cmudb

This is, roughly, a Rust port of CMU's BusTub (https://github.com/cmu-db/bustub). It's a work in progress.

What's done:

• disk-backed pages for tables
• Adaptive replacement cache policy for pages
• B+ tree indexes with optimistic latch coupling
• An in-memory catalog
• Most executors described in the CMU course
• parser (using sqlparser crate)
• binder
• planner

What's missing:

• optimizer
• mvcc
• support for more types, only a few basic types are supported (int, booleans, floats, timestamps)

Things I would change if I have the time:

• Pages as typed views: Currently pages are fully decoded and encoded as we read from and write to them. A better approach is for pages act as typed views over raw bytes, which would save quite a bit of i/o work.
• Improved execution model: A volcano-style iteration model is used here, and notably there is no batching going on, so rows that the db system materializes are pulled 1 by 1.