The interface is localized. Skill descriptions and documents are machine-translated by AI into your language; code, commands, paths and technical terms stay in the original, and the author's document governs.
Explore skills
Development · SKILL.md

build-shiny-module

build-shiny-module

Build reusable Shiny modules with proper namespace isolation using NS(). Covers module UI/server pairs, reactive return values, inter-module communication, and nested module composition. Use when extracting a reusable component from a growing Shiny app, building a UI widget used in multiple places, encapsulating complex reactive logic behind a clean interfa…

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill documentation

GitHub

name: build-shiny-module locale: caveman source_locale: en source_commit: 82c77053 fence_basis_commit: 82c77053 translator: "Julius Brussee homage — caveman" translation_date: "2026-04-19" description: > Build reusable Shiny modules with proper namespace isolation using NS(). Covers module UI/server pairs, reactive return values, inter-module communication, and nested module composition. Use when extracting a reusable component from a growing Shiny app, building a UI widget used in multiple places, encapsulating complex reactive logic behind a clean interface, or composing larger applications from smaller, testable units. license: MIT allowed-tools: Read Write Edit Bash Grep Glob metadata: author: Philipp Thoss version: "1.0" domain: shiny complexity: intermediate language: R tags: shiny, modules, namespace, reactive, composition

Build Shiny Module

Create reusable Shiny UI/server module pairs with proper namespace isolation, reactive communication, composability.

When Use

  • Extracting reusable component from growing Shiny app
  • Building UI widget used in multiple places
  • Encapsulating complex reactive logic behind clean interface
  • Composing larger applications from smaller, testable units

Inputs

  • Required: Module purpose and functionality description
  • Required: Input/output contract (what module receives and returns)
  • Optional: Whether module nests other modules (default: no)
  • Optional: Framework context (golem, rhino, vanilla)

Steps

Step 1: Define the Module Interface

Before writing code, define what module accepts and returns:

Module: data_filter
Inputs: reactive dataset, column names to filter on
Outputs: reactive filtered dataset
UI: filter controls (selectInput, sliderInput, dateRangeInput)

Got: Clear contract specifying reactive inputs, reactive outputs, UI elements.

If fail: Interface unclear? Module probably too broad. Split into smaller modules with single responsibilities.

Step 2: Create the Module UI Function

#' Data Filter Module UI
#'
#' @param id Module namespace ID
#' @return A tagList of filter controls
#' @export
dataFilterUI <- function(id) {
  ns <- NS(id)
  tagList(
    selectInput(
      ns("column"),
      "Filter column",
      choices = NULL
    ),
    uiOutput(ns("filter_control")),
    actionButton(ns("apply"), "Apply Filter", class = "btn-primary")
  )
}

Key rules:

  • Function name follows <name>UI convention
  • First argument is always id
  • Create ns <- NS(id) at top
  • Wrap every inputId and outputId with ns()
  • Return tagList() to allow flexible placement

Got: UI function creating namespaced input/output elements.

If fail: IDs collide when using module twice? Check every ID wrapped with ns(). Common miss: IDs inside renderUI() or uiOutput() — these need ns() too.

Step 3: Create the Module Server Function

#' Data Filter Module Server
#'
#' @param id Module namespace ID
#' @param data Reactive expression returning a data frame
#' @param columns Character vector of filterable column names
#' @return Reactive expression returning the filtered data frame
#' @export
dataFilterServer <- function(id, data, columns) {
  moduleServer(id, function(input, output, session) {
    ns <- session$ns

    # Update column choices when data changes
    observeEvent(data(), {
      available <- intersect(columns, names(data()))
      updateSelectInput(session, "column", choices = available)
    })

    # Dynamic filter control based on selected column
    output$filter_control <- renderUI({
      req(input$column)
      col_data <- data()[[input$column]]

      if (is.numeric(col_data)) {
        sliderInput(
          ns("value_range"),
          "Range",
          min = min(col_data, na.rm = TRUE),
          max = max(col_data, na.rm = TRUE),
          value = range(col_data, na.rm = TRUE)
        )
      } else {
        selectInput(
          ns("value_select"),
          "Values",
          choices = unique(col_data),
          multiple = TRUE,
          selected = unique(col_data)
        )
      }
    })

    # Return filtered data as a reactive
    filtered <- eventReactive(input$apply, {
      req(input$column)
      col <- input$column
      df <- data()

      if (is.numeric(df[[col]])) {
        req(input$value_range)
        df[df[[col]] >= input$value_range[1] &
           df[[col]] <= input$value_range[2], ]
      } else {
        req(input$value_select)
        df[df[[col]] %in% input$value_select, ]
      }
    }, ignoreNULL = FALSE)

    return(filtered)
  })
}

Key rules:

  • Function name follows <name>Server convention
  • First argument is always id
  • Additional arguments are reactive expressions or static values
  • Use moduleServer(id, function(input, output, session) { ... })
  • Use session$ns for dynamic UI created inside server
  • Return reactive values explicitly

Got: Server function processing inputs and returning reactive output.

