🤖 New build scheduled with the buildbot fleet by @rhettinger for commit 4f04b7e 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

🤖 New build scheduled with the buildbot fleet by @rhettinger for commit d3179bc 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

I can confirm the new tests fail without the fix 👍🏻

One thing I'm trying to find is why we needed the AttributeError check for self.__root before. Since this is an double-underscore attribute, intuition tells me the only case when this can happen is unpickling. But when we provide self.__class__ as an "empty object factory" in the result of self.reduce, that will call __init__() first before setting the state, so self.__root won't exist at pickling and unpickling time either. And in fact, I can reproduce that it isn't ever present in those times, for protocols 3 - 5.

So I guess skipping the try:except: is harmless. But I wonder if I'm missing something.

@ambv

One thing I'm trying to find is why we needed the AttributeError check for self.__root before.

The __init__ method can be called more than once for an instance and it must add on to existing state rather than reset it from scratch. Here's an example with both regular dictionaries and OrderedDicts:

>>> d = dict(a=1, b=2)
>>> d.__init__(b=3, c=4)
>>> d
{'a': 1, 'b': 3, 'c': 4}

>>> d = OrderedDict(a=1, b=2)
>>> d.__init__(b=3, c=4)
>>> d
OrderedDict({'a': 1, 'b': 3, 'c': 4})

In contrast, the __new__ method can only be called once because it creates the actual instance. Hence, the try/except is no longer needed — the attribute cannot already exist.

By whom can it be called multiple times? I mean, yes, theoretically you can explicitly grab __init__() from an existing instance and call it again but is this a pattern in actual use anywhere? I haven't seen this before.

(yes, I agree that __new__() can skip the try:except:.)

Thanks @rhettinger for the PR, and @ambv for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11, 3.12.
🐍🍒⛏🤖

GH-108200 is a backport of this pull request to the 3.12 branch.

GH-108201 is a backport of this pull request to the 3.11 branch.

Side note. "gh-" in the title should be in lower case, and the issue number should be separated by a colon from the rest of the text. I think this is the reason why other Raymond's PRs were not automatically linked from the corresponding issue.