René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Performance of attribute lookup for type objects","articleBody":"\u003c!--\r\n  If you're new to Python and you're not sure whether what you're experiencing is a bug, the CPython issue tracker is not\r\n  the right place to seek help. Consider the following options instead:\r\n\r\n  - reading the Python tutorial: https://docs.python.org/3/tutorial/\r\n  - posting in the \"Users\" category on discuss.python.org: https://discuss.python.org/c/users/7\r\n  - emailing the Python-list mailing list: https://mail.python.org/mailman/listinfo/python-list\r\n  - searching our issue tracker (https://github.com/python/cpython/issues) to see if\r\n    your problem has already been reported\r\n--\u003e\r\n\r\n**Bug report**\r\n\r\nThe performance of attribute lookup for type objects is worse than for other objects. A benchmark\r\n```\r\nimport pyperf\r\nrunner=pyperf.Runner()\r\n\r\nsetup=\"\"\"\r\nclass Class:\r\n    def all(self):\r\n        pass\r\n\r\nx=Class()\r\n\"\"\"\r\n\r\nrunner.timeit('hasattr x.all', \"hasattr(x, 'all')\", setup=setup)\r\nrunner.timeit('hasattr x.__array_ufunc__', \"hasattr(x, '__array_ufunc__')\", setup=setup)\r\nrunner.timeit('hasattr Class.all', \"hasattr(Class, 'all')\", setup=setup)\r\nrunner.timeit('hasattr Class.__array_ufunc__', \"hasattr(Class, '__array_ufunc__')\", setup=setup) # worse performance\r\n```\r\nResults:\r\n```\r\nhasattr x.all: Mean +- std dev: 68.1 ns +- 1.1 ns\r\nhasattr x.__array_ufunc__: Mean +- std dev: 40.4 ns +- 0.3 ns\r\nhasattr Class.all: Mean +- std dev: 38.1 ns +- 0.6 ns\r\nhasattr Class.__array_ufunc__: Mean +- std dev: 255 ns +- 2 ns\r\n```\r\n\r\nThe reason seems to be that the [`type_getattro`](https://github.com/python/cpython/blob/364ed9409269fb321dc4eafdea677c09a4bc0d8d/Objects/typeobject.c#L3902) always executes `PyErr_Format`, wheras for the \"normal\" attribute lookup this is avoided (see [here](https://github.com/python/cpython/blob/364ed9409269fb321dc4eafdea677c09a4bc0d8d/Objects/object.c#L952) and [here](https://github.com/python/cpython/blob/364ed9409269fb321dc4eafdea677c09a4bc0d8d/Objects/object.c#L1348))\r\n\r\nNotes:\r\n* The benchmark is from the python side, but we are working with the C-API\r\n* The performance is important for numpy, see https://github.com/numpy/numpy/pull/21423\r\n* Another location where this is a bottleneck: https://github.com/python/cpython/blob/v3.12.0a2/Lib/dataclasses.py#L1301\r\n\r\n**Your environment**\r\n\r\n\u003c!-- Include as many relevant details as possible about the environment you experienced the bug in --\u003e\r\n\r\n- CPython versions tested on: Python 3.11.0a7+\r\n- Operating system and architecture: Linux Ubuntu\r\n\r\n\u003c!--\r\nYou can freely edit this text. Remove any lines you believe are unnecessary.\r\n--\u003e\r\n\r\n\r\n\u003c!-- gh-linked-prs --\u003e\r\n### Linked PRs\r\n* gh-99977\r\n* gh-99979\n\u003c!-- /gh-linked-prs --\u003e\r\n","author":{"url":"https://github.com/eendebakpt","@type":"Person","name":"eendebakpt"},"datePublished":"2022-05-03T06:44:21.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":5},"url":"https://github.com/92216/cpython/issues/92216"}