Network Wrangling Verbs

cograph's verbs for reshaping a network. Every verb takes any supported input (matrix, edge list, igraph, statnet network, tna model, cograph_network), takes its options as named arguments, and returns a cograph_network — or the input format when keep_format = TRUE. There is no pipeline state to activate and nothing to unpack afterwards: use as.data.frame() for the tidy edge or node table.

Value

Each verb returns a cograph_network, except split_components(), which returns a list of them. With keep_format = TRUE a matrix, igraph, statnet network or tna input comes back in that format.

Selecting

filter_nodes(), select_nodes(): Keep nodes by expression, name, index, top-N, neighborhood or component.
filter_edges(), select_edges(): Keep edges by expression, endpoints, bridges, mutuality or top-N.
select_neighbors(), select_component(), select_top(), select_k_core(): Named shorthands for the common selections.
split_components(): One network per connected component.

Weights

threshold_edges(): Keep edges by weight, count, proportion or density.
binarize(): Replace weights with 0/1.
symmetrize(): Combine opposite arcs into one edge.
normalize_weights(): Rescale by row, column, maximum, total, or to [0, 1].
invert_weights(): Turn similarities into distances.

Structure

to_undirected(), to_directed(), reverse_edges(): Change directedness.
remove_isolates(): Drop nodes with no edges.
contract_nodes(): Collapse groups of nodes into one.
spanning_tree(), complement_network(): Derived graphs.
reorder_nodes(), rename_nodes(): Change node order or labels without changing the network.
simplify(): Merge duplicate edges and drop loops.

Editing

add_nodes(), remove_nodes(), add_edges(), remove_edges(): Add and remove.
mutate_nodes(), mutate_edges(): Compute and store attributes.
bind_networks(): Union, intersection or difference of two networks.

Conversion and access

as_cograph(), to_matrix(), to_igraph(), to_network(), to_df(), and as.data.frame() on a cograph_network (see as.data.frame.cograph_network).

Semantics worth knowing

Filtering edges does not remove nodes. This matches igraph::delete_edges() and tidygraph. Nodes left without edges raise a cograph_isolates_created warning; call remove_isolates() to drop them, or pass keep_isolates = FALSE.
Undirected results stay undirected. The weight matrix of an undirected result is symmetric, so nothing downstream re-detects it as directed.
Metadata survives. Node groups, estimation data, layout coordinates and the original source type are carried through every verb.
Malformed selections are errors. Unknown node names, out-of- range or fractional indices, unknown measure names and a malformed between raise a cograph_bad_selection error rather than warning and returning something plausible.

ego_networks(), shortest_paths(), disparity_filter(), detect_communities(), summarize_clusters(), aggregate_layers().

Examples

adj <- matrix(c(0, .5, .8, 0,
                .5, 0, .3, .6,
                .8, .3, 0, .4,
                 0, .6, .4, 0), 4, 4, byrow = TRUE)
rownames(adj) <- colnames(adj) <- c("A", "B", "C", "D")

# One call, named arguments, a tidy table out
as.data.frame(threshold_edges(adj, minimum = 0.4))
#>   from to weight
#> 1    A  B    0.5
#> 2    A  C    0.8
#> 3    B  D    0.6
#> 4    C  D    0.4

# Verbs compose
adj |>
  threshold_edges(minimum = 0.4) |>
  remove_isolates() |>
  mutate_nodes(deg = degree) |>
  as.data.frame(what = "nodes")
#>   id label name  x  y deg
#> 1  1     A    A NA NA   2
#> 2  2     B    B NA NA   2
#> 3  3     C    C NA NA   2
#> 4  4     D    D NA NA   2