MOTOR-1363 - Pass correct parameter to AsyncIOMotorClientEncryption

[NoahStapp]

No description provided.

[NoahStapp]

The tests appear to be failing now due to not closing clients correctly--do we want to wait for that ticket to be done first, or have this PR also skip those warnings for now?

[blink1073]

have this PR also skip those warnings for now?

I'd say skip them. There is a separate set of failures in synchro though

[ShaneHarvey]

Could you explain these changes?

[NoahStapp]

The type guards added in PYTHON-4590 have issues with the synchro import hacking when we import from pymongo.synchronous . I'll have to undo this change since the top-level pymongo.* modules don't export the private members that Motor makes use of.

[ShaneHarvey]

Ah this is exactly the case I was worried about here:

I'm paranoid this might break apps that Mock pymongo classes but it should be fine. Anything that's mocking should inherit from our base classes anyway (instead of trying to ducktype).

mongodb/mongo-python-driver#1820 (comment)

[NoahStapp]

A very accurate prediction, unfortunately I can't find a solution for this issue in Motor that doesn't rely on also changing PyMongo code.

[blink1073]

an option is to get the class names of the full mro and check for the desired name (in PyMongo)

[NoahStapp]

Something like type(client).__name__ == "MongoClient"?

[blink1073]

But also for subclasses:

>>> class Foo(MongoClient):
...    pass
...
>>> Foo.__mro__
(<class '__main__.Foo'>, <class 'pymongo.synchronous.mongo_client.MongoClient'>, <class 'pymongo.common.BaseObject'>, <class 'typing.Generic'>, <class 'object'>)

[NoahStapp]

We'd still run into the same issue with that, right? Since the import paths differ.

[blink1073]

LGTM!