event_data() triggering unexpected warning about registering

Using event_data("plotly_hover", source="mysource") seems to trigger the following warning with the newest plotly version:
Warning: The 'plotly_hover' event tied a source ID of 'mysource' is not registered. In order to obtain this event data, please add event_register(p, 'plotly_hover') to the plot (p) that you wish to obtain event data from.

library(shiny)
library(plotly)

ui <- fluidPage(
        plotlyOutput("myplot"),
        verbatimTextOutput("hovered")
      )

server <- function(input, output, session) {
    
    output$myplot <- renderPlotly({
        p <- plot_ly(faithful, x=~eruptions, y=~waiting, source="mysource")
        return(p)
    })
    
    output$hovered <- renderPrint({
      event_data("plotly_hover", source="mysource")
    })

}

shinyApp(ui = ui, server = server)

The relevant entry from the plotly source code:

# It's possible for event_data() to execute before any 
  # relevant input values have been registered (i.e, before 
  # relevant plotly graphs have been executed). Therefore, 
  # we delay checking that a relevant input value has been 
  # registered until shiny flushes
  session$onFlushed(
    function() {
      eventIDRegistered <- eventID %in% session$userData$plotlyShinyEventIDs
      if (!eventIDRegistered) {
        warning(
          "The '", event, "' event tied a source ID of '", source, "' ",
          "is not registered. In order to obtain this event data, ", 
          "please add `event_register(p, '", event, "')` to the plot (`p`) ",
          "that you wish to obtain event data from.",
          call. = FALSE
        )
      }
    }
  )```

Huh, the warning seems to go away if you avoid return()

library(shiny)
library(plotly)

ui <- fluidPage(
  plotlyOutput("myplot"),
  verbatimTextOutput("hovered")
)

server <- function(input, output, session) {
  
  output$myplot <- renderPlotly({
    plot_ly(faithful, x=~eruptions, y=~waiting, source="mysource")
  })
  
  output$hovered <- renderPrint({
    event_data("plotly_hover", source="mysource")
  })
  
}

shinyApp(ui = ui, server = server)

How would you handle returning NULL?
I need to stop plot when there is no data (or other criteria are not matched). Currently, I return(NULL),
but I guess this is not optimal. I would rather see a message saying that data needs be loaded, but I still haven't figured out how to do this...
And yes, I have the impression that without the return I see fewer warnings, but I still see some, most probably due to the return(NULL)...
Thanks a lot.
B

Are there any docs on how to use event_register? I recently upgrade to the newest plotly r version and am getting event_register errors on all of my scatter plots. I have two different sources for two different plots.

Edit: I figured it out but it would be helpful if there were additional documentation and examples outside of the webinar for this update.

Wrap user-supplied expression in evalq(), closes #1528

How would you handle returning NULL?

Avoid calling event_data() until the relevant output is non-NULL, like this #1538 (comment)

If someone can some with an example where this is sufficiently hard, I might be willing to add an official way to suppress the warning in event_data()

Are there any docs on how to use event_register?

Sort of, see ?event_register. It's worth noting though that you shouldn't need to use event_register() unless you need an event that isn't registered by default...here's a hack to get at the default events:

p <- plotly_build(plot_ly())
p$x$shinyEvents

 [1] "plotly_hover"          
 [2] "plotly_click"          
 [3] "plotly_selected"       
 [4] "plotly_relayout"       
 [5] "plotly_brushed"        
 [6] "plotly_brushing"       
 [7] "plotly_clickannotation"
 [8] "plotly_doubleclick"    
 [9] "plotly_deselect"       
[10] "plotly_afterplot"

Wrap user-supplied expression in evalq(), closes #1528

Merge pull request #1539 from ropensci/renderWidgetPrep

[pull] master from ropensci:master (#2)

I am still running into this issue of Warning: The 'plotly_hover' event tied a source ID of 'mysource' is not registered. In order to obtain this event data, please add event_register(p, 'plotly_hover') to the plot (p) that you wish to obtain event data from.
I've copied Carson's code from this post, and mixed it with my own editable plotly in a test shiny app and can get it to work. When I run it in a larger app I am getting the same warning but the observe() or event_data("plotly_relayout") does not seem to be working with this app. I can get the shape to move but no underlying data is changing. I am completely stuck about how to get an event_register() to work and why I am not seeing any observe occur with the edited plot.... any guidance on how to code in an event_register() or why I can't see any interaction on this plot?

