add python 3.10 compatibility

[christian-steinmeyer]

When trying to run our sacred code in python 3.10, if fails with

Exception originated from within Sacred.
Traceback (most recent calls):
  File "/Users/<USER>/.virtualenvs/<PROJECT>/lib/python3.10/site-packages/sacred/utils.py", line 345, in recursive_update
    if isinstance(v, collections.Mapping):
AttributeError: module 'collections' has no attribute 'Mapping'

But it looks like, in master, this issue is already fixed. I'm not sure, this is the only issue. But if it were, perhaps a new minor release could be published?

[thequilo]

I agree, an update on PyPi would be a good idea. Many issues have been fixed since the last version (which is already over a year old). But before publishing a new version, I would like to have the tests fixed. I'll look into that when I have some time

[christian-steinmeyer]

Thanks for the response @thequilo. Quick question: which tests are you talking about? The last two azure pipelines seem to only have failed because they were cancelled, if I didn't misread some info there. Of course they might have been cancelled because of some upcoming errors?

[thequilo]

I'm talking about the hanging tests of the QueuedMongoObserver. The QueueObserver, introduced in #471, contains a while True loop which just just hangs the whole script forever when the called method continues failing. In the tests, this is currently (after mongomock's latest update) only caused by a small change in the hash computation in pymongo. This issue is also visible in the failing test_mongo_observer_resource_event test case.

This test shouldn't be too hard to fix, but I am unsure whether we should keep the while True loop in the QueueObserver or replace it with for ... in range(some large number) or a timeout-based system. But this may be a different issue.

[thequilo]

Found a new issue with Python 3.10 and mongomock: mongomock/mongomock#758. So, we have to wait until this issue is fixed in mongomock before we can fix the testcases for Python 3.10

[christian-steinmeyer]

I see, thanks for the updates!

[lucasdavid]

@thequilo no news on mongomock' release date, right?

Considering they release quite seldom, and that this is only a develop dependency used to run the pipelines, don't you think it would be a good idea to refer to their main github version when running Sacred's tests?

This way we can have a sacred release without waiting for them :)

[thequilo]

@lucasdavid right, I didn't even get an answer to the question when the next release will be published.

I actually already started preparing a release by simply deactivating the pymongo tests for Python 3.10 in #858. I didn't want to depend on a git commit of another repository, so I thought that deactivating the tests until a new release is there is the better option. But I'm really unsure if this is really better than the GitHub dependency.

I believe that there are currently no urgent PRs or issues for sacred and a new release has been requested frequently, so I'll try to prepare the release as soon as I can.

[thequilo]

I just published the 0.8.3 release on GitHub (and I hope that I didn't forget anything). @Qwlouse, can you push the release to PyPI?

[ahallermed]

Short reminder for @Qwlouse . Can you push the new version to PyPI? I am very keen on using the new version.
Thank in advance!

[vnmabus]

Maybe it would be good if @Qwlouse could create a PyPI token and add it as a repository secret. This way a Github action could be used to push to PyPI on release, without depending on the availability of one person.