This is heavily modified version of what was originally an extremely efficient binary search data structure I'd made a few years back for checking strings against a dictionary, and for counting occurances of strings. It packs strings into arrays of uint64, allowing for much faster comparisons and very low memory overhead.
This modified version is the backend to TokenMonster. It uses a combination of binary search, sorting, hashmaps, lookup tables and bloom filters.
I would write usage instructions but I suspect nobody will find it. In case you really want to use it, you can make an issue and ask for the usage instructions.