Shiny Web Apps

Hello Shiny

library(shiny)
runExample("01_hello")

Anatomy of a Shiny app

• ui function (or ui.R file) controls the layout of outputs and how inputs are collected
• server function (or server.R) controls logic how inputs influence outputs
• R script(s) must be placed in app-specific folder
• RStudio will set all of this up for you if you select Shiny Web App from the New File menu

ui

• fluidPage: renders a webpage that dynamically resizes to fit browser (can also use navbarPage, which accommodates multiple webpages)
  • titlePanel: page title
  • sidebarLayout: creates main panel for outputs and sidebar for input widgets (can also use fluidRow/column to create a custom layout)
    • sidebarPanel: render sidebar
      • sliderInput: widget that collects user input
    • mainPanel: render main panel
      • plotOutput: render output plot

server

• input: list of inputs with names matching widget inputIds
• output: list of reactive outputs with names that match outputIds of ui outputs

A simple example

• see here for a more complex example that illustrates many of the features discussed here

library(shiny)
ui <- fluidPage(
  titlePanel("Hello Shiny!"),
  sidebarLayout(
    sidebarPanel(
      sliderInput(inputId = "bins", label = "Number of bins:", min = 1, max = 50, value = 30)
    ),

    mainPanel(
      plotOutput(outputId = "distPlot")
    )
  )
)

server <- function(input, output) {
  output$distPlot <- renderPlot({
    x    <- faithful$waiting
    bins <- seq(min(x), max(x), length.out = input$bins + 1)
    hist(x, breaks = bins, col = "#75AADB", border = "white",
         xlab = "Waiting time to next eruption (in mins)",
         main = "Histogram of waiting times")
    })
}

Additional UI elements:

• navbarPage allows inclusion of multiple tabPanel elements, each having it’s own sidePanel/mainPanel (or other layout)
  • can include navbarMenu elements, that create drop-down menus with additional tabPanels
• tabsetPanel allows multiple tabPanel elements within a mainPanel (sharing the same sidePanel)
  • conditionalPanel allows sidebarPanel elements to change between tabs
• navlistPanel is similar to tabsetPanel, but the list of tabPanels is in the sidebar rather than across the top of the main panel (similar to a floating table of contents in a rMarkdown document)
• conditionalPanel allows different sidebars for each tab
• see here for more info about Shiny layout options, or see examples in the Shiny Gallery
• text can be formated using Shiny versions of common html tags
• see here for the full list of input widgets

Additional Server elements:

• observe can be used to reactively change the state of a ui widget (eg; update slider based on typed value)

  observe({
  updateSliderInput(session, "pvalue", value = input$typedPval)
  })
  

• req can be used prevent rendering output until input has been selected

  output$Plot1 <- renderPlot({
    req(input$geneID)
    MakePlot(input$geneID , table1)
  })
  

• validate can be used to raise an error if inputs are incorrect

  output$Plot1<- renderPlot({
    validate(
      need(input$Table1_rows_selected != "", "Please select a row from the table")
    )
    MakePlot(input$Table1_rows_selected , table1)
  })
  

  Useful packages

• DT: render fancy tables using DataTables jquery library
  • tables can be searched and sorted by any column
  • supports selection of single rows/ranges of rows
    • selected row number(s) available as an input

      output$Table1 <- DT::renderDataTable({
      DT::datatable(table1, caption = 'Big ass table of results')
      })
      

• shinyBS:
  • bsCollapse creates collapsable panels
  • bsTooltip and bsPopover to add text popups to input widgets and outputs respectively
  • bsModal to put output in a popup window

Performance

• Run as much code as possible server function / render function

• Use reactive to avoid redoing all processing when some inputs change

  • if an output requires inputs from multiple widgets, it will rerun whenever any of the inputs change
  • reactive can be used to store inputs that require a lot of processing
  • if the other outputs change, the processing in the reactive function will not rerun

dataInput <- reactive({
  getSymbols(input$symb, src = "google",
    from = input$dates[1],
    to = input$dates[2],
    auto.assign = FALSE)
})

output$plot <- renderPlot({    
  chartSeries(dataInput(), theme = chartTheme("white"),
    type = "line", log.scale = input$log, TA = NULL)
})

• Use indexing to avoid loading large datasets into memory
  • data can be stored in a database and retrieved as needed
  • Use external tools (tabix, bedtools, samtools) to index data on disk
    • tabix supports bedfiles with arbitrary numbers of columns after three interval columns

genotypes <- system(paste("bcftools view -H -r", snp_pos, "./Data/combined_filtered.vcf.gz"), intern=TRUE) %>%
    str_split('\t')

Deployment

• ShinyApps.io
  • deploy with the press of a button within RStudio
  • free option that will break quite quickly if your app becomes popular (5 apps/25 hours per month), as well as payed options with higher capacity
• ShinyServer
  • locally hosted (specifically, in Mark’s office)
  • allows more flexibility to include external tools and databases (though database servers can be accessed from ShinyApps.io)
• GitHub
  • requires your audience to have RStudio, but combined with private repos, would allow you to control access
  • there is a runGitHub function that allows apps to be run directly from github, but I’m not sure if it would work with a private repo