Sankey layout

Layered flow layout for DAG edge tables: nodes become rectangles (xmin/xmax/ymin/ymax, plus xc/yc centers), and each edge becomes a closed bezier ribbon emitted to a third table named ribbons in long-form polygon coordinates (x, y, .ribbon_id plus all original edge attribute columns for fill mapping).

Usage

layout_sankey(
  plot,
  node_width = 0.04,
  padding = 0.02,
  curvature = 0.5,
  n_points = 50,
  max_sweeps = 4L
)

Arguments

plot

A plotit object holding graph data, or a bare plotit_graph. The graph must be acyclic.

node_width

Horizontal width of node rectangles (unit-square fraction).

padding

Vertical gap between nodes in the same layer.

curvature

Ribbon curvature in [0, 1]; 0.5 gives symmetric horizontal-tangent beziers.

n_points

Samples per ribbon boundary curve.

max_sweeps

Barycenter refinement sweeps (deterministic).

Value

A modified plotit object (pipeline form), or a new plotit_graph with a ribbons table when called on raw graph data.

Details

The layout is fully deterministic: layers follow longest-path depths, refined by barycenter sweeps; no seed is required.

Examples

e <- data.frame(
  source = c("A", "A", "B", "B", "C"),
  target = c("B", "C", "C", "D", "D"),
  value  = c(10, 5, 8, 3, 6)
)
g <- as_graph(e) |> layout_sankey()
names(g)
#> [1] "nodes"   "edges"   "ribbons"
head(g$nodes[, c("id", "xmin", "ymin")])
#>   id xmin       ymin
#> 1  A 0.00 0.00000000
#> 2  B 0.32 0.13333333
#> 3  C 0.64 0.06666667
#> 4  D 0.96 0.20000000

as_graph(e) |>
  plotit() |>
  layout_sankey() |>
  mark_polygon(
    data = ~ribbons,
    encode(fill = source, group = .ribbon_id),
    alpha = 0.5
  ) |>
  mark_rect(data = ~nodes, encode(fill = id))