You could give the warning a class so that these specific warnings could be turned off.

  session$onFlushed(
    function() {
      eventIDRegistered <- eventID %in% session$userData$plotlyShinyEventIDs
      if (!eventIDRegistered) {
        rlang::warn(
          paste0("The '", event, "' event tied a source ID of '", source, "' ",
          "is not registered. In order to obtain this event data, ", 
          "please add `event_register(p, '", event, "')` to the plot (`p`) ",
          "that you wish to obtain event data from."),
          class = "plotly_unregistered_event_warning"
        )
      }
    }
  )

These warnings can be disabled in the shiny server code by wrapping this around the desired code

suppressWarnings(..., classes = "plotly_unregistered_event_warning")

I started getting this warning as soon as I started using source argument in plot_ly(). Seems there is no clear cut solution to this as I see above. Is there any?

You could give the warning a class so that these specific warnings could be turned off.

  session$onFlushed(
    function() {
      eventIDRegistered <- eventID %in% session$userData$plotlyShinyEventIDs
      if (!eventIDRegistered) {
        rlang::warn(
          paste0("The '", event, "' event tied a source ID of '", source, "' ",
          "is not registered. In order to obtain this event data, ", 
          "please add `event_register(p, '", event, "')` to the plot (`p`) ",
          "that you wish to obtain event data from."),
          class = "plotly_unregistered_event_warning"
        )
      }
    }
  )

These warnings can be disabled in the shiny server code by wrapping this around the desired code

suppressWarnings(..., classes = "plotly_unregistered_event_warning")

Where does eventID come from?

[pull] master from ropensci:master (#2)

Plotly june2022 (#18)

@cpsievert

How would you handle returning NULL?

Avoid calling event_data() until the relevant output is non-NULL, like this #1538 (comment)

If someone can some with an example where this is sufficiently hard, I might be willing to add an official way to suppress the warning in event_data()

Are there any docs on how to use event_register?

Sort of, see ?event_register. It's worth noting though that you shouldn't need to use event_register() unless you need an event that isn't registered by default...here's a hack to get at the default events:

@cpsievert I'm a bit confused, if the recommended way to avoid this warning is a req() above the call, why is the warning message itself talking about event_register()?

Also, the req() method isn't working consistently for me in a larger app. IIUC when the renderPlotly() and the reactive holding event_data() are both depending on the same reactive, there's no guarantee as to which will be executed first when that reactive becomes available--so if the shiny reactive queue happens to execute the event_data first, we see the warning.

@dvg-p4

@dvg-p4

Hmm, maybe this will help.

@cpsievert

Addressing the Need to Suppress Warnings in `event_data()`

Regarding this statement, @cpsievert:

If someone can some with an example where this is sufficiently hard, I might be willing to add an official way to suppress the warning in event_data()

Creating a minimal reproducible example from such applications is not only quite time-consuming but often fails to replicate the issue, as the warning tends to disappear in simpler scenarios while persisting in the more complex app. Additionally, the suggestion provided in the warning message itself does not address the root problem effectively.

"In order to obtain this event data, please add event_register(p, 'plotly_click') to the plot (p) that you wish to obtain event data from."

Even when the recommended event_register calls are added to the plot_ly() function, the warnings persist:

output$plotlyplot <- renderPlotly({
  plot_ly(
    data = mtcars,
    x = ~mpg, y = ~carb,
    type = "scatter",
    mode = "markers"
  ) %>%
    event_register('plotly_hover') %>% 
    event_register('plotly_click') %>% 
    event_register('plotly_doubleclick')
})

Cluttered Logs

In larger Shiny applications with multiple pages for example and several interactive Plotly plots on each page, these warnings can quickly accumulate, so that the Log files become cluttered with unnecessary warnings, obscuring genuinely important warnings that need attention. This makes debugging significantly harder. Another complication is that we can't simply use suppressWarnings() to handle this issue effectively. In one of my apps, I had to resort to a workaround like this:

options(warn = -1)
# Perform some calculations and then create a Plotly plot
options(warn = 0)

This is far from ideal and can suppress more warnings than intended.

