shinyreprex

The aim of shinyreprex is to be able to recreate any reactive or output that is available in a Shiny application outside of said application.

In static documents, like Quarto, it is easy to include the code chunk by including code folding. Due to the interactiveness of Shiny, this isn’t as easy to include out of the box. Reactive depend on inputs set by the user, and need to be replaced in the reactive expressions to be able to run in an environment outside of Shiny.

The script alone reproduces the code, but not the environment it ran in. The packages it depends on can also be captured as an renv lockfile, so the environment can be rebuilt rather than guessed at.

Installation

install.packages("shinyreprex")

To get the development version of shinyreprex, install from GitHub:

require(remotes)
remotes::install_github("AscentSoftware/shinyreprex")

Usage

The following examples takes a couple of inputs, and uses one of them in one reactive, and the other in another reactive. The second reactive is a table output, which can be passed to reprex_reactive to re-create the code that generates the table seen in the UI.

library(shiny)
library(shinyreprex)

ui <- fluidPage(
  h1("Reproducible Code Example"),
  inputPanel(
    sliderInput(
      "min_width",
      "Minimum Petal Width",
      min(iris$Petal.Width),
      max(iris$Petal.Width),
      min(iris$Petal.Width),
      step = 0.1
    ),
    selectInput(
      "summary_fn",
      "Summary Function",
      c("Mean" = "mean", "Median" = "median", "SD" = "sd"),
      selected = "mean"
    ),
    actionButton("update", "Update")
  ),
  fluidRow(
    column(
      width = 5,
      h2("Table"),
      tableOutput("table")
    ),
    column(
      width = 7,
      h2("Code"),
      verbatimTextOutput("code")
    )
  )
)

server <- function(input, output, session) {
  iris_filt <- reactive({
    iris[with(iris, Petal.Width > input$min_width), ]
  }) |>
    bindEvent(input$update)

  summary_tbl <- reactive({
    aggregate(
      Sepal.Width ~ Species,
      data = iris_filt(),
      FUN = get(input$summary_fn)
    )
  }) |>
    bindEvent(input$update)

  output$table <- renderTable(summary_tbl())
  output$code <- renderText(reprex_reactive(summary_tbl))
}

shinyApp(ui, server)

Pinning Package Versions

reprex_reactive emits the library() calls a script needs, but not the versions those packages were at. reprex_lockfile records them, along with the R version and the full recursive dependency tree, as an renv lockfile:

output$lockfile <- downloadHandler(
  filename = function() "renv.lock",
  content = function(file) reprex_lockfile(summary_tbl, lockfile = file)
)

Whoever receives the lockfile rebuilds the environment with:

renv::restore(lockfile = "renv.lock")

In a modular application each module can register the reactives it owns, so a whole-application lockfile needs no reactives passed up to the top level:

moduleServer(id, function(input, output, session) {
  summary_tbl <- reactive(...)

  register_reactives(summary_tbl)
})

# Elsewhere, covering every registered reactive
reprex_lockfile(lockfile = file)

reprex_packages reports the detected packages, either to display them or to let the user narrow the set before pinning it via the packages argument.

Example Application

An example covering all of the above ships with the package:

shiny::runExample("lockfile", package = "shinyreprex")