GH-101100: Fix reference warnings for __getitem__

[AA-Turner]

---

📚 Documentation preview 📚: https://cpython-previews--110118.org.readthedocs.build/

• Issue: Fix all Sphinx reference warnings in the documentation #101100

[serhiy-storchaka]

Previously, there was no need to add the object. prefix. All object methods were automatically found. I don't know why it broke and from where. If someone could fix this, it would save us from rewriting a lot of references.

[AA-Turner]

Do you have a rough idea of when it was working? (2010? 2015? 2020?) I could look at this, but unbounded I fear it'd be a fairly time consuming activity.

Prepending references with ~object. is a fairly easy mechanical task, though I appreciate it adds noise. From what I understand of Sphinx though, it does seem the 'proper' thing to do.

Perhaps of interest, this is a table of the frequency of dunder attributes that cause reference warnings (I've filtered a little but not properly):

Count	Type	Target
32	meth	__exit__
31	meth	__getitem__
26	meth	__enter__
21	meth	__get__
21	meth	__init__
20	meth	__iter__
20	meth	__getattr__
20	meth	__new__
16	attr	__doc__
15	meth	__getattribute__
13	meth	__repr__
12	attr	__cause__
12	meth	__set__
12	meth	__contains__
11	attr	__annotations__
11	meth	__hash__
10	attr	__module__
10	attr	__context__
10	meth	__eq__
10	meth	__len__
10	attr	__self__
8	meth	__reduce__
7	meth	__str__
7	meth	__add__
7	attr	__func__
6	attr	__traceback__
6	meth	__delete__
6	meth	__reversed__
6	meth	__dir__
6	meth	__lt__
6	meth	__await__
6	meth	__setstate__
6	meth	__missing__
6	meth	__setitem__
6	meth	__index__
5	meth	__set_name__
5	meth	__radd__
5	meth	__getstate__
5	meth	__cmp__
4	meth	__call__
4	meth	__le__
4	meth	__ge__
4	meth	__format__
4	meth	__delitem__
4	meth	__int__
3	attr	__suppress_context__
3	attr	__members__
3	meth	__getnewargs__
3	attr	__wrapped__
3	meth	__gt__
3	meth	__setattr__
3	meth	__or__
3	attr	__code__
3	meth	__ne__
3	meth	__truediv__
3	meth	__unicode__
2	meth	__reduce_ex__
2	attr	__isabstractmethod__
2	meth	__and__
2	attr	__notes__
2	func	__getattr__
2	attr	__defaults__
2	attr	__globals__
2	attr	__closure__
2	attr	__type_params__
2	meth	__floordiv__
2	meth	__iadd__
2	attr	__methods__
2	meth	__coerce__
2	meth	__complex__
2	meth	__hex__
2	meth	__oct__
2	meth	__matmul__
1	meth	__next__
1	meth	__subclasscheck__
1	meth	__aenter__
1	meth	__aexit__
1	meth	__abs__
1	func	__getattribute__
1	meth	__getnewargs_ex__
1	meth	__bytes__
1	meth	__fspath__
1	attr	__kwdefaults__
1	meth	__pow__
1	meth	__neg__
1	meth	__pos__
1	meth	__invert__
1	meth	__mul__
1	meth	__rmul__
1	meth	__mod__
1	meth	__sub__
1	meth	__lshift__
1	meth	__rshift__
1	meth	__rand__
1	meth	__xor__
1	meth	__rxor__
1	meth	__ror__
1	meth	__isub__
1	attr	__metaclass__
1	attr	__class__
1	meth	__sizeof__
1	data	__metaclass__
1	meth	__getslice__
1	meth	__setslice__
1	meth	__delslice__
1	func	__repr__
1	func	__str__
1	meth	__del__
1	meth	__rmatmul__
1	meth	__imatmul__
1	meth	__rmod__
1	meth	__divmod__
1	meth	__aiter__
1	attr	__all__
1	meth	__class_getitem__

A

[serhiy-storchaka]

Perhaps it worked 10 years ago. There were more Python specific code in Doc/tools/extensions, and Sphinx itself was more Python-oriented. Or maybe it was just a legend, I don't remember.

[serhiy-storchaka]

It perhaps worked by accident on old Sphinx versions.

[serhiy-storchaka]

LGTM.

@AA-Turner, are you going to create a separate PR for every dunder method? Some lines contain references to several dunder methods (for example __getitem__, __iter__, __len__, __setitem__). The same line can be modified several times in a series of PRs. It will make reading history more difficult. Why not merge changes for several related methods in one PR?

[miss-islington-app]

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

[bedevere-app]

GH-111073 is a backport of this pull request to the 3.12 branch.

[bedevere-app]

GH-111074 is a backport of this pull request to the 3.11 branch.