Small improvements for ppl backend

[kliem]

We remove some code duplications in the ppl backend. Along the way we also allow initialization from ppl_polyhedron. This saves almost all the time, when extending base ring to QQ:

Before:

sage: %time P = polytopes.permutahedron(7)                                                                                                                                                                                                                                                                                                                                 
CPU times: user 1.51 s, sys: 19.8 ms, total: 1.53 s
Wall time: 1.52 s
sage: %time P.base_extend(QQ)                                                                                                                                                                                                                                                                                                                                              
CPU times: user 1.17 s, sys: 7.78 ms, total: 1.18 s
Wall time: 1.18 s
A 6-dimensional polyhedron in QQ^7 defined as the convex hull of 5040 vertices

After:

sage: %time P = polytopes.permutahedron(7)                                                                                                                                                                                                                                                                                                                                 
CPU times: user 1.5 s, sys: 23.4 ms, total: 1.52 s
Wall time: 1.52 s
sage: %time P.base_extend(QQ)                                                                                                                                                                                                                                                                                                                                              
CPU times: user 177 ms, sys: 50 µs, total: 177 ms
Wall time: 176 ms
A 6-dimensional polyhedron in QQ^7 defined as the convex hull of 5040 vertices

CC: @jplab @mkoeppe @yuan-zhou

Component: geometry

Author: Jonathan Kliem

Branch/Commit: c82c144

Reviewer: Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/32157

[kliem]

comment:2

This is a preliminary for two things:

• Allowing lazy computation of Hrepresentation and Vrepresentation
• Have interactive backends (add generator/add constraint).

[mkoeppe]

comment:3

Why is the _element_constructor_polyhedron only needed for Polyhedra_QQ_ppl but not Polyhedra_ZZ_ppl?

[sagetrac-git]

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

5124844

add _element_constructore_polyhedron for ppl over ZZ

[sagetrac-git]

Changed commit from 5e81afa to 5124844

[mkoeppe]

comment:5

+
+        See :class:`Polyhedron_normaliz` for a description of the input
+        data.
+

should be ...ppl

[mkoeppe]

comment:6

We forgot to do this in the normaliz and polymake backends, but now that mutability is coming into play I think it's crucial to clarify object ownership: Does the normaliz cone / polymake polytope / ppl polyhedron belong to the created Polyhedron object?

[sagetrac-git]

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

c82c144

remove copy paste typo

[sagetrac-git]

Changed commit from 5124844 to c82c144

[kliem]

comment:8

Replying to @mkoeppe:

We forgot to do this in the normaliz and polymake backends, but now that mutability is coming into play I think it's crucial to clarify object ownership: Does the normaliz cone / polymake polytope / ppl polyhedron belong to the created Polyhedron object?

Yes, I think so. I don't know what exact consequences this has. Do we make a copy, when initializing with a cone etc?

As long as everything is immutable, it's fine to have multiple polyhedra with the same cone. However, once things are mutable, that should be taken care of.

[mkoeppe]

comment:9

Replying to @kliem:

Replying to @mkoeppe:

We forgot to do this in the normaliz and polymake backends, but now that mutability is coming into play I think it's crucial to clarify object ownership: Does the normaliz cone / polymake polytope / ppl polyhedron belong to the created Polyhedron object?

Yes, I think so. I don't know what exact consequences this has. Do we make a copy, when initializing with a cone etc?

If it is clarified that ownership transfers to the new object, then no copy needs to be made.

[kliem]

comment:10

Yes this makes sense. When giving a cone to a normaliz polyhedron, I would usually assume that it may be modified. If the polyhedron is mutable, this mean actual modification and not just computing stuff.

[mkoeppe]

comment:11

Right, and on the other hand it would be an error if the user continues to modify the object that now belongs to the polyhedron.

[mkoeppe]

Reviewer: Matthias Koeppe

[vbraun]

Changed branch from u/gh-kliem/small_improvements_ppl to c82c144