Bloom filters provide a probablistic way to test for membership in a set using a fixed amount of space. This is useful for reducing work when the test definitively rules out the existence of a record.

Connections are stateful components which manage database identity and versioning. A connection is generally backed by a merkledag node store and a ref tracker.

Core database functions.

Common code and utilities for working with MerkleDAG graphs.

The data tree in a table is a B+ tree variant which orders the partitions into a sorted branching structure.

Record keys are immutable byte sequences which uniquely identify a record within a table. Keys are stored in sorted order by comparing them lexicographically.

Partitions contain non-overlapping ranges of the records witin a table. Partition nodes contain metadata about the contained records and links to the tablets where the data for each field family is stored.

Patches are applied on top of tables in order to efficiently store changes while re-using old indexed data.

Core record specs and functions.

Tables are named top-level containers of records. Generally, a single table corresponds to a certain ‘kind’ of data. Tables also contain configuration determining how keys are encoded and records are stored.

Functions for working with tablet data.