get back on track with official d3-sankey

[cjyetman]

The official d3-sankey library has finally gotten some love recently, it even gained an official horizontal alignment feature. Should we consider getting back to tracking the official version? Frankly, I have avoided messing with any of the sankey code here because so many minor changes/additions have been made I'm afraid of messing something up, but we would likely lose some current features.

Maybe we/someone can audit what we have and compare it to the official library and make a list of additional features we have here, so we can assess the cost of losing them or reimplementing them?

[christophergandrud]

That sounds like a good idea. I do wonder what the new tools give us over the ones already in there.

[Sigfried]

Was any work ever done on this? Any idea how much work might be involved? I guess if there were a bunch of minor changes, finding and replicating them could be a slog.

But I started looking at issues in hopes of finding clues about how to make a new networkD3 visualization or modify an existing one. And Sankey is the one I was thinking about modifying.

So, maybe this should be a new issue, but, is there any kind of architecture guide to this project? I have to admit, I haven't even tried to mess with HTMLWidgets in the past, so I have no idea how the R code and JavaScript come together.

[cjyetman]

As far as I can tell, I’m the only “active” developer here, and even that’s a stretch because I don’t have much time for it these days. I would like to make a new release to integrate a handful of existing fixes and pull requests, but I’ve been saying that for months.

If you wanted to tackle this issue, you’d definitely need substantial knowledge of htmlwidgets and D3/JavaScript. And one of the maintainers would have to take the time to thoroughly review it considering how much could go wrong. If you submitted a nice, clean, well-documented PR, I could probably find the time to do that.

[timelyportfolio]

I'd be glad to help but unfortunately can't commit the time to perform all the steps. As a point of reference, @fbreitwieser has done some very fine work creating a package focused on hierarchical network visualization hiervis that incorporates a new condensed d3 sankey for hierarchies based on my condensed version described in https://www.jsinr.me/2017/11/13/visualizing-trees--partition---sankey/. So, I guess there is room for a traditional non-hierarchical sankey including the features discussed earlier in this issue.