ggplot+coord_map自定义tooltip：nearPoint匹配偏移问题求助

I get it—when using coord_map() (especially projections like polyconic), the nearPoints() function keeps mismatching the correct hover location because it’s comparing the plot’s transformed coordinates against your raw geographic (lat/long) data. The projection warps the coordinate system, so your original long/lat values don’t line up with what’s actually rendered on the screen. Here’s how to fix this:

The Root Cause #

coord_map() applies a projection transformation to your data in the background, converting geographic coordinates to flat plot-ready coordinates. But nearPoints() uses your untransformed counties data, while input$plot_hover returns coordinates from the projected plot space. This mismatch is why the tooltips are offset.

Solution: Use Transformed Plot Data for nearPoints #

We need to extract the exact data ggplot uses to render the map (after projection) and use that for the nearPoints() check. Here’s the revised working code:

library(ggplot2)
library(mapdata)
library(data.table)
library(shiny)

map.county <- map_data('county')
counties <- data.table(map.county)

ui <- fluidPage(
  titlePanel("NearPoints using a map"),
  div(
    style = "position:relative",
    plotOutput("county_map", hover = hoverOpts("plot_hover", delay = 100, delayType = "debounce")),
    uiOutput("hover_info")
  )
)

server <- function(input, output) {
  # Create plot and extract transformed data reactively
  map_plot <- reactive({
    # Build the base plot with projection
    p <- ggplot(counties, aes(x=long, y=lat, group = group)) +
      geom_polygon(colour = "grey") +
      coord_map("polyconic")
    
    # Extract the data ggplot actually renders (after projection)
    gb <- ggplot_build(p)
    transformed_data <- gb$data[[1]]
    
    # Merge back original region/subregion labels using group ID
    transformed_data <- merge(
      transformed_data, 
      unique(counties[, .(group, region, subregion)]), 
      by = "group"
    )
    
    # Return both plot and transformed data
    list(plot = p, data = transformed_data)
  })
  
  output$county_map <- renderPlot({
    map_plot()$plot
  })
  
  output$hover_info <- renderUI({
    hover <- input$plot_hover
    transformed_data <- map_plot()$data
    
    # Use transformed data for accurate nearPoints matching
    point <- nearPoints(transformed_data, hover, threshold = 5, maxpoints = 1, addDist = TRUE)
    if (nrow(point) == 0) return(NULL)
    
    # Use direct pixel coordinates for tooltip positioning
    style <- paste0(
      "position:absolute; z-index:100; background-color: rgba(245, 245, 245, 0.85); ",
      "left:", hover$coords_img$x + 10, "px; top:", hover$coords_img$y - 100, "px;"
    )
    
    wellPanel(
      style = style,
      p(HTML(paste0(
        "<b>Region:</b> ", point$region, "<br/>",
        "<b>County:</b> ", point$subregion, "<br/>"
      )))
    )
  })
}

shinyApp(ui = ui, server = server)

Key Changes Breakdown #

1. Extract Transformed Data: ggplot_build() gives us the data after projection, which has x/y values that exactly match the plot’s coordinate space.
2. Merge Metadata: We link the transformed data back to the original region and subregion labels using the group column, so we can still display those details in the tooltip.
3. Simplified Tooltip Placement: Instead of calculating percentages, we use hover$coords_img to get the exact pixel position of the mouse on the plot—this makes tooltip placement precise without extra math.
4. Reactive Plot Object: Wrapping the plot and data in a reactive ensures it’s only built once (or when dependencies change), boosting performance.

This approach makes sure nearPoints() compares like-for-like coordinates, so the hover always matches the correct county—even with complex projections like polyconic.

内容的提问来源于stack exchange，提问作者Rochelle Smits