input$bins,
         col = "steelblue", border = "white")
  })
}
shinyApp(ui, server)

Move the slider — the histogram redraws instantly.

Why Shiny code looks unusual

• Shiny code is declarative: you declare what each output is, not when to update it
• The server function never “runs” in the normal sense — Shiny runs pieces of it reactively
• Nested function calls in the UI are not evaluated immediately; they build an HTML description
• This takes some adjustment but becomes natural quickly

Summary

• Every app: UI + server + shinyApp
• IDs connect UI controls to server logic
• render* functions make outputs reactive to inputs
• Shiny re-runs only the parts of the server that depend on changed inputs