Debounce timer not restarting when input is changed - r

Question

I am trying to create a date input that automatically updates the inputted value to the end of the selected month. The problem I was having was when I ran this first bit of code (below) was that if the user tries to manually change the date by typing in a date, they are unable to because the input is updated instantly to the end of the month before the user can finish typing.
library(shinydashboard)
library(lubridate)
ui <- dashboardPage(
dashboardHeader(),
dashboardSidebar(),
dashboardBody(
dateInput(
inputId = "date",
label = "End of Month Date",
value = ceiling_date(x = Sys.Date() + 365, unit = "month") - 1,
startview = "year"
)
)
)
server <- function(input, output, session) {
observe({
if(is.Date(input$date) & length(input$date) > 0){
if(input$date != ceiling_date(input$date, unit = "month") - 1) {
updateDateInput(
session,
inputId = "date",
value = ceiling_date(x = input$date, unit = "month") - 1
)
}
}
})
}
shinyApp(ui, server)
So, I tried implementing debounce (as shown below - server code change only) so that it would delay the input from updating until the user was finished typing; however, I am running into an issue. The timer does not reset on each key stroke as I understand it should. Instead, the timer operates as if I were using the throttle function and starts when the input is first changed and doesn't reset when the input changes.
server <- function(input, output, session) {
observe({
dateinputdelay <- debounce(r = reactive(input$date), millis = 2000)
if(is.Date(dateinputdelay()) & length(dateinputdelay()) > 0){
if(dateinputdelay() != ceiling_date(dateinputdelay(), unit = "month") - 1) {
updateDateInput(
session,
inputId = "date",
value = ceiling_date(x = input$date, unit = "month") - 1
)
}
}
})
}

Answer 1

Edit: the best I can get is with the code below, but because the "fancy date input thingy" has its own internal updating mechanism separate from shiny's reactive concept, this debouncing is only addressing one source of problem.
The first trick is that the debounce needs to happen before the block of code is even started. That is, it "de-bounces" the start of the dependent code blocks; once they are started, it does not help.
The second is knowing that observe blocks attempt to run their code eagerly (see shiny docs and read "Details"), whereas reactive blocks are relatively lazy -- they only run dependent code as needed. It's the "eager" part that may be hurting.
Additionally, you use side-effect in the observe block to update the input field, but you never store the value elsewhere. I suggest that it might be better to calculate the new value in one place (in a functional manner, like a reactive block that should not operate in side-effect) and then use that later.
server <- function(input, output, session) {
dateinputdelay <- debounce(reactive(input$date), 2000)
end_of_month <- reactive({
# print("react!")
x <- dateinputdelay()
if (is.Date(x) & length(x) > 0) {
if (x != ceiling_date(x, unit = "month") - 1) {
x <- ceiling_date(x = input$date, unit = "month") - 1
}
}
x
})
observe({
# print("observe!")
updateDateInput(
session,
inputId = "date",
value = end_of_month()
)
})
}
(I've kept a couple of print statements in there, not because they help much here, but because they can be a good tool to see when/how-frequently reactitivity is causing code-blocks to run.)
As I said up top, I suspect this behavior is indicative of something within dateInput, not within the subsequent reactive or observe blocks.