Replies: 3 comments 1 reply
|
Comment by @seabbs on 2026-07-22: An extreme option would be to rename one of the packages. I don't think DistSpec.jl makes sense as it doens't do the distribution definition done here. For the same reason I am not sure calling the R package composeddistributions makes sense. |
|
Comment by @seabbs on 2026-07-23: This was more of a discussion issue... |
|
I agree there's overlap but my impression is that the actual shared bit (which could benefit from coordination on terminology) is pretty narrow, e.g. here we don't have any of the declaration on sequencing etc. - although of course this could come in handy if/when implementing the planned observation model generalisation, and if going that way adopting the terminology from ComposedDistributions.jl would make a lot of sense, I think. We could definitely change the Beyond that, as you say, most determined by specific language constraints. We could make a cross-reference table? |
|
The current overlap is sequential I think i.e +. I don't think we particularly need to signal alignment either now or in the future. I think perhaps the only action to take is to think about this vaguely and consider if there comes a time to extend to tree like structures here. |
Uh oh!
There was an error while loading. Please reload this page.
Originally opened by @seabbs on 2026-07-22 as issue #139:
I hadn't really thought about this but looking at the package with fresh eyes thare is a lot of overlap in terms of the + etc part of the infrastructure. Design wise the approaches are different but I think that that is forced by the various language restrictions. I am not sure there is anything to do about this apart from note it.
One option would be trying to standarise some nomenclauture (for example composeddistributions uses compose to mean multi - or at least they are similar concepts).
All reactions