Update logging LogRecord msg attribute

[aminalaee]

Closes #9913

[AlexWaygood]

It looks like this would be a partial revert of #1776 by @rowillia.

Since there are no mypy-primer hits, it seems unlikely that too many people are passing non-str objects in this specific situation. And the inconsistency with the attribute annotation is indeed incorrect -- they should either both be str or both be object.

[AlexWaygood]

But I'm not sure it makes sense to only do a partial revert of #1776. msg annotations should probably be the same across all of logging/__init__.pyi. For example, logging.Logger.debug currently uses object as the annotation for the msg parameter:

typeshed/stdlib/logging/__init__.pyi

Lines 115 to 117 in dd2818a

	def debug(
	self,
	msg: object,

But logging.Logger.debug calls logging.Logger._log, which passes msg to logging._logRecordFactory, and logging._logRecordFactory is an alias for logging.LogRecord. So if you passed a non-str object to logging.Logger.debug, you'd end up passing a non-str object to the logging.LogRecord constructor. Meaning that the annotation for the msg parameter for logging.Logger.debug (and logging.Logger.{info,warning,warn,error,exception,critical,log,_log}) should be kept in sync with the annotation for the msg parameter of logging.LogRecord.__init__.

[AlexWaygood]

In the issue, you wrote:

According to the Python docs for the LogRecord in logging module the msg argument and property is a string:
https://docs.python.org/3/library/logging.html#logging.LogRecord

However, the CPython docs are pretty inconsistent here: later on down in the docs, it's specifically stated that it's okay to pass non-str objects in as messages:

In the preceding sections and examples, it has been assumed that the message passed when logging the event is a string. However, this is not the only possibility. You can pass an arbitrary object as a message, and its __str__() method will be called when the logging system needs to convert it to a string representation. In fact, if you want to, you can avoid computing a string representation altogether - for example, the SocketHandler emits an event by pickling it and sending it over the wire.

This seems to me like a bug in the CPython docs: they shouldn't say one thing up at the top, only to say a different thing later down.

[aminalaee]

I hadn't really checked the other places, you're right.
This is much deeper in the docs than the CPython and typeshed. I will create another PR on CPython docs.

[AlexWaygood]

Since the CPython docs specifically say that it's okay to pass in arbitrary non-str objects to the msg parameter, I think a better fix would be to keep object for the parameter annotation here, but change this attribute annotation so that it is also object:

typeshed/stdlib/logging/__init__.pyi

Line 392 in dd2818a

msg: str

[AlexWaygood]

Thank you! A docs PR over at CPython would also be great. If you ping me on the PR, I can help review it :)

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