Challenge with dynamically generated, interactive R Shiny plots (mostly functioning code) - r

Question

Below is the minimum code. It works, but there is a weird problem. Here is what works:
User can select a number of plots (default is 3).
User can click in a plot and have that value represented (partly works).
Steps to reproduce the "partly works":
At launch, click in plot #3, no problem.
Click in plot #2, nothing happens.
Reduce the number of plots from 3 to 2 and then back to 3.
Click in plot #2, now it works.
Click in plot #1, nothing happens.
Reduce the number of plots from 3 to 1 and then back to 3.
Click in plot #1, now it works.
If you reload the app, and start with step 6 above, all plots are interactive as expected.
rm(list=ls())
library(shiny)
#
# Dynamic number of plots: https://stackoverflow.com/questions/26931173/shiny-r-renderplots-on-the-fly
# That can invalidate each other: https://stackoverflow.com/questions/33382525/how-to-invalidate-reactive-observer-using-code
#
ui <- (fluidPage(sidebarLayout(
sidebarPanel(
numericInput("np", "Plots:", min=0, max=10, value=3, step=1)
)
,mainPanel(
fluidRow(uiOutput("plots"))
)
)))
server <- function(input, output, session) {
val <- reactiveValues()
dum <- reactiveValues(v=0)
obs <- list()
### This is the function to break the whole data into different blocks for each page
plotInput <- reactive({
print("Reactive")
np <- input$np
for(i in 1:np) {
cx <- paste0("clk_p",i); dx <- paste0("dbl_p",i); px <- paste0("p",i)
obs[[cx]] <- observeEvent(input[[cx]], {
req(input[[cx]]); val[[px]] <- input[[cx]]$x; dum$v <- dum$v+1; print(paste("Dum",dum$v))
})
obs[[dx]] <- observeEvent(input[[dx]], {
req(input[[dx]]); val[[px]] <- NULL
})
}
return (list(np=np))
})
##### Create divs######
output$plots <- renderUI({
print("Tag plots")
pls <- list()
for(i in 1:plotInput()$np) {
pls[[i]] <- column(4,
plotOutput(paste0("p",i), height=200, width=200
,click=paste0("clk_p",i)
,dblclick=paste0("dbl_p",i))
)
}
tagList(pls)
})
observe({
print("Observe")
lapply(1:plotInput()$np, function(i){
output[[paste("p", i, sep="") ]] <- renderPlot({
print(paste("Plot",dum$v))
x <- val[[paste0("p",i)]]
x <- ifelse(is.null(x),"NA",round(x,2))
par(mar=c(2,2,2,2))
plot(x=runif(20), y=runif(20), main=i, xlim=c(0,1), ylim=c(0,1), pch=21, bg="gray", cex=1.5)
if(is.numeric(x)) abline(v=x, col="blue")
rm(x)
})
})
})
}
shinyApp(ui, server)

Answer 1

Here is a working version of what you're trying to do:
library(shiny)
ui <- fluidPage(
sidebarPanel(
numericInput("num", "Plots:", 3)
),
mainPanel(
uiOutput("plots")
)
)
server <- function(input, output, session) {
obs <- list()
val <- reactiveValues()
observe({
lapply(seq(input$num), function(i){
output[[paste0("plot", i) ]] <- renderPlot({
click_id <- paste0("clk_p",i);
plot(x = runif(20), y = runif(20), main=i)
if (!is.null(val[[click_id]])) {
abline(v = val[[click_id]], col = "blue")
}
})
})
})
observe({
lapply(seq(input$num), function(i){
id <- paste0("clk_p",i);
if (!is.null(obs[[id]])) {
obs[[id]]$destroy()
}
val[[id]] <- NULL
obs[[id]] <<- observeEvent(input[[id]], {
cat('clicked ', id, ' ', input[[id]]$x, '\n')
val[[id]] <- input[[id]]$x
}, ignoreInit = TRUE)
})
})
output$plots <- renderUI({
lapply(seq(input$num), function(i) {
id <- paste0("plot", i)
plotOutput(id, height=200, width=200, click=paste0("clk_p",i))
})
})
}
shinyApp(ui,server)
A few main pointers for anyone who sees this in the future:
The main issue with the original code was that all the observers were registering only for the last ID. This is a bit of an advanced concept and has to do with the way environments in R work and because they were created in a for loop. The fix for this is to use lapply() instead of a for loop to create the observers
Another issue is that obs was overwriting the observers in the list, but the previous observers still exist and can still fire, so I added logic to destroy() existing observers.
One of the most important rules in shiny is to not place side effects inside reactives (plotInput has side effects) so I rewrote the code in a way that avoids that