Column selector for color

[yvanrichard]

The layer selector is great as it helps explore the data without having to build a Shiny app. In the same vein, how can one add a selector to define the variable used to colour points?

The selector would be a dropdown list of column names for columns containing hex colors, and the javascript would update the deck accordingly. I know it's possible in theory, but I haven't found a solution, despite my many attempts.

It would be extremely useful.

[anthonynorth]

It's possible, and you can achieve it without re-serialising data on selection change. The pattern we've used in the past to do this:

• ensure all columns needed for all selections are serialised
• use update_x_layer() on selection change

Something like this:

library(shiny)
library(rdeck)


ui <- fillPage(
  rdeckOutput("map", height = "100%"),
  absolutePanel(
    top = 10, left = 10,
    selectInput("variable_select", "variable", c("foo", "bar"), "bar")
  )
)

rdeck_data <- vctrs::data_frame(
  position = wk::xy(runif(1e3, -180, 180), runif(1e5, -90, 90), wk::wk_crs_longlat()),
  foo = runif(1e3),
  bar = runif(1e3) ** 2
)

map <- rdeck() |>
  add_scatterplot_layer(
    id = "scatterplot_layer",
    name = "foo/bar",
    data = rdeck_data,
    get_radius = 3,
    radius_units = "pixels",
    get_fill_color = scale_color_linear(col = bar),
    # columns in tooltip ensures they're serialised
    pickable = TRUE,
    tooltip = everything(),
    # alternatively, this hack also works
    # a get_ or _value param with cur_value() -> serialise everything
    # get_whatever_value = cur_value()
  )

server <- function(input, output, session) {
  output$map <- renderRdeck(map)

  observe({
    rng <- range(dplyr::pull(rdeck_data, !!input$variable_select))

    rdeck_proxy("map") |>
      update_scatterplot_layer(
        id = "scatterplot_layer",
        get_fill_color = scale_color_linear(col = !!input$variable_select, limits = rng)
      )
  })
}

shinyApp(ui, server)

[anthonynorth]

Helps if I read your question properly.. The above is for a shiny app -- which you aren't using.

Changing a variable for colour, without shiny? {rdeck} doesn't support this out of the box. Your option right now is to duplicate layers, each using a specific variable for scaling

You can use rdeck.controls rdeck_layer_dropdown() to wrap an rdeck instance in a selector.

[yvanrichard]

Oh right! Both your answers are very informative, actually. Thank you.

In the meantime, I ended up with the attached R function to create an HTML output to meet my needs, that doesn't rely on packages (with some help from Gemini). The function takes a data frame with coordinates and pre-calculated colours for each variable of interest, and it creates a self-contained HTML file to show a scatter plot where the user can change the colour variable and the size and opacity of the points. It also supports HTML in tooltips. Just in case someone else finds it useful.

deck-scatter-function-and-usage.txt