在R Shiny中保存数据

Question

how is it possible to save the data the user is inserting, that I only print:

print(c(input$buttonNext,randomNumber,input_radio))

This should not be just printed, but it should be available to me later. My plan is, that people evaluate my plots and then I see for each user the buttons he clicked for each plot.

ui <- fluidPage(

  actionButton("buttonNext", "Next"),

  radioButtons("radio", "Your Decision:",
               choices = c("No Decision" = 'NoDec', "Yes" = 'yes', "No" = 'no'),
               selected = 'NoDec'),

  plotOutput("TimeSeriesPlot")
)


server <- function(input,output,session) {

  clickNext <- reactive({
    isolate(input_radio <- input$radio)
    randomNumber <- input$buttonNext
    print(c(input$buttonNext,randomNumber,input_radio))
    return(randomNumber)
  })

  observeEvent(input$buttonNext,
               {
                 updateRadioButtons(session,'radio',selected = -1)
               })

  output$TimeSeriesPlot <- renderPlot({ 
    i <- clickNext()
    plot(i)
  })

}
shinyApp(server = server, ui = ui)

Answer 1

The approach is the typical for any similar web app, you can:

• Save data as files in a dedicated folder, either in csv format or rds (R compressed format), probably using a combination of user name, time, session etc. in the file name.
• Use some kind of a database connection (anything like SQLite, mysql, PostgreSQL would do. Here you can decide on best design (eg single table for all users, table for each users etc.) depending on application.

Ultimately the decision will depend on the complexity of the application, the amount of data, concerns about data integrity, portability, environment etc.

Exploiting some of the many nice features of R, you could sore as per your request, the parameters specified by the user, but you could even store the actual resulting plot.

Ideally this would require a blob field in a db, but in principle you could build a list of lists, where one of the entries are the plot parameters, and the other is the actual plot (then save everything in a single rds file.

This is a sample implementation. You can save data in an RDS file, or use the alternative implementation (commented out here) that allows you to save the data once when you kill the app.

library(shiny)

# this is not strictly necessary, but is the way I prefer to define an emply data.frame
resultData <- data.frame(times=character(), inputs=numeric(), random=numeric(),
                     decision=character())

ui <- fluidPage(

  actionButton("buttonNext", "Next"),

  radioButtons("radio", "Your Decision:",
           choices = c("No Decision" = 'NoDec', "Yes" = 'yes', "No" = 'no'),
           selected = 'NoDec'),

  plotOutput("TimeSeriesPlot")
)

server <- function(input,output,session) {

  clickNext <- reactive({
    isolate(input_radio <- input$radio)
    randomNumber <- input$buttonNext
    print(c(input$buttonNext,randomNumber,input$radio))
    return(randomNumber)
  })

  observeEvent(input$buttonNext,
           {
             updateRadioButtons(session,'radio',selected = -1)
           })

  output$TimeSeriesPlot <- renderPlot({

    i <- clickNext()
    a <- plot(i)
    results <-   data.frame(times=Sys.time(),inputs=input$buttonNext,random=clickNext(),
  decision=input$radio)

    # with <<- you  make sure the variable is updated outside of the scope of the function
    resultData <<- rbind(resultData,results)
    saveRDS(resultData, file = "data/resultData.rds")
  })

  # alterantive way to save data, when the app is ended
  # session$onSessionEnded({
  #   saveRDS(resultData, file = "data/resultData.rds")
  #   stopApp
  # })
}
shinyApp(server = server, ui = ui)

# analyse <- readRDS(file = "data/resultData.rds")

Please let me know if you'd like me to expand any of the above.