ObjectBox is an object-oriented DBMS library built for IoT and mobile devices. ObjectBox supports several programming languages (Java, C/C++, Go, Swift, Kotlin, and Python) and runs on multiple platforms (Linux, Windows, Mac/iOS, and Android).
ObjectBox stores objects directly, instead of storing data in relational DBs and accessing them through an ORM layer. It has full ACID compatibility, provides transaction semantics, and uses MVCC for concurrency control.
ObjectBox is presented by greenrobot. In 2011, they first created greenDAO, an Android ORM library that uses SQLite as backend. The team then decided to build a NoSQL object-oriented DBMS from scratch, which is ObjectBox. ObjectBox' first public beta release was on 2017/1/24.
The author of ObjectBox is Markus Junginger.
Foreign keys are maintained by relations in each entity. There are 2 kinds of relations: ToOne and ToMany. By combining the type of the relation and the type of the reverse relation, the user can define one-to-one, one-to-many, and many-to-many relations.
B+ tree based with 3 different options: value, 32 bit hash, 64 bit hash (hashes for string values only)
Joins in ObjectBox is implemented using relations stored in each object. Each object stores reference to the other object(s) in the relation, which is similar to index nested loop join.
A query builder "compiles" query objects by setting up using code pointers that check criteria
Low-level mmap approach taking care of OS in-memory caches and ACID compliant disk storage
The DB itself don't have stored procedures, but it can be implemented using functions in the corresponding programming language when run as an embedded database.
The DB runs on IoT and mobile devices and does not share disk or memory with other devices. However, the ObjectBox supports automatic synchronization of data with the backend server. When the server is unavailable (i.e. network failure), the synchronization will stop, and it is restored seamlessly when the server is available again.