Trie minimize seems to produce solutions of the same value during the minimize mode, and I'm pondering whether that means we don't need to try the trie search for every solution we generate. In particular, if we put all the solutions into the trie already, then all the trie_minimize calls should produce the same result, because they start at the same state. So we should be able to run trie_minimize from one solution instead of all of them.
Additionally, I think we ought to be able to repeatedly call the mutate functions on a solution until we stop improving it, instead of serializing into a list of many disjoint single improvements.
Trie minimize seems to produce solutions of the same value during the minimize mode, and I'm pondering whether that means we don't need to try the trie search for every solution we generate. In particular, if we put all the solutions into the trie already, then all the
trie_minimizecalls should produce the same result, because they start at the same state. So we should be able to runtrie_minimizefrom one solution instead of all of them.Additionally, I think we ought to be able to repeatedly call the mutate functions on a solution until we stop improving it, instead of serializing into a list of many disjoint single improvements.