Creating a new Module

The `ShinyModule` Class

The ShinyModule class is an interface to be inherited to create modules. The module class does some simple, yet convenient things:

It stores the appId used for shiny::NS() or session$ns().
It stores the moduleName or rather the class name of the module.
It stores the instanceId of the class.
It stores the parentNamespace of a parent module.
It enables you to use the re-use references for similar outputs in different modules. The ShinyModule will add the correct namespacing for you when you use shiny::NS()

Overrides

When inheriting the ShinyModule class it is expected that you override the UI() and / or server() methods. The UI() method should return something akin to a shiny::tagList(). By default they are defined, but do nothing. This allows for just rendering simple text in a module, without any back-end like the DarwinShinyModules::Text module. Or a static image.

Defining a new module

The base interface of creating a new module would be this:

library(DarwinShinyModules)

MyModule <- R6::R6Class(
  classname = "MyModule",
  inherit = ShinyModule,

  private = list(
    # Override UI()
    .UI = function() {},

    # Override server()
    .server = function(input, output, session) {}
  )
)

As you can see it is a new R6 class inheriting from DarwinShinyModules::ShinyModule with a UI and a server method. For a new module you can either directly add functionality to these functions, and use private and active fields and methods within the R6 framework. Or if you are more familiar with functions, you may want to call your own server and ui functions in the module, like so:

library(DarwinShinyModules)
library(shiny)

myUi <- fluidPage(
  tableOutput(outputId = shiny::NS("id", "table"))
)

myServer <- function(input, output, session) {
  output$table <- renderTable({
    head(iris)
  })
}

MyModule <- R6::R6Class(
  classname = "MyModule",
  inherit = ShinyModule,

  private = list(
    # Override UI()
    .UI = function() {
      myUi
    },

    # Override server()
    .server = function(input, output, session) {
      myServer(input, output, session)
    }
  )
)

Let’s remake MyModule putting all the functionality directly on the module, and see what extra funcionality and information exists on the module. In this module I want to:

Display the head() of the iris data frame, using "myTable" as the input / output reference.
Display the appId, moduleName, and instanceId, using "moduleData" as the input / output reference.

library(DarwinShinyModules)

MyModule <- R6::R6Class(
  classname = "MyModule",
  inherit = ShinyModule,

  active = list(
    # Additional active binding field, to fetch the private field .data
    data = function() return(private$.data)
  ),

  private = list(
    # Additional private field, containing the `iris` data.frame
    .data = iris,

    # Override UI()
    .UI = function() {
      shiny::tagList(
        # Output a table "input$app-MyModule_ABCDEFGHIJ__myTable"
        shiny::tableOutput(outputId = shiny::NS(private$.namespace, "myTable")),
        # Output HTML "input$app-MyModule_ABCDEFGHIJ_moduleData"
        shiny::uiOutput(outputId = shiny::NS(private$.namespace, "moduleData"))
      )
    },

    # Override server()
    .server = function(input, output, session) {
      output$myTable <- shiny::renderTable({
        head(private$.data)
      })
      
      # Render the HTML to "MyModule_ABCDEFGHIJ_moduleData"
      output$moduleData <- shiny::renderUI({
        shiny::HTML(paste(
          sprintf("App ID: %s", private$.moduleId),
          sprintf("Module Name: %s", private$.moduleName),
          sprintf("Instance ID: %s", private$.instanceId),
          sep = "<br/>"
        ))
      })
    }
  )
)

To initialize our module object we call the new() method on our defined MyModule:

module <- MyModule$new()

To see what the module contains we can simply print() the module instance:

print(module)
#> <MyModule>
#>   Inherits from: <ShinyModule>
#>   Public:
#>     clone: function (deep = FALSE) 
#>     data: active binding
#>     initialize: function () 
#>     instanceId: active binding
#>     moduleId: active binding
#>     moduleName: active binding
#>     namespace: active binding
#>     parentNamespace: active binding
#>     reactiveValues: active binding
#>     server: function (input, output, session) 
#>     UI: function () 
#>     validate: function () 
#>   Private:
#>     .data: data.frame
#>     .init: function () 
#>     .instanceId: SwlKLU5Ed6ieYxombQ_X
#>     .moduleId: MyModule-SwlKLU5Ed6ieYxombQ_X
#>     .moduleName: MyModule
#>     .namespace: MyModule-SwlKLU5Ed6ieYxombQ_X
#>     .parentNamespace: NULL
#>     .reactiveValues: NULL
#>     .server: function (input, output, session) 
#>     .UI: function () 
#>     checkMethodOverrides: function () 
#>     finalize: function () 
#>     makeInstanceId: function (n = 20)

There are several items listed under Public: appId, clone(), data, id() initialize(), instanceId, moduleName, server(), UI(), and validate(). Note that appId, data, instanceId, and moduleName have a Private counter parts: .appId, .data, .instanceId, moduleName.

