Make ppl mutable

[kliem]

We allow for ppl polyhedra to be mutable.

This is a preparation for allowing interactive modification. At the moment not much can be done, but to delay the calculation of Vrepresentation and Hrepresentation.

The goal is that normaliz and polymake use Polyhedron_interactive as well and allow adding generators/inequalities.

Depends on #32157

CC: @jplab @mkoeppe @yuan-zhou

Component: geometry

Keywords: ppl, lazy double description

Author: Jonathan Kliem

Branch/Commit: fce9f98

Reviewer: Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/32158

[mkoeppe]

comment:2

How about "mutable" instead of "interactive"

[mkoeppe]

comment:3

And surely by default, polyhedra should remain immutable, right?

[mkoeppe]

comment:4

IMO, the constructor Polyhedron should get an additional keyword argument immutable=True (unfortunately there's a bit of inconsistency - in the Sage library one finds all three of immutable=, is_mutable=, is_immutable=).
When False, it should select a suitable backend

[mkoeppe]

comment:5

See also #29101 for my proposed "Refined protocol for _element_constructor_ in element classes with mutability"

[kliem]

comment:6

Thanks for the comments. Yes, using immutable is much better. I would prefer to not have another backend though. It's the same with vectors and matrices. One can just make them immutable, if one likes to, but not the other way around.

[mkoeppe]

comment:7

Replying to @kliem:

I would prefer to not have another backend though.

I'm just saying that Polyhedron, if no backend is given and is_mutable=True is passed, should select one of the backends that support mutation. As of this or a follow-up ticket, I guess that's only ppl.

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

8dceb88	Polyhedron_interactive -> Polyhedron_mutable
5124844	add _element_constructore_polyhedron for ppl over ZZ
c82c144	remove copy paste typo
2cfd7a5	Merge branch 'u/gh-kliem/small_improvements_ppl' of git://trac.sagemath.org/sage into u/gh-kliem/lazy_ppl
d4f6fb7	remove redundant lines in test

[sagetrac-git]

Changed commit from 86be3ed to d4f6fb7

[sagetrac-git]

Changed commit from d4f6fb7 to 1ece615

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

1ece615

allow ppl polyhedron to be mutable

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

7548c12

invalidate dependent object on mutable polyhedron

[sagetrac-git]

Changed commit from 1ece615 to 7548c12

[kliem]

comment:13

I fixed faces to be unhashable when taken from a mutable polyhedron (or that actually happens automatically). This means that in particular we cannot create graphs or hasse diagrams for faces.

I also invalidate all objects that point towards the polyhedron, when the polyhedron is changed. I don't know if this is good or if a user is responsible for not keeping objects pointing on polyhedra that might change.

I also changed the behavior to not recycle Vrepresentation and Hrepresentation objects when clearing the cache. This why I at least prevent that an representation object that a user keeps changes at some point.

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

8720572	coverage in base.py
473f729	remove unused imports
fce9f98	check attribute first

[sagetrac-git]

Changed commit from 7548c12 to fce9f98

[mkoeppe]

Reviewer: Matthias Koeppe

[mkoeppe]

comment:15

LGTM.

[kliem]

comment:16

Thank you.

[vbraun]

Changed branch from u/gh-kliem/lazy_ppl to fce9f98