I'm interested in databases and in Rust so i thought why not combine it and learn about databases and Rust by writing a SQL database in Rust. There's a working (although very limited and kinda hacked together) database on the v1 branch which is based on the embedded Sled KV DB as a storage engine.

The goals with v2 are:

• Write my own storage engine
• support all CRUD operations (maybe ALTER TABLE too)
• be multi user and transactional (i'm thinking likely MVCC)
• write a proper cost based based optimizer with a simple cost model (likely DPccp/DPhyp)
• make it an actual database server (maybe support postgres wire protocol)
• actually test it too

I will likely implement create table, insert and select first again and leave the rest for later. All the multi user, transactional and server stuff will likely come last.

The main goal is to write a disk based system that can start out with a rather traditional architecture and then be used as a play-/testing-ground for me to implement all kinds of cool new techniques (JIT compilation using cranelift, fancy new buffer managers, ...). Although the system is supposed to be disk-based, we live in a world where RAM in the tens or even hundreds of GBs is ubiquitous. Therefore we can often just assume that stuff fits into main memory. Especially for System- and Metadata we will mostly not bother having the option of writing them to disk other than for persistence. For execution we will for now also assume that we won't have to spool to disk (otherwise we could still rely on OS swapping for this purpose for now).

• Simple CREATE TABLE
• Simple INSERT (INSERT INTO x VALUES (a,b,c))
• Simple select from where including simple predicates with <,>,>=,<=,= and equi-joins in "attribute = attribute" style (currently only "AND" supported)
• Analysis, planning and optimization (DPccp for now, will do DPhyp later)
• Execution through a simple single threaded volcano style interpreted engine printing the results to stdout (for now)
• crashing and burning (no, but it panics) when someone enters a query that results in a cross product