PyCQA / astroid / 2053

91%

main: 92%

push

travis-ci

Bug pylint 4206 (#921)

* Takes into account the fact that inferring subscript when the node is a class may use the __class_getitem__ method of the current class instead of looking for __getitem__ in the metaclass.

* OrderedDict in the collections module inherit from dict which is C coded and thus have no metaclass but starting from python3.9 it supports subscripting thanks to the __class_getitem__ method.

* check_metaclass becomes a static class method because we need it in the class scope.
The brain_typing module does not add a ABCMeta_typing class thus there is no need to test it. Moreover it doesn't add neither a __getitem__ to the metaclass

* The brain_typing module does not add anymore _typing suffixed classes in the mro

* The OrderedDict class inherits from C coded dict class and thus doesn't have a metaclass.

* When trying to inherit from typing.Pattern the REPL says :  TypeError: type 're.Pattern' is not an acceptable base type

* The REPL says that Derived as ABCMeta for metaclass and the mro is Derived => Iterator => Iterable => object

* Adds comments

* Starting with Python39 some collections of the collections.abc module support subscripting thanks to __class_getitem__ method. However the wat it is implemented is not straigthforward and instead of complexifying the way __class_getitem__ is handled inside the getitem method of the ClassDef class, we prefer to hack a bit.

* Thanks to __class_getitem__ method there is no need to hack the metaclass

* SImplifies the inference system for typing objects before python3.9. Before python3.9 the objects of the typing module that are alias of the same objects in the collections.abc module have subscripting possibility thanks to the _GenericAlias metaclass. To mock the subscripting capability we add __class_getitem__ method on those objects.

* check_metaclass_is_abc become global to be shared among different classes

* Create a test class dedicated to the Collections ... (continued)

54 of 54 new or added lines in 3 files covered. (100.0%)

7777 of 8619 relevant lines covered (90.23%)

4.47 hits per line