R Shiny: How to dynamically append arbitrary numbe

问题:

The goal

I am working on a Shiny app that allows the user to upload their own data and focus on the entire data or a subset by providing data filtering widgets described by the below graph The select input "Variable 1" will display all the column names of the data uploaded by the user and the selectize input "Value" will display all the unique values of the corresponding column selected in "Variable 1". Ideally, the user will be able to add as many such rows ("Variable X" + "Value") as possible by some sort of trigger, one possibility being clicking the "Add more" action button.

A possible solution

After looking up online, I've found one promising solution given by Nick Carchedi pasted below

ui.R

library(shiny)

shinyUI(pageWithSidebar(

    # Application title
    headerPanel("Dynamically append arbitrary number of inputs"),

    # Sidebar with a slider input for number of bins
    sidebarPanel(
        uiOutput("allInputs"),
        actionButton("appendInput", "Append Input")
    ),

    # Show a plot of the generated distribution
    mainPanel(
        p("The crux of the problem is to dynamically add an arbitrary number of inputs
          without resetting the values of existing inputs each time a new input is added.
          For example, add a new input, set the new input's value to Option 2, then add
          another input. Note that the value of the first input resets to Option 1."),

        p("I suppose one hack would be to store the values of all existing inputs prior
          to adding a new input. Then,", code("updateSelectInput()"), "could be used to 
          return inputs to their previously set values, but I'm wondering if there is a 
          more efficient method of doing this.")
    )
))

server.R

library(shiny)

shinyServer(function(input, output) {

    # Initialize list of inputs
    inputTagList <- tagList()

    output$allInputs <- renderUI({
        # Get value of button, which represents number of times pressed
        # (i.e. number of inputs added)
        i <- input$appendInput
        # Return if button not pressed yet
        if(is.null(i) || i < 1) return()
        # Define unique input id and label
        newInputId <- paste0("input", i)
        newInputLabel <- paste("Input", i)
        # Define new input
        newInput <- selectInput(newInputId, newInputLabel,
                                c("Option 1", "Option 2", "Option 3"))
        # Append new input to list of existing inputs
        inputTagList <<- tagAppendChild(inputTagList, newInput)
        # Return updated list of inputs
        inputTagList
    })

})

The downside

As pointed by Nick Carchedi himself, all the existing input widgets will undesirably get reset every time when a new one is added.

---

A promising solution for data subsetting/filtering in Shiny

As suggested by warmoverflow, the datatable function in DT package provides a nice way to filter the data in Shiny. See below a minimal example with data filtering enabled.

library(shiny)
shinyApp(
    ui = fluidPage(DT::dataTableOutput('tbl')),
    server = function(input, output) {
        output$tbl = DT::renderDataTable(
            iris, filter = 'top', options = list(autoWidth = TRUE)
        )
    }
)

If you are going to use it in your Shiny app, there are some important aspects that are worth noting.

1. Filtering box type
   • For numeric/date/time columns: range sliders are used to filter rows within ranges
   • For factor columns: selectize inputs are used to display all possible categories
   • For character columns: ordinary search boxes are used
2. How to obtain the filtered data
   • Suppose the table output id is tableId, use input$tableId_rows_all as the indices of rows on all pages (after the table is filtered by the search strings). Please note that input$tableId_rows_all returns the indices of rows on all pages for DT (>= 0.1.26). If you use the DT version by regular install.packages('DT'), only the indices of the current page are returned
   • To install DT (>= 0.1.26), refer to its GitHub page
3. Column width
   • If the data have many columns, column width and filter box width will be narrow, which makes it hard to see the text as report here

---

Still to be solved

Despite some known issues, datatable in DT package stands as a promising solution for data subsetting in Shiny. The question itself, i.e. how to dynamically append arbitrary number of input widgets in Shiny, nevertheless, is interesting and also challenging. Until people find a good way to solve it, I will leave this question open :)

Thank you!

