Increased efficiency of UnionFind.__getitem__ slightly.

[simonsteinberg]

The last element in path is the root itself.

[boothby]

Are you sure this increases efficiency? A microbenchmark tells me otherwise.

In [1]: def original(parents, path): 
   ...:     for node in path: 
   ...:         parents[node] = None 
   ...: def improved(parents, path): 
   ...:     for node in path[:-1]: 
   ...:         parents[node] = None 
   ...: path = list(range(10)) 
   ...: parents = {} 
   ...: print("original:") 
   ...: %timeit original(parents, path) 
   ...: print("improved:") 
   ...: %timeit improved(parents, path) 
   ...: print()                                                                                          
original:
623 ns ± 18 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
improved:
701 ns ± 31.7 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

It should be a tiny bit faster to call path.pop or avoid appending the root in the first place

[dschult]

Yes, it looks like the faster method is to construct the path without the last node.
Something like:

        path = []
        root = self.parents[object]
        while root != object:
            path.append(object)
            object = root
            root = self.parents[object]
        # Note that root is not added to the path

[rossbar]

Gentle ping @simonsteinberg - do you want to give the above suggestion a try? No worries if you don't have time - we can always convert this to an issue to be picked up later!

[MridulS]

Thanks for opening this @simonsteinberg, I've created a new PR with the suggested change in #5844. I'll go ahead and close this :)