Physics layout for rectangular (non 1:1 aspect ratio) canvas

[sneumann]

Hi, I'd like a visualisation of a medium-sized network (~100 nodes) with ~7 disconnected subcomponents, and have a standard visNetwork() with standard visPhysics() and standard visLayout(). The resulting network after a while settles into a square-ish shape, i.e. 1:1 aspect ratio.

For publication I want a wide layout, but setting visNetwork(width = xxx, height = yyy) just changes the canvas. I guess what I am looking for is to extend centralGravity to have different gravity in x and y dimensions. Maybe using non-euclidean distance (mahalanobis ?) for the force calculations.

Did I miss something obvious ? I am also happy to use visIgraphLayout() if there is a way to pass the desired aspect ratio.

Thanks in advance, yours, Steffen

[terenceltl]

hi sneumann have you found a solution to coerce the aspect ratio of the clusters into a non-square? I run into the same spot too.

[sneumann]

Hi, afraid I don't think so. Our visualisation code is in https://github.com/rformassspectrometry/metaRbolomics-book/blob/master/rmd/99-Appendices/9902-Appendices-Appendix_2_metaRbolomics_dependencies.Rmd#L120 and the results in https://rformassspectrometry.github.io/metaRbolomics-book/9902-Appendices-Appendix_2_metaRbolomics_dependencies.html and it is still square. Any suggestions still welcome. Yours, Steffen

[terenceltl]

adding the following line kinda works for me:

visIgraphLayout(type="full", randomSeed = 999)

if you want physics effect:

visIgraphLayout(physics = T, type="full", randomSeed = 999)

and you may have to further tune the gravity, and for my case I have to trial and error for the optimal value:

visPhysics(solver = "barnesHut", barnesHut = list(gravitationalConstant = -2700))

[terenceltl]

#Steffen, so does my suggestions work in your case?

So far I find that it is a pain to control the layout of nodes.

[renanxcortes]

Hi, I'd like a visualisation of a medium-sized network (~100 nodes) with ~7 disconnected subcomponents, and have a standard visNetwork() with standard visPhysics() and standard visLayout(). The resulting network after a while settles into a square-ish shape, i.e. 1:1 aspect ratio.

For publication I want a wide layout, but setting visNetwork(width = xxx, height = yyy) just changes the canvas. I guess what I am looking for is to extend centralGravity to have different gravity in x and y dimensions. Maybe using non-euclidean distance (mahalanobis ?) for the force calculations.

Did I miss something obvious ? I am also happy to use visIgraphLayout() if there is a way to pass the desired aspect ratio.

Thanks in advance, yours, Steffen

Multiple gravity centers, possibly allowing "drag and drop" gravity centers would be absolutely great!

[renanxcortes]

adding the following line kinda works for me:

visIgraphLayout(type="full", randomSeed = 999)

if you want physics effect:

visIgraphLayout(physics = T, type="full", randomSeed = 999)

and you may have to further tune the gravity, and for my case I have to trial and error for the optimal value:

visPhysics(solver = "barnesHut", barnesHut = list(gravitationalConstant = -2700))

I tried this workaround in my use cases, however, it was not enough for the networks I'm generating :(

[bschilder]

In my use case, I think i managed to spread things out horizontally a bit more by supplying a start matrix with x/y coordinates for each node. Hard to tell though since the physics takes over after this:

 start <- data.table::as.data.table(g)[,list(x=ifelse(group=="group1",
                                                         -100,100),
                                                y=0)] |> as.matrix()
 visNetwork::visIgraphLayout(layout = "layout_with_graphopt",
                                    type = "full",
                                    start = start)