I undertook a similar endeavor a while back. My solution[1] rested on the observation that you don’t need to have a B-tree to do a binary search; one need only

I undertook a similar endeavor a while back. My solution[1] rested on the observation that you don’t need to have a B-tree to do a binary search; one need only be able to calculate the correct byte offset of the Nth hash. With some optimizations, this approach produced a 9.9GB file with similarly fast lookups. [1]: https://github.com/tylerchr/pwnedpass/blob/master/README.md#file-format https://github.com/tylerchr/pwnedpass/blob/master/README.md#...