downloadHandler: Set reactiveValues inside content function - r

Question

Is it possible to set reactiveValues inside the content part of the downloadHandler? I tried it and don't understand the behavior.
A simple example could be a counter showing how often the download button has been clicked:
library(shiny)
ui <- fluidPage(
downloadButton("downloadData", "Download"),
textOutput("nDownloads"),
actionButton("trig", "get number")
)
server <- function(input, output) {
# Our dataset
data <- mtcars
r.nDownloads <- reactiveValues(n=0)
output$nDownloads <- renderText({
input$trig
paste("number of downloads:", r.nDownloads$n)
})
output$downloadData <- downloadHandler(
filename = function() {
paste("data-", Sys.Date(), ".csv", sep="")
},
content = function(file) {
r.nDownloads$n <- r.nDownloads$n + 1
write.csv(data, file)
}
)
}
shinyApp(ui, server)
If the download button is clicked, the textOutput is grayed out, but not updated. I added an action button as a trigger to force the renderText to be updated. Surprisingly (at least to me) that works: the correct number is shown.
So, somehow the reactiveValue is changed by the downloadHandler, but its dependencies are only invalidated, not updated.
Of course, the proper way to do it would be making the "data"-object reactive and doing the counting there. But I'm curious how the described behavior can be explained.
EDIT:
OK, now I get really confused: I tried what I mentioned above: making "data" reactive and doing the counting there. This could not be simple counting of downloads anymore, because the data-reactive gets only recalculated if it's invalid.
Here is an example with an additional input for the invalidation of "data":
library(shiny)
ui <- fluidPage(
numericInput("nRows", label = "nRows", min=1, max=32, value=15),
downloadButton("downloadData", "Download"),
textOutput("nDownloads")
)
server <- function(input, output) {
r.nDownloads <- reactiveValues(n=0)
# Our dataset
data <- reactive({
isolate({
r.nDownloads$n <- r.nDownloads$n + 1
})
mtcars[1:input$nRows,]
})
output$nDownloads <- renderText({
paste("number of downloads:", r.nDownloads$n)
})
output$downloadData <- downloadHandler(
filename = function() {
paste("data-", Sys.Date(), ".csv", sep="")
},
content = function(file) {
write.csv(data(), file)
}
)
}
shinyApp(ui, server)
But still, I see a similar behavior: Clicking the download button grays the text out, changing "nRows" makes the expected number of downloads (which is now downloads after a change in nRows ;-)) to show up.
Now it gets an actual problem for me: In my real app, a rather complex Excel file can be downloaded. While preparing and formatting the Excel file there can occur events that should lead to some reaction of the app. That's why the download should trigger something. The alternative I can see is, to prepare the Excel file before the user clicks on download (what I would like to avoid, because this can take a few seconds depending on the complexity of the file/formatting).
Am I missing something obvious? If not, I'd appreciate any ideas, how the download event can trigger something in the rest of the app.

Answer 1

The solution is to remove the isolation of the reactiveValues as this prevents the counter from being updated until the numericInput is triggered. This is because data() is dependent on input$nrows.
library(shiny)
ui <- fluidPage(
numericInput("nRows", label = "nRows", min = 1, max = 32, value = 15),
downloadButton("downloadData", "Download"),
textOutput("nDownloads")
)
server <- function(input, output) {
r.nDownloads <- reactiveValues(n = 0)
# Our dataset
data <- reactive({
r.nDownloads$n <- r.nDownloads$n + 1
mtcars[1:input$nRows,]
})
output$nDownloads <- renderText({
paste("number of downloads:", r.nDownloads$n)
})
output$downloadData <- downloadHandler(
filename = function() {
paste("data-", Sys.Date(), ".csv", sep = "")
},
content = function(file) {
write.csv(data(), file)
}
)
}
shinyApp(ui, server)
With regards to the deeper problem, it would be inefficient to constantly prepare a complex Excel file if there is no guarantee that the user would download the file. What you can try to do is:
Keep your data in a reactive method (e.g. data()).
Write a method to prep your data for downloading (e.g. prepExcel(data)) which returns your prepped data.
Pass (1) and (2) into the content of the downloadHandler() like this: write_xx(prepExcel(data())) or pipe the data into the write_xx function like this data() %>% prepExcel() %>% write_xx() where xx is the method used to output your final file e.g. write_xlsx or write_csv etc.
I hope this helps.