bpo-42064: Optimise sqlite3 state access, part 1

[erlend-aasland]

Prepare for module state:

• Add "get state by defining class" and "get state by module def" stubs
• Add AC defining class when needed
• Add state pointer to connection context
• Pass state as argument to utility functions

https://bugs.python.org/issue42064

Automerge-Triggered-By: GH:encukou

[encukou]

Have you thought about caching the state on Connection objects? That way you could get the state from a cursor by following two pointers. (And similarly for other cases where you have a Connection, which is most of pysqlite).
Connections are already quite heavy, so adding one more pointer might be a good trade-off.

pysqlite_get_state_by_cls looks like it'll eventually use _PyType_GetModuleByDef, which isn't as fast as following two pointers.

[erlend-aasland]

Have you thought about caching the state on Connection objects? That way you could get the state from a cursor by following two pointers. (And similarly for other cases where you have a Connection, which is most of pysqlite).
Connections are already quite heavy, so adding one more pointer might be a good trade-off.

Yes, I have though about it, but I don't remember why I didn't follow that idea :) As you say, it's probably a very good trade-off. I'll have a go at it.

pysqlite_get_state_by_cls looks like it'll eventually use _PyType_GetModuleByDef, which isn't as fast as following two pointers.

Both true. No, it was meant to use PyType_GetModuleState.

[erlend-aasland]

PTAL!

[encukou]

Looks good, I just did a trivial merge to resolve a conflict.

[miss-islington]

@erlend-aasland: Status check is done, and it's a success ✅ .

[erlend-aasland]

Looks good, I just did a trivial merge to resolve a conflict.

Great, thanks and thank you for reviewing.