René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Cache the hash value of fractions","articleBody":"In both 3.10 and 3.11, hashing a small fraction is 20x slower than for a decimal.  For fractions with large denominators, it is slower still.\r\n\r\nWe could sacrifice the space for one slot to cache the hash value:\r\n\r\n    __slots__ = ('_numerator', '_denominator', '_hash')\r\n\r\n    def __hash__(self):\r\n        try:\r\n            return self._hash\r\n        except AttributeError:\r\n            self.hash = ... # current code goes here\r\n        return self._hash     \r\n\r\nAnother approach would be to have a single, size limited cache for the computation:\r\n\r\n```\r\ndef __hash__(self):\r\n\r\n    @lru_cache(maxsize=1 \u003c\u003c 16)\r\n    def current_code(numerator, denominator):\r\n        try:\r\n            dinv = pow(denominator, -1, _PyHASH_MODULUS)\r\n        except ValueError:\r\n            # ValueError means there is no modular inverse.\r\n            hash_ = _PyHASH_INF\r\n        else:\r\n            hash_ = hash(hash(abs(numerator)) * dinv)\r\n        result = hash_ if numerator \u003e= 0 else -hash_\r\n        return -2 if result == -1 else result\r\n\r\n    return current_code(self._numerator, self._denominator)\r\n\r\n```\r\n\r\nExample of code that would benefit:\r\n\r\n```\r\n@lru_cache(maxsize=100_000)\r\ndef pabs(n: Fraction | int) -\u003e Fraction:\r\n    \"P-adic absolute value\"\r\n    ...\r\n\r\nfor x, y in product(pool, repeat=2):            # pool is a set of fractions used as test values\r\n    assert (pabs(x)==0) == (x==0)               # |x|=0 if and only if x=0\r\n    assert pabs(x * y) == pabs(x) * pabs(y)     # |xy| = |x||y|\r\n    assert pabs(x + y) \u003c= pabs(x) + pabs(y)     # |x + y| ≤ |x| + |y| (Triangle Inequality)\r\n    assert pabs(x + y) \u003c= max(pabs(x), pabs(y)) # Strong triangle inequality\r\n    assert pabs(x) == pabs(y) or pabs(x + y) == max(pabs(x), pabs(y)) # Lemma 2.1: If |x| ≠ |y|, |x+y| = max{|x|, |y|}\r\n```\r\n\r\n","author":{"url":"https://github.com/rhettinger","@type":"Person","name":"rhettinger"},"datePublished":"2022-08-31T18:04:04.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":4},"url":"https://github.com/96465/cpython/issues/96465"}