René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Add typing.override decorator","articleBody":"# Feature or enhancement\r\n\r\nSee [PEP 698](https://peps.python.org/pep-0698/) for details.\r\n\r\nThe `typing.override` decorator should, at runtime, attempt to set the `__override__` attribute on its argument to `True` and then return the argument. If it cannot set the `__override__` flag it should return its argument unchanged.\r\n\r\n# Pitch\r\n\r\nThe purpose of `typing.override` is to inform static type checkers that we expect this method to override some attribute of an ancestor class. By having this decorator in place, a developer can ensure that static type checkers will warn them if a base class method name changes.\r\n\r\nTo quote the PEP consider a code change where we rename a parent class method.\r\n\r\nThe original code looks like this:\r\n```python\r\nclass Parent:\r\n    def foo(self, x: int) -\u003e int:\r\n        return x\r\n\r\nclass Child(Parent):\r\n    def foo(self, x: int) -\u003e int:\r\n        return x + 1\r\n\r\ndef parent_callsite(parent: Parent) -\u003e None:\r\n    parent.foo(1)\r\n\r\ndef child_callsite(child: Child) -\u003e None:\r\n    child.foo(1)\r\n```\r\nAnd after our rename it looks like this:\r\n```python\r\nclass Parent:\r\n    # Rename this method\r\n    def new_foo(self, x: int) -\u003e int:\r\n        return x\r\n\r\nclass Child(Parent):\r\n    # This (unchanged) method used to override `foo` but is unrelated to `new_foo`\r\n    def foo(self, x: int) -\u003e int:\r\n        return x + 1\r\n\r\ndef parent_callsite(parent: Parent) -\u003e None:\r\n    # If we pass a Child instance we’ll now run Parent.new_foo - likely a bug\r\n    parent.new_foo(1)\r\n\r\ndef child_callsite(child: Child) -\u003e None:\r\n    # We probably wanted to invoke new_foo here. Instead, we forked the method\r\n    child.foo(1)\r\n```\r\nIn the code snippet above,  renaming `foo` to `new_foo` in `Parent` invalidated the override `foo` of `Child`. But type checkers have no way of knowing this, because they only see a snapshot of the code.\r\n\r\nIf we mark `Child.foo` as an override, then static type checkers will catch the mistake when we rename only `Parent.foo`:\r\n```py\r\nfrom typing import override\r\n\r\nclass Parent:\r\n    def new_foo(self) -\u003e int:\r\n        return 1\r\n\r\n    def bar(self, x: str) -\u003e str:\r\n        return x\r\n\r\nclass Child(Parent):\r\n    @override\r\n    def foo(self) -\u003e int:   # Type Error: foo does not override an attribute of any ancestor\r\n        return 2\r\n```\r\n\r\n# Previous discussion\r\n\r\n[PEP 698](https://peps.python.org/pep-0698/) has details about the proposal itself.\r\n\r\nDiscussion on this proposal has happened on [typing-sig](https://mail.python.org/archives/list/typing-sig@python.org/thread/V23I4D6DEOFW4BBPWBMYTHZUOMKR7KQE/#MZMBDDKOOH2KUE2IRRMQLOEN6MJ2WHCN) and on [discuss.python.org)[https://discuss.python.org/t/pep-698-a-typing-override-decorator/20839].\r\n\n\n\u003c!-- gh-linked-prs --\u003e\n### Linked PRs\n* gh-101564\n\u003c!-- /gh-linked-prs --\u003e\n","author":{"url":"https://github.com/stroxler","@type":"Person","name":"stroxler"},"datePublished":"2023-02-04T16:09:55.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/101561/cpython/issues/101561"}