回答1:

are you looking for something like this?

library(shiny)


LHSchoices <- c("X1", "X2", "X3", "X4")


#------------------------------------------------------------------------------#

# MODULE UI ----
variablesUI <- function(id, number) {

  ns <- NS(id)

  tagList(
    fluidRow(
      column(6,
             selectInput(ns("variable"),
                         paste0("Select Variable ", number),
                         choices = c("Choose" = "", LHSchoices)
             )
      ),

      column(6,
             numericInput(ns("value.variable"),
                          label = paste0("Value ", number),
                          value = 0, min = 0
             )
      )
    )
  )

}

#------------------------------------------------------------------------------#

# MODULE SERVER ----

variables <- function(input, output, session, variable.number){
  reactive({

    req(input$variable, input$value.variable)

    # Create Pair: variable and its value
    df <- data.frame(
      "variable.number" = variable.number,
      "variable" = input$variable,
      "value" = input$value.variable,
      stringsAsFactors = FALSE
    )

    return(df)

  })
}

#------------------------------------------------------------------------------#

# Shiny UI ----

ui <- fixedPage(
  verbatimTextOutput("test1"),
  tableOutput("test2"),
  variablesUI("var1", 1),
  h5(""),
  actionButton("insertBtn", "Add another line")

)

# Shiny Server ----

server <- function(input, output) {

  add.variable <- reactiveValues()

  add.variable$df <- data.frame("variable.number" = numeric(0),
                                "variable" = character(0),
                                "value" = numeric(0),
                                stringsAsFactors = FALSE)

  var1 <- callModule(variables, paste0("var", 1), 1)

  observe(add.variable$df[1, ] <- var1())

  observeEvent(input$insertBtn, {

    btn <- sum(input$insertBtn, 1)

    insertUI(
      selector = "h5",
      where = "beforeEnd",
      ui = tagList(
        variablesUI(paste0("var", btn), btn)
      )
    )

    newline <- callModule(variables, paste0("var", btn), btn)

    observeEvent(newline(), {
      add.variable$df[btn, ] <- newline()
    })

  })

  output$test1 <- renderPrint({
    print(add.variable$df)
  })

  output$test2 <- renderTable({
    add.variable$df
  })

}

#------------------------------------------------------------------------------#

shinyApp(ui, server)

回答2:

You need to check for existing input values and use them if available:

  # Prevent dynamic inputs from resetting
  newInputValue <- "Option 1"
  if (newInputId %in% names(input)) {
    newInputValue <- input[[newInputId]]
  }
  # Define new input
  newInput <- selectInput(newInputId, newInputLabel, c("Option 1", "Option 2", "Option 3"), selected=newInputValue)

A working version of the gist (without the reset problem) can be found here: https://gist.github.com/motin/0d0ed0d98fb423dbcb95c2760cda3a30

Copied below:

ui.R

library(shiny)

shinyUI(pageWithSidebar(

  # Application title
  headerPanel("Dynamically append arbitrary number of inputs"),

  # Sidebar with a slider input for number of bins
  sidebarPanel(
    uiOutput("allInputs"),
    actionButton("appendInput", "Append Input")
  ),

  # Show a plot of the generated distribution
  mainPanel(
    p("This shows how to add an arbitrary number of inputs
      without resetting the values of existing inputs each time a new input is added.
      For example, add a new input, set the new input's value to Option 2, then add
      another input. Note that the value of the first input does not reset to Option 1.")
  )
))

server.R

library(shiny)

shinyServer(function(input, output) {

  output$allInputs <- renderUI({
    # Get value of button, which represents number of times pressed (i.e. number of inputs added)
    inputsToShow <- input$appendInput
    # Return if button not pressed yet
    if(is.null(inputsToShow) || inputsToShow < 1) return()
    # Initialize list of inputs
    inputTagList <- tagList()
    # Populate the list of inputs
    lapply(1:inputsToShow,function(i){
      # Define unique input id and label
      newInputId <- paste0("input", i)
      newInputLabel <- paste("Input", i)
      # Prevent dynamic inputs from resetting
      newInputValue <- "Option 1"
      if (newInputId %in% names(input)) {
        newInputValue <- input[[newInputId]]
      }
      # Define new input
      newInput <- selectInput(newInputId, newInputLabel, c("Option 1", "Option 2", "Option 3"), selected=newInputValue)
      # Append new input to list of existing inputs
      inputTagList <<- tagAppendChild(inputTagList, newInput)
    })
    # Return updated list of inputs
    inputTagList
  })

})

(The solution was guided on Nick's hints in the original gist from where you got the code of the promising solution)