R shiny ERROR: object of type 'closure' is not subsettable [duplicate]

Question

I am using the following code and I always get the this subsettable error. What am I subsetting and where am I wrong. This should be some basic entry code that I modified and which did work at some point and I can't see the error.

Thank you

server.R

library(shiny)

# Define a server for the Shiny app
shinyServer(function(input, output) {

  # Filter data based on selections
  output$table <- renderDataTable({
    data <- read.table("my.csv", sep =',', header =TRUE)
    if (input$shortdesc != "All"){
      data <- data[data$ShortDescription == input$shortdesc,]
    }
    if (input$taken != "All"){
      data <- data[data$Taken == input$taken,]
    }
    if (input$location != "All"){
      data <- data[data$Location == input$location,]
    }
    data
  })

})

ui.R

library(shiny)
# Define the overall UI

shinyUI(
  fluidPage(
    titlePanel("My Items"),

    # Create a new Row in the UI for selectInputs
    fluidRow(
      column(4, 
             selectInput("man", 
                         "What:", 
                         c("All", 
                           unique(as.character(data$ShortDescription))))
      ),
      column(4, 
             selectInput("trans", 
                         "Where:", 
                         c("All", 
                           unique(as.character(data$Location))))
      ),
      column(4, 
             selectInput("cyl", 
                         "Who:", 
                         c("All", 
                           unique(as.character(data$Taken))))
      )        
    ),
    # Create a new row for the table.
    fluidRow(
      dataTableOutput(outputId="table")
    )    
  )  
)

Update:

Why does the example (see below) work and the moment I change it to my.csv it breaks ? If "data" is a buildin function wouldn't that collide also with the example below ? Sorry for not understanding, but this puzzles me.

server.R

library(shiny)

# Load the ggplot2 package which provides
# the 'mpg' dataset.
library(ggplot2)

# Define a server for the Shiny app
shinyServer(function(input, output) {

  # Filter data based on selections
  output$table <- renderDataTable({
    data <- mpg
    if (input$man != "All"){
      data <- data[data$manufacturer == input$man,]
    }
    if (input$cyl != "All"){
      data <- data[data$cyl == input$cyl,]
    }
    if (input$trans != "All"){
      data <- data[data$trans == input$trans,]
    }
    data
  })

})

ui.R.

library(shiny)

# Load the ggplot2 package which provides
# the 'mpg' dataset.
library(ggplot2)

# Define the overall UI
shinyUI(
  fluidPage(
    titlePanel("Basic DataTable"),

    # Create a new Row in the UI for selectInputs
    fluidRow(
      column(4, 
          selectInput("man", 
                      "Manufacturer:", 
                      c("All", 
                        unique(as.character(mpg$manufacturer))))
      ),
      column(4, 
          selectInput("trans", 
                      "Transmission:", 
                      c("All", 
                        unique(as.character(mpg$trans))))
      ),
      column(4, 
          selectInput("cyl", 
                      "Cylinders:", 
                      c("All", 
                        unique(as.character(mpg$cyl))))
      )        
    ),
    # Create a new row for the table.
    fluidRow(
      dataTableOutput(outputId="table")
    )    
  )  
)

Answer 1

Expanding on @Roland 's comment: you have a namespace collision going on. There's a function data in base R, so if R can't find an object data in the current environment, the function data is referenced from the global environment. In your particular case, this happens because ui.R and server.R are in different environments, and, moreover, the individual function bodies all have their own environments. So the data in fluidRow(...) doesn't reference the datafrom output$table. You need to pass around arguments and/or dynamically construct the UI using the functions for that. See for example here.

Update for the updated question:

Replacing datawith mpg in ui.R fixes the problem because mpg is defined as a dataset in the global environment (this is a side effect of library(ggplot2)). So mpg is (almost) always accessible and has the necessary properties. For a fairer comparison, replace mpg in ui.R with data, which should bring back the old problem, because data in the global environment refers to a function and not the data frame you're trying to manipulate.

Super Update with more general solution for dynamically defining and loading selection elements for each dataset:

The server code loops through all the columns of the chosen dataframe and dynamically generates a selection box for every column that has a type other than double. (Uniqueness and equality with doubles is just asking for trouble.) This avoids the scoping issues because the UI elements are created in server.R after a call to the reactive function that loads the data.

server.R

library(shiny)
library(ggplot2)
# Define a server for the Shiny app
shinyServer(function(input, output) {

  get.data <- reactive({
    switch(input$dataset,
           "rock" = rock,
           "pressure" = pressure,
           "cars" = cars,
           "mpg" = mpg,
           "mtcars" = mtcars,
           "diamonds" = diamonds)
  })

  # Filter my.data based on selections
  output$table <- renderDataTable({
    my.data <- get.data()
    for(n in names(my.data)){
        # avoid too many cases ... 
        # unique() with double is just asking for trouble
        if(typeof(my.data[,n]) != "double"){ 
            val <- eval(parse(text=paste0("input$",n)))
            print(val)
            if(val != "All"){
                my.data <- eval(parse(text=paste0("subset(my.data,",n,"==",val,")")))
            }
        }
    }
    my.data
  })

  output$dyn.ui <- renderUI({
      my.data <- get.data()
      sel <- NULL
      for(n in names(my.data)){
          # avoid too many cases ... 
          # unique() with double is just asking for trouble
          if(typeof(my.data[,n]) != "double"){ 
              sel <- c(sel,
                   selectInput(n, n, choices=c("All",unique(my.data[,n])))
                   )
          }
      }
      sel
  })

})

ui.R

library(shiny)

# Define the overall UI

shinyUI(fluidPage(
    titlePanel("Displaying tables dynamically with renderUI() and eval()"),

    sidebarLayout(
        sidebarPanel(h2("Selection"),
                     selectInput("dataset", "Dataset", c("rock", "pressure", "cars","mtcars","diamonds")),
                     # Create a new Row in the UI for selectInputs
                     uiOutput("dyn.ui")

        )
        ,mainPanel(h2("Data"),
           dataTableOutput(outputId="table")       
        )
    )


))