DB
Get started
Blog

Engineering notes on how the engine works

How NYXDB keeps your tables exact, your derived tables current and your queries open, written by the people building it. These notes run on the same two tables as the docs, so you can run each statement yourself.

All notes

Every note, newest first

03 notes

Several writers, one row: attribute groups

Two pipelines that update different parts of the same entity should never overwrite each other. Attribute groups give each writer its own columns and its own version order on one row.

5 min read

Reading a table as it was: versioned tables and AS OF COMMIT

A versioned table keeps every accepted version beside the current row and settles which one wins, so you can read any key as it stood at a past commit with one clause.

5 min read

Why derived tables belong in the database

An aggregate maintained by a stream processor is a copy you have to reconcile. A transformation keeps it exact inside the engine, fed by the same publication of committed writes as everything else.

5 min read

Read it, then run it

Install a node, create the two tables the docs build, and run every statement from these notes on your own machine.

Get startedRead the docs