Debugging Becomes Impossible with `options(warn = 2)`

Setting options(warn = 2) -which turns warnings into errors- causes the app to crash simply because interactive Plotly plots are used. This behavior blocks effective debugging.

Example Shiny App

To illustrate, consider a Shiny app where users dynamically generate multiple plots based on selected time ranges. For simplicity, I’ve adapted the scenario to use a numeric input, allowing users to specify between 1 and 5 plots. Each plot supports interactivity via hover, click, and double-click events.
In such a case, the repetitive warnings for multiple plots add unnecessary noise to the logs (especially since everything seems to be working correctly!). While it might technically be possible to suppress these warnings correctly in this scenario, the process is neither obvious nor straightforward. This is evidenced by the number of related GitHub issues and StackOverflow questions, highlighting that many developers struggle with this same problem. The current behavior adds unnecessary complexity and frustration, making a proper solution or clear guidance essential.

library(shiny)
library(plotly)

# Simulated dataset generator
generate_data <- function(id) {
  data.frame(
    x = rnorm(50, mean = id * 5),
    y = rnorm(50, mean = id * 3),
    group = sample(letters[1:3], 50, replace = TRUE)
  )
}

# UI
ui <- fluidPage(
  titlePanel("Dynamic Plotly Plots"),
  sidebarLayout(
    sidebarPanel(
      numericInput("num_plots", "Number of Plots:", value = 1, min = 1, max = 5),
      verbatimTextOutput("hover_info"),
      verbatimTextOutput("click_info"),
      verbatimTextOutput("dblclick_info")
    ),
    mainPanel(
      uiOutput("plots_ui") # Placeholder for dynamically generated plots
    )
  )
)

# Server
server <- function(input, output, session) {
  plot_data <- reactive({
    lapply(1:input$num_plots, generate_data)
  })
  
  # Dynamically generate UI for plots
  output$plots_ui <- renderUI({
    num <- input$num_plots
    plot_output_list <- lapply(1:num, function(i) {
      plotlyOutput(outputId = paste0("plot_", i), height = "300px")
    })
    do.call(tagList, plot_output_list)
  })
  
  # Render dynamic plots
  observe({
    lapply(1:input$num_plots, function(i) {
      local({
        id <- i
        output[[paste0("plot_", id)]] <- renderPlotly({
          plot_ly(
            data = plot_data()[[id]],
            x = ~x, y = ~y,
            type = "scatter",
            mode = "markers",
            color = ~group,
            source = paste0("plot_source_", id),
            text = ~paste("Point ID:", 1:nrow(plot_data()[[id]]))
          ) %>%
            layout(title = paste("Dynamic Plot", id))
        })
      })
    })
  })
  
  # Observe Events (Click, Dbclick, Hover) for all plots
  observe({
    lapply(1:input$num_plots, function(i) {
      local({
        id <- i
        # observeEvent(event_data("plotly_hover", source = paste0("plot_source_", id)), {
        observe({
          req(plot_data())
          hover <- event_data("plotly_hover", source = paste0("plot_source_", id))
          if (!is.null(hover)) {
            output$hover_info <- renderText({
              paste("Hover on Plot", id, ":\nX =", hover$x, "\nY =", hover$y)
            })
          }
        })
        # observeEvent(event_data("plotly_click", source = paste0("plot_source_", id)), {
        observe({
          req(plot_data())
          click <- event_data("plotly_click", source = paste0("plot_source_", id))
          if (!is.null(click)) {
            output$click_info <- renderText({
              paste("Click on Plot", id, ":\nX =", click$x, "\nY =", click$y)
            })
          }
        })
        observe({
          req(plot_data())
          click <- event_data("plotly_doubleclick", source = paste0("plot_source_", id))
          if (!is.null(click)) {
            output$dblclick_info <- renderText({
              paste("DBL-Click on Plot", id, ":\nX =", click$x, "\nY =", click$y)
            })
          }
        })
      })
    })
  })
}

# Run the app
shinyApp(ui = ui, server = server)

@cpsievert

@cpsievert, I have an example where this is sufficiently hard-- seemingly impossible for all practical purposes.

