Should we consider using geom_forecast() from the forecast package for vis?

[nickreich]

Something that @elray1 has mentioned several times is the possibility of using a ggplot "geom" in hubvis. Recently @sbfnk mentioned that the forecast package has a geom_forecast() in it. Here is the documentation: https://pkg.robjhyndman.com/forecast/reference/geom_forecast.html

Maybe worth thinking about whether or not we can/should take advantage of this as a part of hubVis? I don't know how readily the tools/formats would translate, but maybe worth looking into.

[elray1]

I suggest moving this discussion over here: hubverse-org/hubVis#22

[aaronger]

(Unsure if discussion moved yet)
My sense is that there is not going to be much support in forecast for the general problem we have of representing time series of predictive distributions. The only facility for this that I can see at a first glance is this: https://github.com/robjhyndman/forecast/blob/1f537ebddf318050f7c5ed362bf92cf618041edc/R/ggplot.R#L2090

The lineribbon stat and geom from package ggdist seem like a closer match to our needs, but I have struggled even for those with the question of weather shoe-horning our out_type_id's into the line_ribbon aesthetics is feasible and worthwhile.

[elray1]

Let's close this discussion and move all related comments to that issue.