If the keys you want to store are in random order, this is a very good choice of datastructure for O(log2(n)) lookups including find_min and find_max. However, if your keys are ordered, this is probably not a good datastructure for you (however, sometimes you can do an O(n) shuffle before storing things in a binary tree, to get a random order from ordered).
Want to store duplicates? Check out dupdict_mod.
See also this list of datastructures I've worked on.
You can download it from here.
You can e-mail the author with questions or comments: