Closed cpaniaguam closed 1 month ago
Can you create a unique ID based on index and floe ID and then do the lookup using the unique ID, but still drop rows by only index when needed?
I like that idea. A function could be created to drop by floe ID in a safer way than by row index.
This problem went away when switching to initial floe pairing method
https://github.com/WilhelmusLab/IceFloeTracker.jl/blob/290d7bdbd691f83f5fecca19b7a2d6ad8719eb0b/src/tracker/tracker-funcs.jl#L443
When floes share several potential matching floes, the collision resolution routine drops rows that later are looked up by row index which throws an out of bounds error when checking goodness of fit for floes having dropped floes in its set of potential matches.