Better materialized views in 1 million LOC: Inside Materialize's PostgresSQL src

Instructions for getting into the building:

When you arrive at 436 Lafayette St, just select "Materialize" on the on the intercom outside and someone will buzz you into the building. Take the elevator to floor six.
*
Full address: 436 Lafayette St floor 6, New York, NY 10003

PostgreSQL's materialized views are a great tool to provide low-latency answers to complex queries – but did you know they could offer serializable consistency if you used cutting-edge streaming engines, many millions of dollars, and countless hours of work from talented systems engineers and academics? Oh, and I guess you could then also join them with arbitrarily complex queries over Kafka topics…

In this talk, we'll take a look at what Materialize does from a few different points of views: how it differs from PostgreSQL, and how that fits into the modern data landscape. We'll then explore the architectural considerations we made to bring PostgreSQL data into an operational data warehouse. Lastly, we'll dive into the implementation details and the first correctness bug any of us have ever uncovered in PG.