`bundle_strength` and `mode` arguments in `edge_bundle_path`

[noriakis]

Thank you very much for developing this useful library. I have made following changes to the edge_bundle_path function and thought if it could be merged for the general use.

• bundle_strength argument Implemenation of the original JS function (subdivide) that controls the strength of bundling.

• mode argument The argument to be passed to igraph::shortest_paths function. Currently the mode parameter will be fixed to out and I thought it would be useful to pass this parameter arbitrarily, such as all.

I would be grateful if you could review this pull request.

[thomasp85]

An alternative approach to bundle_strength which I have used when I ported edge path bundling over to ggraph is to use the same tension approach used in hierarchical edge bundling. It works by moving the control points of the b-spline closer to the direct line (original edge)

[noriakis]

Thank you very much for the insights! I am glad to hear that the edge bundling is now available in ggraph. I don't know if the pull request here should be modified to use the alternative approach, but I'm ready to implement it.

[thomasp85]

I'll leave that up to David - just sharing an alternative approach to loosening the bundles

[thomasp85]

Come to think of it, the relaxation I have might be the inverse of what your parameter adds, right. Your subdivision will make the edges run closer to the shortest path, not farther from it

[noriakis]

Thank you for the reply. Yes, the bundle_strength parameter will make the edge closer to the shortest paths, and takes long time if the strength is high.

library(igraph)
#> 
#> Attaching package: 'igraph'
#> The following objects are masked from 'package:stats':
#> 
#>     decompose, spectrum
#> The following object is masked from 'package:base':
#> 
#>     union
library(ggraph)
#> Loading required package: ggplot2
library(patchwork)
library(edgebundle)
# simple edge-path bundling example from README
g <- graph_from_edgelist(matrix(c(1, 2, 1, 6, 1, 4, 2, 3, 3, 4, 4, 5, 5, 6), 
                                ncol = 2, byrow = TRUE), FALSE)
xy <- cbind(c(0, 10, 25, 40, 50, 50), c(0, 15, 25, 15, 0, -10))
res <- edge_bundle_path(g, xy, bundle_strength=1)
res5 <- edge_bundle_path(g, xy, bundle_strength=5)

bs1 <- ggplot() +
  geom_path(data = res, aes(x, y, group = group)) +
  geom_point(data = as.data.frame(xy), aes(V1, V2), size = 3) +
  theme_void() + ggtitle("bundle_strength = 1")

bs5 <- ggplot() +
  geom_path(data = res5, aes(x, y, group = group)) +
  geom_point(data = as.data.frame(xy), aes(V1, V2), size = 3) +
  theme_void() + ggtitle("bundle_strength = 5")

t05 <- ggraph(g, layout="manual", x=xy[,1], y=xy[,2]) +
  geom_edge_bundle_path(tension=0.5) +
  geom_point(data = as.data.frame(xy), aes(V1, V2), size = 3) +
  theme_void() + ggtitle("tension = 0.5")

t1 <- ggraph(g, layout="manual", x=xy[,1], y=xy[,2]) +
  geom_edge_bundle_path(tension=1) +
  geom_point(data = as.data.frame(xy), aes(V1, V2), size = 3) +
  theme_void() + ggtitle("tension = 1")

wrap_plots(list(bs1, bs5, t05, t1))

^{Created on 2024-01-17 with reprex v2.1.0}

[schochastics]

@noriakis I will leave it as is, I think. If you want to change it though, go ahead, otherwise I will merge this in the next few days. I think in the long run, I will not put too much work in this package anymore, now that the most important algorithms are migrating to ggraph.

[noriakis]

@schochastics Thank you for your reply, and I apologize for the delayed response. I would be grateful if you could proceed with merging in its current form. I will try to implement the feature suggested by @thomasp85 anyway. Thank you again for developing this important package.

[schochastics]

Ok will merge this soon. Thanks again for implementing it.

[noriakis]

Thank you very much for merging!