Inspect the compiled-scene provenance of a plot

Compiles x and returns its provenance table: one record per emitted mark grob, tying each low-level primitive back to the data rows and trained scales that produced it. Each record's id matches the grob's data-vellum-id in the SVG output (and the id column of vellum::scene_model()), so it is a stable join key between the rendered scene and the grammar — the substrate for interactivity, linked views, and accessibility.

Usage

plot_provenance(x)

Arguments

x

A PlotSpec or plot composition.

Value

A list of provenance records (see Details). Empty for a plot that emits no mark grobs.

Details

Each record is a plain, serializable list:

id

stable node id, equal to grob@id / SVG data-vellum-id (join key).

layer

1-based layer index within the (sub-)plot spec.

mark

mark type ("point", "line", …).

kind

"mark" (the core layer) or "effect" (a decorative underlay).

panel

facet panel key ("panel-r-c"), or NA for a single panel.

channels

aesthetic → trained-scale reference (scale, type, domain).

rows

the original input-data row indices this grob draws.

The rows field is refined per element for marks that map rows one-to-one to drawn elements (points, bars, tiles, segments, lines, areas, ribbons, steps, text, boxplots, sf features, edges); for aggregating marks (histogram, density, smooth, datashade, …) it is the whole layer, since rows no longer map to single elements.

See also

vellum::scene_model() and the scene-contract vignette in vellum.

Examples

df <- data.frame(wt = mtcars$wt, mpg = mtcars$mpg, model = rownames(mtcars))
p <- vplot(df) |> mark_point(x = wt, y = mpg, data_id = model)
prov <- plot_provenance(p)
prov[[1]]$id
#> [1] "layer-1-point-g1"