overlayServer: Add interactive overlays to a Shiny plot

Description

This function sets up server-side infrastructure to support draggable and resizable overlays on a plot. This may be useful in applications where users need to define regions on the plot for further input or processing. Currently, the overlays are only designed to move along the x axis of the plot.

Usage

overlayServer(
  outputId,
  nrect,
  width = NULL,
  snap = "none",
  colours = overlayColours,
  opacity = 0.25,
  icon = shiny::icon("gear"),
  stagger = 0.045,
  style = list(),
  debug = FALSE
)

Value

A shiny::reactiveValues() object with the following named fields:

n: Number of overlays (read-only).
active: Logical vector of length n; indicates which overlays are active.
show: Logical vector; controls whether overlays are visible.
editing: Index of the overlay currently being edited via the dropdown menu, if any; NA otherwise (read-only).
last: Index of the most recently added overlay (read-only).
snap: Coordinate snapping function.
px, pw: Numeric vector; overlay x-position and width in pixels (see note).
py, ph: Numeric vector; overlay y-position and height in pixels (read-only).
cx0, cx1: Numeric vector; overlay x-bounds in plot coordinates (see note).
label: Character vector of labels shown at the top of each overlay.
outputId: The output ID of the plot display area (read-only).
bound_cx, bound_cw: x-position and width of the bounding area in plot coordinates (read-only).
bound_px, bound_pw: x-position and width of the bounding area in pixels (read-only).
bound_py, bound_ph: y-position and height of the bounding area in pixels (read-only).
stagger: Amount of vertical staggering, as proportion of height.
style: Named list of character vectors; additional styling for rectangular overlays.
update_cx(i): Function to update cx0/cx1 from px/pw for overlays i (see note).
update_px(i): Function to update px/pw from cx0/cx1 for overlays i (see note).

Note: Fields marked "read-only" above should not be changed. Other fields can be changed in your reactive code and this will modify the overlays and their properties. The fields px and pw which specify the pixel coordinates of each overlay can be modified, but any modifications should be placed in a shiny::isolate() call, with a call to ov$update_cx(i) at the end to update cx0 and cx1 and apply snapping. Similarly, the fields cx0 and cx1 which specify the plot coordinates of each overlay can be modified, but modifications should be placed in a shiny::isolate() call with a call to ov$update_px(i) at the end to update px and pw

and apply snapping. The i parameter to these functions can be left out to apply changes to all overlays, or you can pass in the indices of just the overlay(s) to be updated.

Arguments

outputId: The ID of the plot output (as used in overlayPlotOutput()).
nrect: Number of overlay rectangles to support.
width: Optional default overlay width in plot coordinates. If NULL (default), set to 10% of the plot width.
snap: Function to "snap" overlay coordinates to a grid, or "none" (default) for no snapping. See details for how to specify the snap function.
colours: A function to assign custom colours to the overlays. Should be a function that takes a single integer (the number of overlays) and returns colours in hexadecimal notation (e.g. "#FF0000"). Do not provide opacity here as a fourth channel; use the opacity argument instead.
opacity: Numeric value (0 to 1) indicating overlay transparency.
icon: A Shiny icon to show the dropdown menu.
stagger: Vertical offset between stacked overlays, as a proportion of height.
style: Named list of character vectors with additional CSS styling attributes for the overlays. If an element is named "background-color" then this will override the colours and opacity arguments. Vectors are recycled to length nrect.
debug: If TRUE, prints changes to input values to the console for debugging purposes.

Details

Call this function once from your server code to initialise a set of overlay rectangles for a specific plot. It creates reactive handlers for move, resize, and dropdown menu actions, and allows adding new overlays by dragging an overlayToken() onto the plot. The function returns a shiny::reactiveValues() object which you should keep for further use; in the examples and documentation, this object is typically called ov.

This function also defines a dynamic output UI slot with ID paste0(outputId, "_menu"), which can be rendered using shiny::renderUI(). When a user clicks the overlay's dropdown icon, this menu becomes visible and can be populated with inputs for editing overlay-specific settings, e.g. labels or numeric parameters tied to that overlay.

If you provide a coordinate snapping function (snap argument), it should have the signature function(ov, i) where ov is the shiny::reactiveValues() object defining the overlays and their settings, and i is the set of indices for the rectangles to be updated. When the position of any of the overlays is changed, the snapping function will be applied. In this function, you should make sure that all ov$cx0[i] and ov$cx1[i] are within the coordinate bounds defined by the plot, i.e. constrained by ov$bound_cx and ov$bound_cw, when the function returns. This means, for example, if you are "rounding down" ov$cx0[i] to some nearest multiple of a number, you should make sure it doesn't become less than ov$bound_cx. Finally, the snapping function will get triggered when the x axis range of the plot changes, so it may be a good idea to provide one if the user might place an overlay onto the plot, but then change the x axis range of the plot such that the overlay is no longer visible. You can detect this by verifying whether the overlay rectangles are "out of bounds" at the top of your snapping function. See example below.

Examples

Run this code

# Example of a valid snapping function: snap to nearest round number and
# make sure the overlay is at least 2 units wide.
mysnap <- function(ov, i) {
    # remove any "out of bounds" overlays
    oob <- seq_len(ov$n) %in% i &
        (ov$cx0 < ov$bound_cx | ov$cx1 > ov$bound_cx + ov$bound_cw)
    ov$active[oob] <- FALSE

    # adjust position and with
    widths <- pmax(2, round(ov$cx1[i] - ov$cx0[i]))
    ov$cx0[i] <- pmax(round(ov$bound_cx),
        pmin(round(ov$bound_cx + ov$bound_cw) - widths, round(ov$cx0[i])))
    ov$cx1[i] <- pmin(round(ov$bound_cx + ov$bound_cw), ov$cx0[i] + widths)
}

ui <- shiny::fluidPage(
    useOverlay(),
    overlayPlotOutput("my_plot", 640, 480),
    overlayToken("add", "Raise")
    # further UI elements here . . .
)

server <- function(input, output) {
    ov <- overlayServer("my_plot", 4, 1, snap = mysnap)

    output$my_plot_menu <- renderUI({
        i <- req(ov$editing)
        textInput("label_input", "Overlay label", value = ov$label[i])
    })

    observeEvent(input$label_input, {
        i <- req(ov$editing)
        ov$label[i] <- input$label_input
    })

    output$my_plot <- shiny::renderPlot({
        df <- data.frame(x = seq(0, 2 * pi, length.out = 200))
        df$y <- sin(df$x) + 0.1 * sum(ov$active * (df$x > ov$cx0 & df$x < ov$cx1))
        plot(df, type = "l")
        overlayBounds(ov, "base")
    })
    # further server code here . . .
}

if (interactive()) {
    shiny::shinyApp(ui, server)
}