Reactive variables

Case Studies: Building Web Applications with Shiny in R

Dean Attali

Shiny Consultant

Code duplication

data <- gapminder
data <- subset(
    data,
    lifeExp >= input$life[1] & lifeExp <= input$life[2]
)
if (input$continent != "All") {
    data <- subset(
        data,
        continent == input$continent
   )
}

Duplicated 3 times
1. renderTable()
2. renderPlot()
3. downloadHandler()

Reactive variables reduce code duplication

Duplicated code ⇒ multiple places to maintain
- When code needs updating
- When bugs need fixing
Easy to forget one instance, leading to bugs
Use reactive() variables instead of code duplication

Reactive variables

output$my_table <- renderTable({
    data <- gapminder
    data <- subset(
        data,
        lifeExp >= input$life[1] & lifeExp <= input$life[2]
    )
})

my_data <- reactive({

  data <- gapminder
  data <- subset(
         data,
         lifeExp >= input$life[1] & lifeExp <= input$life[2]
     )

})

output$my_table <- renderTable({
    my_data()
})

Reactive variables caching

Reactive variables cache their values
Remember their own value
Do not run again if dependencies didn't change

Reactive variables caching

  output$table <- renderTable({
      fit_model(input$num)
  })

  output$plot <- renderPlot({
      ggplot(
        fit_model(input$num), ...)
  })

fit_model() takes 5s
fit_model() called twice = 10s

x <- reactive({
    fit_model(input$num)
})

output$table <- renderTable({
    x()
})

output$plot <- renderPlot({
    ggplot(x(), ...)
})

x() called twice, code inside x runs once
fit_model() called once = 5s

Reactive variables are lazy

Lazy variable = not calculated until value is needed

x <- reactive({
    fit_model(input$num)
})

output$download <- downloadHandler(
    filename = "x.csv",
    content = function(file) {
        write.csv(x(), file)
    }
)

x() only runs when download is requested, not every time input$num changes

Let's practice!

Case Studies: Building Web Applications with Shiny in R