I like vizNetwork in principle but I've been struggling a bit with it.
Another consideration is how this interacts with topological sort and node placement. In general we want arrows generally flowing in the same direction (with the exception of things like waning immunity) and to minimize the crossing of arrows. The optimal node placement literature emphasizes computational complexity issues, which make me nervous. We don't want to spend too much time on this deep problem, but rather find the right point on the user-value-development-effort tradeoff.
Both @Flynn-Primrose and @wzmli have pointed out that this interacts with model structure (see #207, #219), in that users will generally just want to see visualizations of sub-models. This should simplify the task of finding good layouts because sub-models are obviously simpler than full models.
Here's an example with vizNetwork that's not great. https://canmod.github.io/macpan-book/flow-between-states.html#topological-sort
I like vizNetwork in principle but I've been struggling a bit with it.
Another consideration is how this interacts with topological sort and node placement. In general we want arrows generally flowing in the same direction (with the exception of things like waning immunity) and to minimize the crossing of arrows. The optimal node placement literature emphasizes computational complexity issues, which make me nervous. We don't want to spend too much time on this deep problem, but rather find the right point on the user-value-development-effort tradeoff.
Both @Flynn-Primrose and @wzmli have pointed out that this interacts with model structure (see #207, #219), in that users will generally just want to see visualizations of sub-models. This should simplify the task of finding good layouts because sub-models are obviously simpler than full models.