Remove duplicate paths

[espottesmith]

Because of prerequisite solving and the use of split reaction nodes, often, there are multiple chemically equivalent pathways that can be taken from the same reactants to the same products. When performing pathfinding, one will often encounter these duplicate pathways, which are uninteresting and redundant.

Ideally, there would be some procedure to prune the shortest path list to return only unique pathways. This shouldn't require any major changes to how MR.Net conducts pathfinding; rather, this can just be a postprocessing step, where the reactions in each path are compared.

[samblau]

I know we probably want to be clever here, and I think that Daniel is working on a strategy for catching duplicate paths from kMC that would work, but after spending some time digging through paths the past couple of days, there is something pretty obvious that should work - just prune paths with identical costs. For example:

0 3.4620678646635246894
1 3.4620678646635246894
2 3.4736261363211290221
3 3.75510952395117456
4 3.7861278042505688611
5 3.7861278042505688611
6 4.2238517847547954525
7 4.3857509970409720313
8 4.3857509970409720313
9 4.477374250818638558
10 4.477374250818638558

paths 0 and 1 are duplicates, paths 4 and 5 are duplicates, paths 7 and 8 are duplicates, paths 9 and 10 are duplicates.

[danielbarter]

Here is how i solve this: I have a integer index for each reaction. A pathway is then a list of indices which you can convert into a frozen set. frozen sets are hashable, so you can easily detect duplicates and count how many times each pathway happens

I think i like Sam's solution better for path finding though