This function returns another function that can extract edges from a ggraph_layout object. The functionality of the returned function is decided by the arguments to get_edges. The need for get_edges is mainly to pass to the data argument of the different geom_edge_* functions in order to present them with the right kind of data. In general each geom_edge_* has the default set correctly so there is only need to modify the data argument if parallel edges should be collapsed.

get_edges(format = "short", collapse = "none", ...)

get_sf_edges()

Arguments

format

Either 'short' (the default) or 'long'. See details for a descriptions of the differences

collapse

Either 'none' (the default), 'all' or 'direction'. Specifies whether parallel edges should be merged. See details for more information

...

Additional data that will be cbind'ed together with the returned edge data. Accepts expressions that will be evaluated on the edge data

Value

A data.frame with columns dependent on format as well as the graph type. In addition to the columns discussed in the details section, the data.frame will always contain the columns from, to and circular, the two former giving the indexes of the start and end node and the latter if the layout is circular (needed for correct formatting of some geom_edge_*). The graph dependent information is:

dendrogram

A label column will hold the value of the edgetext attribute. In addition any value stored in the edgePar attribute will be added. Lastly a direction column will hold the relative position between the start and end nodes (needed for correct formatting of geom_edge_elbow()).

igraph

All edge attributes of the original graph object is added as columns to the data.frame

Details

There are two types of return formats possible for the result of the returned function:

short

In this format each edge is described in one line in the format expected for ggplot2::geom_segment(), that is, the start node position is encoded in the x and y column and the end node position is encoded in the xend and yend column. If node parameters are added to the edge the name of the parameters will be prefixed with node1. for the start node and node2. for the end node.

long

In this format each edge consists of two rows with matching edge.id value. The start and end position are both encoded in the x and y column. The relative position of the rows determines which is the start and end node, the first occurring being the start node. If node parameters are added to the edge data the name of the parameters will be prefixed with node..

Node parameters are automatically added so it is possible to format edge aesthetics according to start or end node parameters, or interpolate edge aesthetics between start and end node parameters. Node parameters will be prefixed to avoid name clash with edge parameters. The prefix depends on the format (see above).

If the graph is not simple (it contains at most one edge between each node pair) it can be collapsed so either all edges between two nodes or all edges of the same direction between two nodes are merged. The edge parameters are taken from the first occurring edge, so if some more sophisticated summary is needed it is suggested that the graph be tidied up before plotting with ggraph.

See also

Other extractors: get_con(), get_sf_nodes()