We can get the active fields of our module MyModule like so:

module$moduleId
#> [1] "MyModule-SwlKLU5Ed6ieYxombQ_X"
module$instanceId
#> [1] "SwlKLU5Ed6ieYxombQ_X"
module$moduleName
#> [1] "MyModule"
head(module$data)
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#> 1          5.1         3.5          1.4         0.2  setosa
#> 2          4.9         3.0          1.4         0.2  setosa
#> 3          4.7         3.2          1.3         0.2  setosa
#> 4          4.6         3.1          1.5         0.2  setosa
#> 5          5.0         3.6          1.4         0.2  setosa
#> 6          5.4         3.9          1.7         0.4  setosa

Notice that if we initialize multiple instances of the same type of module the instanceId is different per instance:

mod1 <- MyModule$new()
mod2 <- MyModule$new()
mod3 <- MyModule$new()

mod1$instanceId
#> [1] "uRHfdV8IyvQgF7i1jX4q"
mod2$instanceId
#> [1] "Z4vEafHYT6yUXKhGFRsB"
mod3$instanceId
#> [1] "BVkzIbwPcLd3aMfvs5NX"

We can spin up a simple shiny app using our module:

ui <- shiny::fluidPage(
  module$UI()
)

server <- function(input, output, session) {
  module$server(input, output, session)
}

shiny::shinyApp(ui, server)

Shiny applications not supported in static R Markdown documents

Multiple Modules in one ShinyApp

We can also combine multiple modules into one ShinyApp.

We can make a new module called: MyModule2 which inherits from MyModule. I want the module to be exactly the same, but instead of looking at the iris data, I want to look at mtcars.

MyModule2 <- R6::R6Class(
  classname = "MyModule2",
  inherit = MyModule,
  private = list(
    .data = mtcars
  )
)

# initialize a MyModule2 instance
module2 <- MyModule2$new()

We can spin up a new ShinyApp:

ui <- shiny::fluidPage(
  module$UI(),
  module2$UI()
)

server <- function(input, output, session) {
  module$server(input, output, session)
  module2$server(input, output, session)
}

shiny::shinyApp(ui, server)

Shiny applications not supported in static R Markdown documents

Nesting modules

Modules may be nested in other modules. The nested modules can communicate data to one another.

Let’s conciser the following plot function, for the iris dataset. We add an arbitrary size column for this example, which we will manipulate later to change the plot.

data <- iris
data$size <- 0.5

plotFun <- function(data) {
  ggplot2::ggplot(data = data) +
    ggplot2::geom_point(mapping = ggplot2::aes(x = Sepal.Width, y = Sepal.Length, size = size)) +
    ggplot2::facet_grid(. ~ Species)
}

plotFun(data)

What we will do is create two modules, a Table and plotStatic module, in a parent module, ExampleModule.

Let’s create our base module first: ExampleModule.

ExampleModule <- R6::R6Class(
  classname = "ExampleModule",
  inherit = ShinyModule,

  private = list(
    .UI = function() {},
    
    .server = function(input, output, session) {}
  )
)

We can now add our modules that we want to nest into this ExampleModule, when the module is initialized.

ExampleModule <- R6::R6Class(
  classname = "ExampleModule",
  inherit = ShinyModule,
  
  public = list(
    # We are overriding the initialize of `ShinyModule` here
    initialize = function(data) {
      # So we call `super$initialize()` to make sure we still initialize properly
      super$initialize()
      # We are now free to add extra things

      # We add a `Table` module, with the `iris` data set, with no title.
      private$.table <- Table$new(data = data, title = NULL)

      # We add a `PlotStatic` module with same data, also the  `iris` dataset.
      private$.plot <- PlotStatic$new(fun = plotFun, args = list(data = data))
      # In order to add the `private$.table` and `private$.plot` we add them to
      # our private environment. So they are accessable in all other methods
      # like the `.server()` and `.UI()` methods.

      # in order to make sure the namespacing is done correctly, we must update
      # the `parentNamespace` fields in both the `Table` and `PlotStatic` modules.
      # The `parentNamespace` is the namespace of the overarching module, which
      # we can access with `self$namespace`
      private$.table$parentNamespace <- self$namespace
      private$.plot$parentNamespace <- self$namespace
    }
  ),

  private = list(
    # We MUST initialize the private fields. This makes sure the fields exist
    # when we write to them in `initialize()`.
    .table = NULL,
    .plot = NULL,

    .UI = function() {},
    
    .server = function(input, output, session) {}
  )
)

With our modules added, we can call their respective server() and UI() methods in our module.

