Start a graph (network) plot

vgraph() begins a node-link diagram from an igraph graph. It computes a layout (vertex x/y), builds a node table and an edge table, and returns a PlotSpec with the aspect locked (coord_equal) and a void theme (no axes or gridlines) – publication-quality defaults with no tuning. Add layers with mark_edges(), mark_nodes(), and mark_node_text(); edges default to the edge table and nodes to the node table.

Usage

vgraph(
  g,
  layout = "stress",
  ...,
  augment = FALSE,
  filter_nodes = NULL,
  filter_edges = NULL,
  k_core = NULL,
  seed = 42L,
  width = 6,
  height = 4,
  dpi = 96
)

Arguments

g

An igraph graph (or a two-column edge-list matrix).

layout

A layout: a name ("stress" (default), "sparse_stress", "backbone", "fr", "kk", "circle", "tree", "sugiyama", ...), a supplied N x 2 coordinate matrix, or a function f(g, ...) returning one.

...

Extra arguments forwarded to the layout function (e.g. pivots = for sparse stress).

augment

Opt-in vertex metrics to attach as graph attributes (so they can be mapped – mark_nodes(size = degree) – or used by filter_nodes). FALSE (default) attaches nothing; TRUE attaches degree and components; a character vector picks from "degree", "in_degree", "out_degree", "betweenness", "closeness", "eigen", "coreness", "components", "community". Never computed silently – vellum does not guess which centrality you meant.

filter_nodes, filter_edges

Data-masked predicates that keep the vertices / edges for which they are TRUE, evaluated against the vertex / edge attribute table – e.g. filter_edges = weight > 0.5, filter_nodes = degree >= 2 (with augment). The graph is reduced before the layout, so the picture is of the subgraph, not a full layout with holes.

k_core

Keep only the k-core: vertices with coreness >= k_core (and their induced edges). NULL (default) keeps everything.

seed

Integer seed for stochastic layouts ("fr", "drl"), making the figure reproducible. The global RNG stream is restored afterwards.

width, height

Page size in inches.

dpi

Output resolution in dots per inch (see vplot()).

Value

A PlotSpec whose data is the node table and whose edge_data is the edge table.

Details

The default layout is stress majorization (graphlayouts::layout_with_stress): it is deterministic (same graph -> same picture) and minimizes the difference between drawn and graph-theoretic distance. Above 10^4 nodes it falls back to sparse stress automatically. Reciprocal and parallel edges are drawn as parallel straight lines offset off the centre line (not curved); self-loops as small loops.

igraph (and graphlayouts, for its layouts) are optional dependencies (in Suggests); vgraph() errors with an install hint if they are not available.

Reductions apply in a fixed order: augment (metrics computed on the input graph) -> filter_edges -> filter_nodes -> k_core. Augmented metrics thus reflect the input graph, not the post-filter subgraph.

See also

mark_edges(), mark_nodes(), mark_node_text()

Examples

if (FALSE) { # \dontrun{
g <- igraph::make_graph("Zachary")
vgraph(g, layout = "stress") |>
  mark_edges() |>
  mark_nodes(size = 3)

# attach degree, then plot only the 2-core, sized by degree
vgraph(g, augment = TRUE, k_core = 2) |>
  mark_edges() |>
  mark_nodes(size = degree)
} # }