Hi, In the past few hours I committed some simple NGP prototype code that implements a journal of binary records and a simple tree content structure based on such records. I'm combining ideas from the append-only data store concept and a more optimized node serialization format. Eventually I'd like to be able to just map the journal file to memory and implement things like Node.getNodes() backed directly on the serialized node format without having to explicitly parse it into an in-memory Java object structure.
If you're interested, check out the journal implementation in o.a.j.ngp.journal and tree implementation in o.a.j.ngp.tree. Comments and improvements are welcome. Feel free to dive in and change things if you like. BR, Jukka Zitting
