Feature req: have clustering work with hierarchical layout

[abcbaby]

Edges becomes much longer and nodes cannot be move up or down until the cluster are open.

http://jsbin.com/cofewo/edit?html,output

[AlexDM0]

Hi,

Clustering does not work with hierarchical layout at the moment.

There are many more issues going on here though, I'll see what I can do.

Regards

[AlexDM0]

Sorry, there is actually nothing else going wrong. All issues are because of using clustering with hierarchical view. This is not supported at this time. I cannot say when this will be supported at the moment.

Regards,

Alex

[abcbaby]

That's fine. I have a workaround for it. That is, whenever I need to change the layout, I'll just clear the whole network, draw the whole thing again, and then recluster all the clusters, which I keep track of. Not the best way, but it works.

BTW, is v4.5.1 be out today?

[AlexDM0]

It should be

[wimrijnders]

Looking into this. OK, I admit it, it's much harder than I thought.

I've had the following hypotheses:

• This is the 'exploding networks' issue hampering clustering for hierarchical networks. Fix exploding networks, refactoring LayoutEngine #3017 solves this for non-clustered networks. This is not the problem of this issue.
• This is the unlawful use of dynamic edges in hierarchical networks, solved by Network: Smooth type 'dynamic' adjusted for node-specific options in hierarchical layout #3130. Also not the problem.

Current hypothesis:
(Part of) the reason that clustering and hierarchical do not mix, is that, for some hierarchical code, all nodes and edges are taken into consideration, while only the visible ones should be. Currently investigating if this can be changed.

---

Update: The final hypothesis was indeed the problem.

[wimrijnders]

I tested the jsbin with the latest stable version of vis.js v4.20.0.

All shown layouts work, with the exception of 'Changing to UD layout', which actually does work, but:

• shows some funky, 'too long' edges
• with the default sorting hubsize, the order is totally not what you would expect.

Visualization:

See previous comment for the current work motivation.