Edges

Link

While you don’t have to use a straight line for edges it is certainly possible and geom_edge_link() is here to serve your needs:

ggraph(hairball, layout = 'stress') + 
  geom_edge_link(aes(colour = year))

There’s really not much more to it — every edge is simply a straight line between the terminal nodes. Moving on…

Fan

Sometimes the graph is not simple, i.e. it has multiple edges between the same nodes. Using links is a bad choice here because edges will overlap and the viewer will be unable to discover parallel edges. geom_edge_fan() got you covered here. If there are no parallel edges it behaves like geom_edge_link() and draws a straight line, but if parallel edges exists it will spread them out as arcs with different curvature. Parallel edges will be sorted by directionality prior to plotting so edges flowing in the same direction will be plotted together:

ggraph(hairball, layout = 'stress') + 
  geom_edge_fan(aes(colour = year))

Parallel

An alternative to geom_edge_fan() is geom_edge_parallel(). It will draw edges as straight lines but in the case of multi-edges it will offset each edge a bit so they run parallel to each other. As with geom_edge_fan() the edges will be sorted by direction first. The offset is done at draw time and will thus remain constant even during resizing:

ggraph(hairball, layout = 'stress') + 
  geom_edge_parallel(aes(colour = year))

Loops

Loops cannot be shown with regular edges as they have no length. A dedicated geom_edge_loop() exists for these cases:

# let's make some of the student love themselves
loopy_hairball <- hairball |> 
  bind_edges(tibble::tibble(from = 1:5, to = 1:5, year = rep('1957', 5)))
ggraph(loopy_hairball, layout = 'stress') + 
  geom_edge_link(aes(colour = year), alpha = 0.25) + 
  geom_edge_loop(aes(colour = year))

The direction, span, and strength of the loop can all be controlled, but in general loops will add a lot of visual clutter to your plot unless the graph is very simple.

Density

This one is definitely strange, and I’m unsure of it’s usefulness, but it is here and it deserves an introduction. Consider the case where it is of interest to see which types of edges dominates certain areas of the graph. You can colour the edges, but edges can tend to get overplotted, thus reducing readability. geom_edge_density() lets you add a shading to your plot based on the density of edges in a certain area:

ggraph(hairball, layout = 'stress') + 
  geom_edge_density(aes(fill = year)) + 
  geom_edge_link(alpha = 0.25)

## Warning: The following aesthetics were dropped during statistical transformation: xend
## and yend.
## ℹ This can happen when ggplot fails to infer the correct grouping structure in
##   the data.
## ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
##   variable into a factor?

Arcs

While some insists that curved edges should be used in standard “hairball” graph visualisations it really is a poor choice, as it increases overplotting and decreases interpretability for virtually no gain (unless complexity is your thing). That doesn’t mean arcs have no use in graph visualizations. Linear and circular layouts can benefit greatly from them and geom_edge_arc() is provided precisely for this scenario:

ggraph(hairball, layout = 'linear') + 
  geom_edge_arc(aes(colour = year))

Arcs behave differently in circular layouts as they will always bend towards the center no matter the direction of the edge (the same thing can be achieved in a linear layout by setting fold = TRUE).

ggraph(hairball, layout = 'linear', circular = TRUE) + 
  geom_edge_arc(aes(colour = year)) + 
  coord_fixed()

Bundling

Edge bundling is a technique to reduce clutter in a network visualization by bundling edges that flows in the same direction. There are various ways of doing this, many with heavy computational cost and the potential to mislead. The technique were initially confined to connections between nodes with a hierarchical structure but has been expanded to general graphs. ggraph provides 3 different bundling geoms with various up- and downsides.

ggraph(hairball) + 
  geom_edge_bundle_force(n_cycle = 2, threshold = 0.4)

## Using "stress" as default layout

ggraph(hairball) + 
  geom_edge_bundle_path()

## Using "stress" as default layout

ggraph(hairball) + 
  geom_edge_bundle_minimal()

## Using "stress" as default layout

Elbow

Aah… The classic dendrogram with its right angle bends. Of course such visualizations are also supported with the geom_edge_elbow(). It goes without saying that this type of edge requires a layout that flows in a defined direction, such as a tree:

ggraph(hierarchy, layout = 'dendrogram', height = height) + 
  geom_edge_elbow()

Diagonals

If right angles aren’t really your thing ggraph provides a smoother version in the form of geom_edge_diagonal(). This edge is a quadratic bezier with control points positioned at the same x-value as the terminal nodes and halfway in-between the nodes on the y-axis. The result is more organic than the elbows:

ggraph(hierarchy, layout = 'dendrogram', height = height) + 
  geom_edge_diagonal()

It tends to look a bit weird with hugely unbalanced trees so use with care…

Bends

An alternative to diagonals are bend edges which are elbow edges with a smoothed corner. It is implemented as a quadratic bezier with control points at the location of the expected elbow corner:

ggraph(hierarchy, layout = 'dendrogram', height = height) + 
  geom_edge_bend()

Hive

This is certainly a very specific type of edge, intended only for use with hive plots. It draws edges as quadratic beziers with control point positioned perpendicular to the axes of the hive layout:

ggraph(hairball, layout = 'hive', axis = pop_devel, sort.by = popularity) + 
  geom_edge_hive(aes(colour = year)) + 
  geom_axis_hive(label = FALSE) + 
  coord_fixed()

Span

As with the hive edge the geom_edge_span() is made in particular for a specific layout - the fabric layout. It draws the edge as a vertical line connecting the horizontal node lines of the layout, potentially with a terminal shape.

