[Idea] loop detect algorithm (scratchpad)

[Marchowes]

• iterate through only non-disqualified saddles/links.
• start at saddle A (root node)
• Find all non-self, non-disqualified saddles via connected summit. This function already exists.
• create either an object, or a hash (hash will be faster, object slower but more robust.) looks like: {'CURRENT SADDLE': {'parent': PREVIOUS, 'next': [NEXT1, NEXT2...]}} this way we can keep track of the path. we will also need to keep track of points we've already visited. so to not create in-transit loops. IF we land on our root node, then we've got a closed circuit. If that is the case, we can backtrace the loop through the hash above. We can collect each hop, and find the lowest -- which can then be marked as a basin saddle.

Instead of leveraging a recursive function we can do a bredth search. used deque() in collections to manage the queue.

might be prudent to set a max depth -- but maybe not.

?runoffs? ?edge saddles?

[Marchowes]

I've got a working beta of this feature which looks very promising.

TODO: -Write Unit Tests --Basically Make up scenarios --Edge? -Test if this can effectively eliminate the need for the redundancy checkers. This should work just fine without them, tho it might be more performant to keep them. -Mark Saddles as "Basin" rather than "toolow" -Mark the other type of Saddle (same summit) as such, and delete entirely. These are useless -Mark alternate basin saddles. -Clean up and Lint code -Write tests for supporting functions (like the linker neighbor ones) -Commit KML changes seperately.

[Marchowes]

The story so far:

Most of this algorithm works, however I'm getting more accurate results when using it in conjunction with redundant linker checks. I am not entirely sure why.