I have an app where the plot height is set dynamically. This leads to a catch-22 situation where it is impossible to get the correct behavior without needless warnings that can't be suppressed.

Here's a simplified version:

# dynamic plot height with click event
library(shiny)
library(plotly)

ui = fluidPage(
    titlePanel('event_data bug report'),
    radioButtons('height', 'Plot height',
                 c('Short' = '200', 'Tall' = '400'), '200'),
    h3('Last click update'),
    verbatimTextOutput('clickInfo'),
    uiOutput('dynamicTimeSeries')
)

server = function(input, output) {
  time_series_output = renderPlotly({
    y = if (input$height == '200') 1:5 else seq(1, 9, 2)
    plot_ly(x = 1:5, y = y, type = 'scatter', mode = 'lines',
            source = 'example') %>%
      event_register('plotly_click')
  })
  output$timeSeries = reactive(time_series_output())
  output$dynamicTimeSeries = renderUI({
    plotlyOutput('timeSeries', height = as.numeric(input$height))
  })
  output$clickInfo = renderPrint({
    req(time_series_output())
    print(Sys.time())
    event_data('plotly_click', 'example')
  })
}

shinyApp(ui = ui, server = server)

In general, using event_data with a dynamic plot has been a huge pain.

First, whatever you're trying to do with session$onFlushed when emitting the warning doesn't seem to work with a dynamic plot. I get false positive warnings.

To fix that, when responding to clicks, I need to call req to make sure the plot exists. I had to create a layer of indirection with a new variable, time_series_output, so that I have something to call req on.

Here's the catch-22: when I add this req statement before checking the click event, it adds a reactive dependency to the plot. That means every time the plot is updated, this code runs again. So now I can't update the plot without the click event running. When you run this example app, it shows that the click event is run every time the plot height is changed. Not only is it run again, but there's no way to distinguish real clicks from plot updates.

So I can either remove the false warnings, or I can have an app that responds correctly to click events, but I can't do both.

I propose a rule of thumb: if the warning is not suppressible, there should be an option to remove it. There's too much weird stuff happening within Shiny for you to guarantee users that it'll only warn at the correct times, and the hoops you want us to jump through are a lot more burdensome than simply turning off the warning.

Edit: Ok, I actually did find a workaround. It requires manually removing the event from the session object after event_data is called.

# remove the click event
eventID = paste('plotly_click', 'example', sep = "-")
session = shiny::getDefaultReactiveDomain()
session$userData$plotlyInputStore[[eventID]] = NULL

Navigation Menu

Repository navigation

event_data() triggering unexpected warning about registering #1528

Addressing the Need to Suppress Warnings in `event_data()`

Cluttered Logs

Debugging Becomes Impossible with `options(warn = 2)`

Example Shiny App

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

event_data() triggering unexpected warning about registering #1528

Description

Activity

bklingen commented on May 10, 2019

cpsievert commented on May 14, 2019

baj12 commented on May 15, 2019

dchen71 commented on May 17, 2019

cpsievert commented on May 24, 2019

CDorich commented on Sep 15, 2021

samssann commented on Nov 2, 2021

sanjmeh commented on Dec 4, 2021

davidseres commented on Apr 27, 2022

dvg-p4 commented on Jan 31, 2024

dvg-p4 commented on Jan 31, 2024

asadow commented on Oct 16, 2024

trafficonese commented on Dec 3, 2024

Addressing the Need to Suppress Warnings in event_data()

Cluttered Logs

Debugging Becomes Impossible with options(warn = 2)

Example Shiny App

wmay commented on Mar 18, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Addressing the Need to Suppress Warnings in `event_data()`

Debugging Becomes Impossible with `options(warn = 2)`