René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Accuracy issues of sum() specialization for floats/complexes","articleBody":"# Bug report\r\n\r\n### Bug description:\r\n\r\nUnfortunately, #121176 was merged with a bug:\r\nhttps://github.com/python/cpython/blob/e9681211b9ad11d1c1f471c43bc57cac46814779/Python/bltinmodule.c#L2749-L2755\r\nL2751 lacks cs_add().  Sorry for that.  Reproducer: ``sum([2j, 1., 10E100, 1., -10E100])`` (should be ``2+2j``).  I'll provide a patch.\r\n\r\nBut maybe cases for integer arguments also should use compensated summation?  E.g.:\r\nhttps://github.com/python/cpython/blob/e9681211b9ad11d1c1f471c43bc57cac46814779/Python/bltinmodule.c#L2689-L2698\r\non L2694 (and use ``PyLong_AsDouble()``).  An example:\r\n```pycon\r\n\u003e\u003e\u003e sum([1.0, 10E100, 1.0, -10E100])\r\n2.0\r\n\u003e\u003e\u003e sum([1.0, 10**100, 1.0, -10**100])  # huh?\r\n0.0\r\n```\r\nI would guess, that integer values in this case are treated as exact and they are allowed to smash floating-point result to garbage.  But...  This looks as a bug for me.  ``fsum()`` also chooses ``2.0``:\r\n```pycon\r\n\u003e\u003e\u003e math.fsum([1.0, 10**100, 1.0, -10**100])\r\n2.0\r\n```\r\n\r\n\r\n### CPython versions tested on:\r\n\r\nCPython main branch\r\n\r\n### Operating systems tested on:\r\n\r\n_No response_\n\n\u003c!-- gh-linked-prs --\u003e\n### Linked PRs\n* gh-122236\n* gh-122406\n\u003c!-- /gh-linked-prs --\u003e\n","author":{"url":"https://github.com/skirpichev","@type":"Person","name":"skirpichev"},"datePublished":"2024-07-24T14:07:07.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":9},"url":"https://github.com/122234/cpython/issues/122234"}