machristie
commented
2 years ago So we should just remove this optimization and have the community case study portion of the maptool build a candidate network first and pass to mps generation (like we already do with national cost surface and build tool).
I thought we would need to have the frontend generate the candidate network. We can do that but we don't need to. The mps generation code will already generate the candidate network if it is missing and whatever it generates we can save with the experiment and reuse.
As an optimization, in #72 we reuse the community case study's candidate network if one isn't provided. The idea was that a candidate network for all sources and sinks would include all paths of the candidate network generated for some subset of those sources and sinks. However, this turns out to be demonstrably not true. It's easy for example to pick a source and sink in the Southeast US community case study and generate a candidate network that draws out a path that includes edges that are not in the overall candidate network.
So we should just remove this optimization and have the community case study portion of the maptool build a candidate network first and pass to mps generation (like we already do with national cost surface and build tool).