how to delete warnings in reactive inputs in shiny - r

Question

Could anyone can tell me why I get an error when I change a dataset in first selectInput widget? When I change a dataset from diamonds to mtcars I get an error Could not find 'carat' in input$bins and in the plot just for one second and after that everything works fine. Why it happened?
library(shiny)
library(ggplot2)
data(diamonds)
data(mtcars)
ui <- fluidPage(
column(3,
selectInput("data", "", choices = c('mtcars', 'diamonds')),
uiOutput('server_cols'),
uiOutput('server_bins')
),
column(9,
plotOutput("plot")
)
)
server <- function(input, output) {
data <- reactive({
switch(input$data,
diamonds = diamonds,
mtcars = mtcars)
})
output$server_cols <- renderUI({
data <- data()
nam <- colnames(data)
selectInput('cols', "Choose numeric columns:", choices = nam[sapply(data, function(x) is.numeric(x))])
})
output$server_bins <- renderUI({
if (!is.null(input$cols)) {
df <- data()
x <- eval(input$cols)
max_value <- max(df[,x])
sliderInput('bins','Choose number of bins:', min = 0.1,
max = max_value,
value = max_value/2)
}
})
output$plot <- renderPlot({
if (!is.null(input$cols) & !is.null(input$bins)) {
basicData <- data()
var <- eval(input$cols)
ggplot(basicData, aes_string(var)) +
geom_histogram(binwidth = input$bins, color = 'white', fill = 'red')
}
})
}
shinyApp(ui, server)

Answer 1

Your respective output objects respond to any changes of your input variables. Thus, when you change your dataset via input$data, the plot rebuilds itself, although input$cols did not yet adjust. Actually, try inserting some print("a") inside the output$plot to see that it is called up to three times if you change input$data.
The fix is to rethink your reaction logic and let your elements respond only to specific changes, to get some kind of response "thread".
For example, input$data should only trigger output$server_cols. And output$server_bins should only be triggered by input$cols (because this already implies that input$data changed earlier). Ultimately, output$plot just has to listen to changes of input$bins (because changes in input$cols and input$data always result in changes of input$bins since it is at the end of the thread).
Here is my suggestion using isolate.
library(shiny)
library(ggplot2)
data(diamonds)
data(mtcars)
ui <- fluidPage(
column(3,
selectInput("data", "", choices = c('mtcars', 'diamonds')),
uiOutput('server_cols'),
uiOutput('server_bins')
),
column(9,
plotOutput("plot")
)
)
server <- function(input, output) {
data <- reactive({
switch(input$data, diamonds = diamonds, mtcars = mtcars)
})
output$server_cols <- renderUI({
data <- data()
nam <- colnames(data)
selectInput('cols', "Choose numeric columns:", choices = nam[sapply(data, function(x) is.numeric(x))])
})
output$server_bins <- renderUI({
if (!is.null(input$cols)) {
df <- isolate(data())
x <- eval(input$cols)
max_value <- max(df[,x])
sliderInput('bins','Choose number of bins:', min = 0.1, max = max_value, value = max_value/2)
}
})
output$plot <- renderPlot({
if (!is.null(isolate(input$cols)) & !is.null(input$bins)) {
basicData <- isolate(data())
var <- eval(isolate(input$cols))
ggplot(basicData, aes_string(var)) +
geom_histogram(binwidth = input$bins, color = 'white', fill = 'red')
}
})
}
shinyApp(ui, server)
You might also want to look into updateSelectInput and updateSliderInput if you want to alter Input Elements depending on other input.