py3: hash collisions of Laurent polynomials

[mwageringel]

This ticket adjusts the hash of multivariate Laurent polynomials, so that it agrees with the hash of univariate Laurent polynomials.

This solves the following problem: Using Python 3, this doctest in laurent_polynomial.pyx fails about 1 out of 4 times.

            sage: L.<w,z> = LaurentPolynomialRing(QQ)
            sage: len({hash(w^i*z^j) for i in [-2..2] for j in [-2..2]})
            25

Due to hash collisions, the result can be smaller than 25 (such as 23 or 21). This gets even worse when using a larger range of monomials.

Regardless of that, it is desirable that univariate and multivariate Laurent polynomials have the same hash anyway. The univariate hash implementation does not seem to have these collisions, so adopting that implementation solves this problem.

For reference, the univariate and multivariate hashes were implemented in #21272 and #23864.

Component: python3

Author: Markus Wageringel

Branch/Commit: 26eb5fd

Reviewer: Frédéric Chapoton

Issue created by migration from https://trac.sagemath.org/ticket/27914

[mwageringel]

Branch: u/gh-mwageringel/py3_hash_laurent

[mwageringel]

Author: Markus Wageringel

[mwageringel]

New commits:

26eb5fd

Trac #27914. py3: fix hash of multivariate Laurent polynomials

[mwageringel]

Commit: 26eb5fd

[fchapoton]

Reviewer: Frédéric Chapoton

[fchapoton]

comment:2

ok, let it be

[vbraun]

Changed branch from u/gh-mwageringel/py3_hash_laurent to 26eb5fd