Good! I got always fascinated by non-boolean indexing (probabilistic instead). I've remember having implemented an OOP indexed version of U.C. Berkeley's Chesh

Good! I got always fascinated by non-boolean indexing (probabilistic instead). I've remember having implemented an OOP indexed version of U.C. Berkeley's Cheshire-II in 2007 with Dolphin Smalltalk. I was using BTrees tho, hence O(log N). The non-boolean part was in the values of the keys which were probability of good match against your search target.