If fail: Reactive values don't update? Check inputs from dynamic UI use session$ns (not outer ns). Module returns NULL? Ensure return() is last expression inside moduleServer().

Step 4: Wire the Module into the Parent App

# In app_ui.R or ui
ui <- page_sidebar(
  title = "Analysis App",
  sidebar = sidebar(
    dataFilterUI("filter1")
  ),
  card(
    DT::dataTableOutput("table")
  )
)

# In app_server.R or server
server <- function(input, output, session) {
  # Raw data source
  raw_data <- reactive({ mtcars })

  # Call module — capture its return value
  filtered_data <- dataFilterServer(
    "filter1",
    data = raw_data,
    columns = c("cyl", "mpg", "hp", "wt")
  )

  # Use the module's returned reactive
  output$table <- DT::renderDataTable({
    filtered_data()
  })
}

Got: Module appears in UI and its returned reactive flows into downstream outputs.

If fail: Module UI doesn't render? Verify id string matches between UI and server calls. Returned reactive is NULL? Check server function actually returns value.

Step 5: Compose Nested Modules (Optional)

For modules containing other modules:

analysisUI <- function(id) {
  ns <- NS(id)
  tagList(
    dataFilterUI(ns("filter")),
    plotOutput(ns("plot"))
  )
}

analysisServer <- function(id, data) {
  moduleServer(id, function(input, output, session) {
    # Call inner module with namespaced ID
    filtered <- dataFilterServer("filter", data = data, columns = names(data()))

    output$plot <- renderPlot({
      req(filtered())
      plot(filtered())
    })

    return(filtered)
  })
}

Key rule: In UI, nest with ns("inner_id"). In server, call with just "inner_id" — moduleServer handles namespace chaining.

Got: Inner module renders correctly within outer module's namespace.

If fail: Inner module's UI doesn't appear? Likely forgot ns() around inner module's ID in outer UI function. Server communication breaks? Check inner module ID matches (no ns() in server call).

Step 6: Test the Module in Isolation

# Quick test app for the module
if (interactive()) {
  shiny::shinyApp(
    ui = fluidPage(
      dataFilterUI("test"),
      DT::dataTableOutput("result")
    ),
    server = function(input, output, session) {
      data <- reactive(iris)
      filtered <- dataFilterServer("test", data, names(iris))
      output$result <- DT::renderDataTable(filtered())
    }
  )
}

Got: Module works correctly in minimal test app.

If fail: Module fails in isolation but works in full app (or vice versa)? Check for implicit dependencies on global variables or parent session state.

Checks

  • Module UI function accepts id as first argument, uses NS(id)
  • Every input/output ID in UI wrapped with ns()
  • Module server uses moduleServer(id, function(input, output, session) { ... })
  • Dynamic UI in server uses session$ns for IDs
  • Module can be instantiated multiple times without ID collisions
  • Reactive return values accessible to parent app
  • Module works in minimal standalone test app

Pitfalls

  • Forgetting ns() in renderUI(): Dynamic UI created inside server must use session$ns — outer ns not available inside moduleServer().
  • Passing non-reactive data: Module arguments that change over time must be reactive expressions. Pass reactive(data) not data.
  • ID mismatch: id string in UI call must exactly match id in server call.
  • Not returning reactives: Module computes something parent needs? Must return() a reactive. Forgetting this is silent bug.
  • Namespace in nested modules: In UI: ns("inner_id"). In server: just "inner_id". Mixing these up causes namespace double-wrapping or missing prefixes.

See Also

  • scaffold-shiny-app — set up app structure before adding modules
  • test-shiny-app — test modules with testServer() unit tests
  • design-shiny-ui — bslib layout and theming for module UIs
  • optimize-shiny-performance — cache and async patterns within modules
Saved skills