Linking myIO with Crosstalk

Crosstalk is a standard for widget-to-widget communication that lets users interact with several visualizations at once — brushing points in one chart dims matching rows in a companion table, filtering a table hides the non-matching points on the chart, and so on. myIO participates in this protocol via setLinked().

Which myIO layers participate

As of v1.2, nine chart-layer types respond to crosstalk selection and filter events:

Layer type	Selection (target)	Filter (target)	Brush source
point	yes	yes	yes (with setBrush())
bar	yes	yes	yes (with setBrush())
groupedBar	yes	yes	yes (with setBrush())
histogram	yes	yes	yes (with setBrush())
hexbin	yes	yes	yes (with setBrush())
waffle	yes	yes	—
beeswarm	yes	yes	yes (with setBrush())
lollipop	yes	yes	—
dumbbell	yes	yes	—

Aggregate layer types (boxplot, violin, qq, regression, density, ridgeline, survfit, comparison) do not participate in v1.2. To drive selection over aggregate views, layer a linked scatter next to the aggregate chart.

Selection vs filter

• Selection dims non-matching elements (sets their CSS opacity to the --chart-brush-dim-opacity variable). Other widgets are not affected by their layout.
• Filter hides non-matching elements entirely (display: none). Use it when linked partners have a filter UI — e.g. a crosstalk::filter_select() dropdown — that genuinely narrows the dataset.

Pass filter = TRUE to setLinked() to enable filter-driven hiding.

Source and target modes

setLinked(mode = ...) accepts three values:

• "source" — this chart emits selections (via its brush) but does not react to incoming ones.
• "target" — this chart reacts to incoming selections but does not emit its own.
• "both" (default) — bidirectional.

Example 1: scatter + DT

library(crosstalk)
library(myIO)
library(DT)

shared <- SharedData$new(mtcars, key = ~rownames(mtcars))

bscols(
  widths = c(7, 5),
  myIO(data = shared$data()) |>
    addIoLayer(
      type = "point", label = "cars",
      mapping = list(x_var = "wt", y_var = "mpg")
    ) |>
    setBrush(direction = "xy") |>
    setLinked(shared),
  datatable(shared, rownames = FALSE, options = list(pageLength = 5))
)

Brush a rectangle in the chart: the matching rows in the table stay bold and the rest dim. Click rows in the table: the matching points stay solid, the rest dim.

Example 2: lollipop + reactable

library(crosstalk)
library(myIO)
library(reactable)

# Aggregate mtcars by cyl for a simple lollipop
by_cyl <- data.frame(
  cyl = c("4", "6", "8"),
  avg_mpg = c(
    mean(mtcars$mpg[mtcars$cyl == 4]),
    mean(mtcars$mpg[mtcars$cyl == 6]),
    mean(mtcars$mpg[mtcars$cyl == 8])
  )
)
shared <- SharedData$new(by_cyl, key = ~cyl)

bscols(
  widths = c(6, 6),
  myIO(data = shared$data()) |>
    addIoLayer(
      type = "lollipop", label = "by-cyl",
      mapping = list(x_var = "cyl", y_var = "avg_mpg")
    ) |>
    setLinked(shared),
  reactable(shared, selection = "multiple", onClick = "select")
)

Example 3: plotly bridge (static listing)

plotly’s own crosstalk integration means you can link a myIO chart directly to a plotly chart over the same SharedData:

library(crosstalk)
library(myIO)
library(plotly)

shared <- SharedData$new(mtcars, key = ~rownames(mtcars))
bscols(
  myIO(data = shared$data()) |>
    addIoLayer(type = "point", label = "m",
               mapping = list(x_var = "wt", y_var = "mpg")) |>
    setBrush(direction = "xy") |>
    setLinked(shared),
  plot_ly(shared, x = ~hp, y = ~mpg, type = "scatter", mode = "markers")
)

Not evaluated in this vignette. Install plotly and run locally.

Example 4: leaflet map (static listing)

library(crosstalk)
library(myIO)
library(leaflet)

quakes_small <- head(quakes, 50)
shared <- SharedData$new(quakes_small)
bscols(
  leaflet(shared) |> addTiles() |> addMarkers(),
  myIO(data = shared$data()) |>
    addIoLayer(type = "point", label = "mag",
               mapping = list(x_var = "depth", y_var = "mag")) |>
    setLinked(shared)
)

Not evaluated in this vignette. Install leaflet and run locally.

Troubleshooting

• My boxplot/violin doesn’t respond. Aggregate layer types are not linkable in v1.2. Drive selection through a sibling scatter or bar layer.
• My chart is in the linked group but brushing doesn’t emit. Add setBrush() to the chart layer. Without a brush, the chart can only be a target, not a source.
• Selections don’t visually change anything on the linked widget. Confirm the SharedData$new() key vector matches the row identity the other widget expects. Mismatched keys silently fail.