Update Sankey algorithm to match latest d3-sankey

[philippjfr]

Michael Bostock has made some major improvements to the overlap avoidance algorithms in d3-sankey over the last week. Since the code in holoviews is a fairly straightforward Python port of that code it should be reasonably straightforward to update our code to match. Here's a link to the diff for all the updates.

[Yura52]

I'd like to work on this. Are there any good examples where the improvements after the update can be observed?

[Yura52]

@philippjfr @jbednar

Hi! I am almost done with the implementation, just want to double check if there is still interest in it.

[philippjfr]

@Yura52 Amazing! We are definitely still interested. Would love to see a PR.

[Yura52]

Sorry for the delay. The implementation is finished. However, I cannot find data where improvements can be observed. Do you have any examples?

The screenshot is based on this example. 1.13.5. on the left, my implementation on the right. The first row shows nodes, the second row shows edges. The "problem" is that the final sankey plot is absolutely the same in both cases and looks like in the gallery example.

So it is either my mistake or "bad" data. The only thing that my implementation changes is how the graph is filled (lines 48-52 here), maybe this is not enough.

[Yura52]

@philippjfr I can create a PR, but I would prefer to make sure that the results I see are expected. Any additional infomation will be helpful.

[philippjfr]

Sorry for not getting back to you before. I'd say open the PR and then I can play around with it and give feedback.

[jnettels]

Hello there, just an occasional user here. I was curious about the progress and saw the images in the PR.

As another user mentioned here #3501 (comment) the biggest issue with holoviews sankey currently seems to be how nodes are aligned towards the center, creating unnecessary overlap of edges and nodes.
I wonder if this PR is a chance to change this behaviour or (maybe instead) expose some parameters to control this behaviour.

Regards, and thanks for your work!

[Yura52]

@jnettels Hi! Have you seen the latest PR update ? What do you think of the (possibly) upcoming changes?

P.S. I would suggest leaving comments in that PR since it contains screenshots and some other information.

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.