David Schoch

Yes. The fixed coordinates can be arbitrary and dont need to include both x and y values

Yes, I verfied that :-)

here is a "workaround" that I [came up with](http://blog.schochastics.net/posts/2020-03-14_ggraph-tricks-for-common-problems/) ``` r library(igraph) #> #> Attaching package: 'igraph' #> The following objects are masked from 'package:stats': #> #> decompose, spectrum #>...

Debugging datashader is always tricky for me. Did you use the automatic install shown here: https://github.com/schochastics/edgebundle/blob/main/R/bundle_hammer.R#L62-L68 Or did you install it manually?

what happens when you run `install_bundle_py`?

I'd recommend switching to ggraph for edgebundling. Most of the algorithms implemented here are now ported in a better way to that package

@fcorowe Yes, most functionality was ported to ggraph and is available under `geom_edge_bundle_*` see the release blog post: https://www.data-imaginist.com/posts/2024-02-15-ggraph-2-2-0/

seems doable with the [python code](https://github.com/devkotasabin/SPX-graph-layout)

To understand this correctly: this is about moving `arrow` or `readODS` from Suggests to Import? Maybe the dependencies should be taken into consideration? ``` r rang::resolve("readODS") #> resolved: 1 package(s)....

hmm tough decision, but i think my vote is on arrow given its importance for DS.