Add pure Python dbshell fallback

[Jake Howard]

dbshell attempts to shell out to the client tooling for the relevant database engine (sqlite3, psql etc) for dbshell. If the tooling isn't installed, it can't shell. However, for a more naive shell implementation, it would be simple to pipe commands through to connection.cursor().execute and get at least some of the benefit without needing to install the client command-line tooling.

Whilst this wouldn't have the client-side magic (eg backslash commands in Postgres or . commands in sqlite), nor tab complete, it can still be a lot more useful than nothing. An implementation of this shipped in Python 3.12 for SQLite (​https://github.com/python/cpython/blob/main/Lib/sqlite3/__main__.py), which can easily be adapted and made generic for all database engines.

Doing this has a side benefit of also giving support for a dbshell for in-memory SQLite connections, which are currently misleading through dbshell, as it doesn't reuse the same connection that Django does, meaning any database bootstrapping done during startup (best-practice aside) is lost. An in-process connection reuses the same connnection, and thus allows access to that in-memory database.

[Mariusz Felisiak]

Thanks for this ticket, however I don't see much burnout in installing database clients if users want to use database in a shell. It's also not something that needs to be shipped by Django itself. It sounds like a third-party package is the best way to proceed so that people can try it out and then approaching the DevelopersMailingList to reach a wider audience and see what other think.

which can easily be adapted and made generic for all database engines.

I don't think it is, but you may prove that I'm wrong. The devil is in the details.

Doing this has a side benefit of also giving support for a dbshell for in-memory SQLite connections, which are currently misleading through dbshell, as it doesn't reuse the same connection that Django does, meaning any database bootstrapping done during startup (best-practice aside) is lost. An in-process connection reuses the same connnection, and thus allows access to that in-memory database.

I'm not sure how would you like to achieve this, dbshell is a management command. How would you like to share the same in-memory database as e.g. runserver?

[Jake Howard]

How would you like to share the same in-memory database as e.g. runserver?

Sadly, you couldn't - but at least it'd make testing any of the bootstrapping during startup easier, and quickly checking any data.

It sounds like a third-party package is the best way to proceed

On reflection, I agree!