ExampleModule <- R6::R6Class(
  classname = "ExampleModule",
  inherit = ShinyModule,
  
  public = list(
    initialize = function(data) {
      super$initialize()

      # Create modules
      private$.table <- Table$new(data = data, title = NULL)
      private$.plot <- PlotStatic$new(fun = plotFun, args = list(data = data))

      # Set `parentNamespace`
      private$.table$parentNamespace <- self$namespace
      private$.plot$parentNamespace <- self$namespace
    }
  ),

  private = list(
    .table = NULL,
    .plot = NULL,

    .UI = function() {
      # We must wrap the individual components in something like a `tagList()`,
      # `fluidPage()`, or something similar.
      shiny::tagList(
        private$.plot$UI(),
        private$.table$UI()
      )
    },
    
    .server = function(input, output, session) {
      private$.plot$server(input, output, session)
      private$.table$server(input, output, session)
    }
  )
)

And now that we have something to look at, let’s preview our module

mod <- ExampleModule$new(data = data)

preview(mod)

Shiny applications not supported in static R Markdown documents

The Table and PlotStatic modules are now independent from one another. However, we can change that by adding some extra code in the .server() method.

ExampleModule <- R6::R6Class(
  classname = "ExampleModule",
  inherit = ShinyModule,
  
  public = list(
    initialize = function(data) {
      super$initialize()

      # Create modules
      private$.table <- Table$new(data = data, title = NULL)
      private$.plot <- PlotStatic$new(fun = plotFun, args = list(data = data))

      # Set `parentNamespace`
      private$.table$parentNamespace <- self$namespace
      private$.plot$parentNamespace <- self$namespace
    }
  ),

  private = list(
    .table = NULL,
    .plot = NULL,

    .UI = function() {
      shiny::tagList(
        private$.plot$UI(),
        private$.table$UI()
      )
    },
    
    .server = function(input, output, session) {
      private$.plot$server(input, output, session)
      private$.table$server(input, output, session)

      # `private$.table$bindings$rows_selected` is a binding exposed by the
      # `Table` module. It is a binding from the `DT` package that allows is to
      # see which rows are selected in the table. We can observe this specific
      # event with `observeEvent()`
      shiny::observeEvent(private$.table$bindings$rows_selected, {
        # The `PlotX` modules expose an `args` binding to get access to it's
        # arguments that it uses for plotting. Our `data` argument is there that
        # we specified in `plotFun()`
        private$.plot$args$data[, "size"] <- 0.5
        # We reset the `size` to 0.5 whenever the event triggers.
        
        # We then fetch the rows which rows are selected in the `Table` module
        selectedRows <- private$.table$bindings$rows_selected
        
        # We then update the`"size"` column to change the size to `10` for all
        # the selected rows.
        private$.plot$args$data[selectedRows, "size"] <- 10
      })
    }
  )
)

We can preview our module again.

mod <- ExampleModule$new(data = data)

preview(mod)

Shiny applications not supported in static R Markdown documents

We can see that if we select rows in our table, the size updates in the plot for the selected row.

All the code we used then looks like this:

# Format data
data <- iris
data$size <- 0.5

# Plot function
plotFun <- function(data) {
  ggplot2::ggplot(data = data) +
    ggplot2::geom_point(mapping = ggplot2::aes(x = Sepal.Width, y = Sepal.Length, size = size)) +
    ggplot2::facet_grid(. ~ Species)
}

# Module definition
ExampleModule <- R6::R6Class(
  classname = "ExampleModule",
  inherit = ShinyModule,
  
  public = list(
    initialize = function(data) {
      super$initialize()

      # Create modules
      private$.table <- Table$new(data = data, title = NULL)
      private$.plot <- PlotStatic$new(fun = plotFun, args = list(data = data))

      # Set `parentNamespace`
      private$.table$parentNamespace <- self$namespace
      private$.plot$parentNamespace <- self$namespace
    }
  ),

  private = list(
    .table = NULL,
    .plot = NULL,

    .UI = function() {
      shiny::tagList(
        private$.plot$UI(),
        private$.table$UI()
      )
    },
    
    .server = function(input, output, session) {
      # Pre-specified plot and table code
      private$.plot$server(input, output, session)
      private$.table$server(input, output, session)

      # Communication code
      shiny::observeEvent(private$.table$bindings$rows_selected, {
        private$.plot$args$data[, "size"] <- 0.5
        selectedRows <- private$.table$bindings$rows_selected
        private$.plot$args$data[selectedRows, "size"] <- 10
      })
    }
  )
)

The ShinyModule Class

Overrides

Defining a new module

Multiple Modules in one ShinyApp

Nesting modules

The `ShinyModule` Class