provenance_join() compiles x once and returns a tidy table with one row
per emitted mark element, tying each drawn primitive to both the source
data rows that produced it (rows, a list column) and its device-pixel
geometry (x0, y0, x1, y1, x, y, w, h) from
vellum::scene_model(). It is the consumer of the provenance table that
plot_provenance() emits — the substrate for click-to-source interactivity,
auditing "which rows made this element?", and linked views.
Arguments
- x
A PlotSpec or composition.
Value
A data frame: id, layer, mark, kind, panel, the pixel bbox
(x0, y0, x1, y1, x, y, w, h), n_rows, and rows (a list
column of integer source-row indices). Empty for a plot with no mark grobs.
Examples
df <- data.frame(wt = mtcars$wt, mpg = mtcars$mpg, cyl = factor(mtcars$cyl))
p <- vplot(df) |> mark_point(x = wt, y = mpg, color = cyl)
pj <- provenance_join(p)
pj[, c("id", "mark", "n_rows")]
#> id mark n_rows
#> 1 layer-1-point-g1 point 7
#> 2 layer-1-point-g2 point 11
#> 3 layer-1-point-g3 point 14