When adding or retrieving fields, the DocumentStorage class uses an internal hash table. Using pre-computed hashes will allow us to avoid repeated hashing of the same keys for e.g. time-series bucket unpacking.