ggraph(hairball, layout = 'fabric', sort.by = node_rank_fabric()) + 
  geom_node_range(colour = 'grey') + 
  geom_edge_span(end_shape = 'circle') + 
  coord_fixed()

Point and tile

It may seem weird to have edge geoms that doesn’t have any span, but the matrix layout calls for exactly that. The terminal nodes of the edge are determined by the vertical and horizontal position of the mark, and for that reason the geom doesn’t need any extend. The point and tile geoms serve the same purpose but are simply different geometry types:

ggraph(hairball, layout = 'matrix', sort.by = bfs_rank()) + 
  geom_edge_point() + 
  coord_fixed()

ggraph(hairball, layout = 'matrix', sort.by = bfs_rank()) + 
  geom_edge_tile() + 
  coord_fixed()

Base variant

The basic edge geom is drawn by calculating a number of points along the edge path and draw a line between these. This means that you’re in control of the detail level of curved edges and that all complex calculations happens up front. Generally you will see better performance using the base variant rather than the 0-variant that uses grid grobs, unless you set the number of points to calculate to something huge (50–100 is usually sufficient for a smooth look). Apart from better performance you also get a nice bonus (you actually get several, but only one is discussed here): The possibility of drawing a gradient along the edge. Each calculated point gets an index value between 0 and 1 that specifies how far along the edge it is positioned and this value can be used to e.g. map to an alpha level to show the direction of the edge:

ggraph(hairball, layout = 'linear') + 
  geom_edge_arc(aes(colour = year, alpha = after_stat(index))) + 
  scale_edge_alpha('Edge direction', guide = 'edge_direction')

2-variant

Like the base variant the 2-variant calculates points along the edge and draws a path along them. The difference here is that in this variant you can map node attributes to the edge and the aesthetics are then interpolated along the edge. This is easier to show than to explain:

ggraph(hierarchy, layout = 'dendrogram', height = height) + 
  geom_edge_elbow2(aes(colour = node.Class))

There are considerably more computation going on than in the base variant so unless you need to interpolate values between the terminal nodes you should go with the base variant.

0-variant

This is, sadly, the boring one at the end. You don’t get the luxury of smooth gradients over the edge and often you have a considerably worse performance. What you gain though is tack sharp resolution in the curves so if this is of utmost importance you are covered by this variant.

Arrows

While we saw above that direction can be encoded as a gradient, the good old arrow is still available. As with the standard ggplot2 geoms an arrow can be added using the arrow argument:

ggraph(simple, layout = 'graphopt') + 
  geom_edge_link(arrow = arrow(length = unit(4, 'mm'))) + 
  geom_node_point(size = 5)

I hope you think Ugh at the sight of this. The edges naturally extend to the node center and nodes are thus drawn on top of the arrow heads. There’s a solution to this in the form of the start_cap and end_cap aesthetics in the base and 2-variant edge geoms (sorry 0-variant). This can be used to start and stop the edge drawing at an absolute distance from the terminal nodes. Watch this:

ggraph(simple, layout = 'graphopt') + 
  geom_edge_link(arrow = arrow(length = unit(4, 'mm')), 
                 end_cap = circle(3, 'mm')) + 
  geom_node_point(size = 5)

Using the circle(), square(), ellipsis(), and rectangle() helpers it is possible to get a lot of control over how edges are capped at either end. This works for any edge, curved or not:

ggraph(simple, layout = 'linear', circular = TRUE) + 
  geom_edge_arc(arrow = arrow(length = unit(4, 'mm')), 
                start_cap = circle(3, 'mm'),
                end_cap = circle(3, 'mm')) + 
  geom_node_point(size = 5) + 
  coord_fixed()

When plotting node labels you often want to avoid that incoming and outgoing edges overlaps with the labels. ggraph provides a helper that calculates the bounding rectangle of the labels and cap edges based on that:

ggraph(simple, layout = 'graphopt') + 
  geom_edge_link(aes(start_cap = label_rect(node1.name),
                     end_cap = label_rect(node2.name)), 
                 arrow = arrow(length = unit(4, 'mm'))) + 
  geom_node_text(aes(label = name))

The capping of edges is dynamic and responds to resizing of the plot so the absolute size of the cap areas are maintained at all time.

Labels

You would expect that edge labels would be their own geom(s), but ggraph departs from the stringent grammar interpretation here. This is because the label placement is dependent on the choice of edge. Because of this edge labeling is bundled with each edge geom (but not the 0-variant) through the label aesthetic

ggraph(simple, layout = 'graphopt') + 
  geom_edge_link(aes(label = type), 
                 arrow = arrow(length = unit(4, 'mm')), 
                 end_cap = circle(3, 'mm')) + 
  geom_node_point(size = 5)

Usually you would like the labels to run along the edges, but providing a fixed angle will only work at a very specific aspect ratio. Instead ggraph offers to calculate the correct angle dynamically so the labels always runs along the edge. Furthermore it can offset the label by an absolute length:

ggraph(simple, layout = 'graphopt') + 
  geom_edge_link(aes(label = type), 
                 angle_calc = 'along',
                 label_dodge = unit(2.5, 'mm'),
                 arrow = arrow(length = unit(4, 'mm')), 
                 end_cap = circle(3, 'mm')) + 
  geom_node_point(size = 5)

ggraph offers a lot of additional customization of the edge labels but this shows the main features. As with arrowheads labels can severely clutter your visualization so it is only advisable on very simple graphs.