R Shiny - Isolating a reactive expression that uses req() to check preconditions - r

Question

The app below contains a selectInput with two options iris and mtcars and a header that displays the current selection.
If the user selects iris, a DT of the corresponding dataset is rendered below the header.
If the user selects mtcars, nothing is rendered below the header.
Here is a screenshot:
I store the selected dataset in a reactive expression, sel_df. The expression checks if the user has selected iris using req(input$dataset=='iris') before returning the corresponding dataset:
sel_df = reactive({
req(input$dataset=='iris')
iris
})
sel_df is passed to renderDT which renders the datatable:
output$df = renderDT({
sel_df()
})
I then render some UI to display the current value of the selectInput using an h3 header, the datatable and a label for the datatable:
output$tbl = renderUI({
tagList(
h3(paste0('Selected:', input$dataset)), # Header should be visible regardless of the value of input$dataset
tags$label(class = 'control-label', style = if(!isTruthy(isolate(sel_df()))) 'display:none;', `for` = 'df', 'Data:'), # Label should only show if input$dataset == 'iris'
DTOutput('df')
)
})
I would like the datatable and its label to only be visible if sel_df outputs a dataset. But due to the way the app is structured, this requires output$tbl (the renderUI above) to take a dependency on sel_df, so that the entire UI chunk disappears whenever input$dataset == 'mtcars'.
My desired output requires output$tbl to only take a dependency on input$dataset, so that the h3 header is always visible regardless of the value of input$dataset. To do this, I tried 'isolating' sel_df using isolate, but output$tbl still calls sel_df each time it's invalidated.
I am not sure where I am going wrong here. I think I may be using isolate incorrectly but I don't know why and was wondering if someone could shed some light.
Here is the app in full:
library(shiny)
library(DT)
ui <- fluidPage(
selectInput('dataset', 'Dataset', c('iris', 'mtcars')),
uiOutput('tbl')
)
server <- function(input, output, session) {
sel_df = reactive({
req(input$dataset=='iris')
iris
})
output$df = renderDT({
sel_df()
})
output$tbl = renderUI({
tagList(
h3(paste0('Selected:', input$dataset)), # Header should be visible regardless of the value of input$dataset
tags$label(class = 'control-label', style = if(!isTruthy(isolate(sel_df()))) 'display:none;', `for` = 'df', 'Data:'), # Label should only show if input$dataset == 'iris'
DTOutput('df')
)
})
}
shinyApp(ui, server)

Answer 1

output$tbl depends on input$dataset, so naturally it is called each time the value of input$dataset changes. sel_df() also depends on input$dataset and gets called whenever it changes. This is all how it is expected to be, I don't think your label is called because it depends on sel_df().
However, please note that when sel_df is NULL, the taglist() call will also return NULL. This is because your sel_df() call fails silently when input$dataset != "iris", and consequently tagList fails as well:
If any of the given values is not truthy, the operation is stopped by raising a
"silent" exception (not logged by Shiny, nor displayed in the Shiny app's UI).
Try this:
server <- function(input, output, session) {
sel_df = reactive({
if(input$dataset=='iris') {
iris
} else {
NULL
}
})
You will find that with mtcars, the h3() tag is shown, but the label is hidden as desired.

Answer 2

If you would like to use req in sel_df() you could use a trycatch in renderDT this addresses the problem mentioned by #January, of tagsList failing when you do not select iris.
You will also need to modify the if statement to use is.null rather, as I use this as the default return value in the trycatch.
library(shiny)
library(DT)
ui <- fluidPage(
selectInput('dataset', 'Dataset', c('iris', 'mtcars')),
uiOutput('tbl')
)
server <- function(input, output, session) {
sel_df = reactive({
req(input$dataset=='iris')
iris
})
output$df = renderDT({
out <- tryCatch(sel_df(), error = function(e) NULL)
return(out)
})
output$tbl = renderUI({
tagList(
tags$h3(paste0('Selected:', input$dataset)), # Header should be visible regardless of the value of input$dataset
tags$label(class = 'control-label', style = if(is.null('df')) 'display:none;', `for` = 'df', 'Data:'), # Label should only show if input$dataset == 'iris'
DTOutput('df')
)
})
}
shinyApp(ui, server)