I'm building a Shiny application where I want to use the dyRangeSelector from dygraphs to provide the input period.
My problem is that I only want the reactive change to fire when the selector receives a "MouseUp"-event, ie., when the user is done with choosing the period. Right now events are dispatched as the selector is moved which results in a lagged app since the computations done for each period take a few seconds. Essentially, Shiny is too reactive for my taste here (I know this it the wrong way round - normally we want the apps to be super reactive).
Can I modify when the reactive request is dispatched?
Here's a small example that shows the problem.
library(quantmod)
library(shiny)
library(dygraphs)
library(magrittr)
# Create simple user interface
ui <- shinyUI(fluidPage(
sidebarLayout(
sidebarPanel(
dygraphOutput("dygraph")
),
mainPanel(
plotOutput("complicatedPlot")
)
)
))
server <- shinyServer(function(input, output) {
## Read the data once.
dataInput <- reactive({
getSymbols("NASDAQ:GOOG", src = "google",
from = "2017-01-01",
auto.assign = FALSE)
})
## Extract the from and to from the selector
values <- reactiveValues()
observe({
if (!is.null(input$dygraph_date_window)) {
rangewindow <- strftime(input$dygraph_date_window[[1]], "%Y-%m-%d")
from <- rangewindow[1]
to <- rangewindow[2]
} else {
from <- "2017-02-01"
to <- Sys.Date()+1
}
values[["from"]] <- from
values[["to"]] <- to
})
## Render the range selector
output$dygraph <- renderDygraph({
dygraph(dataInput()[,4]) %>% dyRangeSelector() %>% dyOptions(retainDateWindow = TRUE)
})
## Render the "complicated" plot
output$complicatedPlot <- renderPlot({
plot(1,1)
text(1,1, values[["from"]])
Sys.sleep(1) ## Inserted to represent computing time
})
})
## run app
runApp(list(ui=ui, server=server))
There is a function in shiny called debounce which might pretty much suit your needs. If you rewrite the limits to a reactive expression (as opposed to observe), you can wrap it into debounce with a specification of time in milliseconds to wait before evaluation. Here is an example with 1000ms:
library(quantmod)
library(shiny)
library(dygraphs)
library(magrittr)
# Create simple user interface
ui <- shinyUI(fluidPage(
sidebarLayout(
sidebarPanel(
dygraphOutput("dygraph")
),
mainPanel(
plotOutput("complicatedPlot")
)
)
))
server <- shinyServer(function(input, output) {
## Read the data once.
dataInput <- reactive({
getSymbols("NASDAQ:GOOG", src = "google",
from = "2017-01-01",
auto.assign = FALSE)
})
## Extract the from and to from the selector
values <- reactiveValues()
limits <- debounce(reactive({
if (!is.null(input$dygraph_date_window)) {
rangewindow <- strftime(input$dygraph_date_window[[1]], "%Y-%m-%d")
from <- rangewindow[1]
to <- rangewindow[2]
} else {
from <- "2017-02-01"
to <- Sys.Date()+1
}
list(from = from,
to = to)
}), 1000)
## Render the range selector
output$dygraph <- renderDygraph({
dygraph(dataInput()[,4]) %>% dyRangeSelector() %>% dyOptions(retainDateWindow = TRUE)
})
## Render the "complicated" plot
output$complicatedPlot <- renderPlot({
plot(1,1)
text(1,1, limits()[["from"]])
Sys.sleep(1) ## Inserted to represent computing time
})
})
## run app
runApp(list(ui=ui, server=server))
This basically means that the reactive expression must be returning the same value for at least 1s to be send to its dependencies. You can experiment with the best time.