That was nearly exactly my initial approach! :) However I realized actual diffing was really slow, especially for large collections, as it involves scanning th

That was nearly exactly my initial approach! :) However I realized actual diffing was really slow, especially for large collections, as it involves scanning the entire structure for changes, when in fact all changes could simply be flagged upon mutation. Conceptually, it performs diffing (ie. it gives you a diff object), its just much, much faster!