I think in the case of yesterday’s puzzle, a shortest path can be computed w/o visiting every node.

Hmm…. maybe not. Looks like my solution visited 249,999 out of 250,000 :)

It seems like it should be possible though because of the specifics of the puzzle - we only need paths from top left of grid to lower right, and we know the minimum path is <= (499 + 499) * 9, so it makes no sense to go up, or left, if our path length is already too long.

Okay, it does depend on the graph. If the graph has a particular and known structure then it’s possible to avoid the overhead.

Moreover, if it’s a grid then plain dynamic programming may be much faster than Dijkstra’s algorithm.

Hmm I take it back. I don’t know if the grid is directd or undirected.

Rhombus meeting today! We’re talking about equality operations, how they’ll relate to collections, and fancy-app.

Agenda: https://github.com/racket/rhombus-prototype/discussions/180#discussioncomment-1814253 Zoom link: https://utah.zoom.us/j/96590513005