OOP for Shiny

R is a functional programming language. Built with R, the web framework Shiny also employs this paradigm. This document shows one of many ways to employ OO patterns in a Shiny app with Q7 type system.

Each module is an object that has the following members:

• id: constructor parameter, an aribitrary string to represent the whole module
• ns: constructor parameter, a Shiny namespace function, defaults to depend on id, but can be overidden
• dev_mode: constructor parameter, a simple boolean value to tell the module to turn some things on or off
• ui: a ui module fragment
• server: definition and call to a server module fragment
• abstractions necessary to the above
• state: an reference object made by reactiveValues() or new.env(), which esposes the what’s meaningful to other modules.

ShinyModule <- type(function(
  id = uuid::UUIDgenerate(FALSE), 
  ns = NS(id), 
  dev_mode = FALSE
){
  # The following are not strictly necessary, as they will be overidden,
  # but serve as good reminders
  ui <- function(){tagList(
    # ...
  )} 
  
  server <- function(){callModule(function(input, output, session){
    # ...
  }, id)}
  
  state <- reactiveValues()
})

LoadData <- ShinyModule %>% 
  implement({
    ui <- function(){
      
    }
    
    server
    
    
  })

PlotChart <- ShinyModule %>% 
  implement({
    ui
    
    server
    
    
  })

require(shiny)
require(Q7)

loadData <- LoadData()
plotChart <- PlotChart()

shinyApp(
  function(){fluidPage(sidebarLayout(  # UI
    loadData$ui(),
    plotChart$ui()
  ))},
  function(input, output, session){ # Server
    loadData$server()
    plotChart$server()
  }
)

Segregation of namespace and state.

If you use shinydashboard, each is naturally a candidate for a top-level module.

If you just need to define a input or output panel that has similarly organized and configured, you don’t need a module.