Documentation says they use LMDB for the backend. Looking over the documentation, it looks like you could readily use pumpkin be directly to implement the datab

Documentation says they use LMDB for the backend. Looking over the documentation, it looks like you could readily use pumpkin be directly to implement the database/datacaching scheme and interface with it from R. Unless your thesis is on implementation of B-trees, definitely try bootstrapping on something like this first. BTW, lmdb provides memory mapping which can be very fast for computations.