Consider deprecating (or at least downplaying) the notion of a "default driver"

[gordthompson]

Further to comment from @aminalaee on #382 :

#382 (comment)

Consider moving away from the notion of a "default driver" and encourage users to specify the driver that they actually want to use. In #382 it is asyncmy vs. aiomysql, but with the release of psycopg 3.x is it also an issue for PostgreSQL since the new driver also supports async (so asyncpg vs. psycopg).

We are having a similar discussion for SQLAlchemy:

sqlalchemy/sqlalchemy#6960

I would be happy to work on a PR here if that would further the cause.

[aminalaee]

@gordthompson
I think that makes sense.
Now that the drivers are being extended, I think keeping that can be limiting and our changes to that won't be backwards-compatible, so I'd rather drop them sooner.

[aminalaee]

@gordthompson Have you had any progress with this? I can pick that up.

[gordthompson]

Hi @aminalaee - Feel free to run with this if you like. SQLAlchemy 1.4+ supports async directly and my primary goal was to provide async support for the CockroachDB dialect so I took that approach

https://github.com/gordthompson/fastapi-tutorial-cockroachdb-async

There is also a related POC for sqlite+aiosqlite at

https://github.com/gordthompson/fastapi-tutorial-aiosqlite