Upgrade pplpy to 0.8.9

[videlec]

Upgrade the standard package pplpy to version 0.8.9 (released on PyPI https://pypi.org/project/pplpy/0.8.9/)

📝 Checklist

• The title is concise, informative, and self-explanatory.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation accordingly.

[mkoeppe]

Lots of test failures

[videlec]

Lots of test failures

Indeed. I do not know what happens. It looks like that the Cython compilation happened with language_level=2.

[antonio-rojas]

Many of the failures have nothing to do with cython, it's an actual bug in pplpy introduced in sagemath/pplpy@5fa3580

More precisely, in the Linear_Expression constructor with two arguments, there is a bool check for the first argument. If this evaluates to false (like it happens, for instance, with a=Cone([(0,0)])._lattice(0), then space_dimension is never set, which is what is causing most of the test failures.

[videlec]

Many of the failures have nothing to do with cython, it's an actual bug in pplpy introduced in sagemath/pplpy@5fa3580

More precisely, in the Linear_Expression constructor with two arguments, there is a bool check for the first argument. If this evaluates to false (like it happens, for instance, with a=Cone([(0,0)])._lattice(0), then space_dimension is never set, which is what is causing most of the test failures.

Indeed, bool([0, 0]) and bool(vector([0, 0])) behave differently! Thanks.

[antonio-rojas]

The failures in sage/geometry/lattice_polytope.py, sage/geometry/polyhedron/backend_ppl.py, sage/geometry/polyhedron/ppl_lattice_polytope.py and the first failure in sage/geometry/cone.py are caused by sagemath/pplpy@3ed9df4, but not by the change in Cython language level, but rather by the __mul__ change in the Variable class

[videlec]

The failures in sage/geometry/lattice_polytope.py, sage/geometry/polyhedron/backend_ppl.py, sage/geometry/polyhedron/ppl_lattice_polytope.py and the first failure in sage/geometry/cone.py are caused by sagemath/pplpy@3ed9df4, but not by the change in Cython language level, but rather by the __mul__ change in the Variable class

Hmm... this should have been catched by the pplpy doctest (https://github.com/sagemath/pplpy/blob/master/ppl/linear_algebra.pyx#L1192)

>>> from ppl import Variable
>>> from gmpy2 import mpz
>>> x = Variable(0)
>>> mpz(3) * x * mpz(5)

[videlec]

I indeed misinterpreted https://cython.readthedocs.io/en/latest/src/userguide/special_methods.html#arithmetic-methods (which affects __add__ but not __mul__).

[videlec]

I indeed misinterpreted https://cython.readthedocs.io/en/latest/src/userguide/special_methods.html#arithmetic-methods (which affects __add__ but not __mul__).

This was correctly interpreted, but works only with Cython>=3.0.0 which is not the default version in sage. This explains why doctests pass in conda.

[videlec]

Hopefully fixed in sagemath/pplpy#27

[antonio-rojas]

Can confirm that it fixes all issues in the downstream Arch sagemath package

[github-actions]

Documentation preview for this PR (built with commit 525dedd; changes) is ready! 🎉