Matrix should not inherit from AlgebraElement

[roed314]

This ticket makes Matrix only inherit from ModuleElement.

Warning: it will cause practically all Cython files to be rebuilt.

See also #15215 (duplicate of this ticket).

CC: @jasongrout @pjbruin

Component: linear algebra

Keywords: AlgebraElement, Matrix

Author: Peter Bruin

Branch/Commit: 05c7c30

Reviewer: Luis Felipe Tabera Alonso

Issue created by migration from https://trac.sagemath.org/ticket/804

[sagetrac-mabshoff]

comment:1

Maybe now is a good time to do it?

Cheers,

Michael

[mwhansen]

comment:2

What are the reasons for the change in organization?

[robertwb]

comment:4

The reason for the change is that not all matrices are Algebra Elements.

[kcrisman]

comment:5

What should it inherit from instead? This is a naive question, but perhaps someone with not much skill but much patience could fix this :)

[robertwb]

comment:6

ModuleElement

[pjbruin]

Branch: u/pbruin/804-Matrix_inheritance

[pjbruin]

comment:11

Given that this ticket has been open for more than seven years, it turned out to be surprisingly straightforward. There is one small simplification: Matrix used to have two identical methods _mul_() and __mul__(), now it only needs __mul__(). On the other hand, new (but very straightforward) __pow__() and __div__() methods were required.

[pjbruin]

Author: Peter Bruin

[pjbruin]

Commit: 0fd3d31

[kcrisman]

comment:12

Wow, nice necromancy! Dumb question - any other translations of tutorials have this bit which would need to be translated?

[pjbruin]

comment:13

Replying to @kcrisman:

any other translations of tutorials have this bit which would need to be translated?

It seems not; grep RingElement src/doc/*/tutorial/* only finds the English and French ones.

[kcrisman]

comment:14

any other translations of tutorials have this bit which would need to be translated?

It seems not; grep RingElement src/doc/*/tutorial/* only finds the English and French ones.

Interesting - apparently that must have been added after the other translations were made.

[videlec]

comment:15

Hi,

At first glance, I thought this ticket would help to build tropical matrices as discussed in #14507... but no, the "base ring" for tropical matrix is a semiring and not a ring (no requirement of an additive inverse). Am I right?

Vincent

[pjbruin]

comment:16

Replying to @videlec:

At first glance, I thought this ticket would help to build tropical matrices as discussed in #14507... but no, the "base ring" for tropical matrix is a semiring and not a ring (no requirement of an additive inverse). Am I right?

I think you are. Everywhere in Sage, matrices and vectors are assumed to have coefficients in some base ring. This would probably be much harder to change than the inheritance issue addressed in this ticket. If enough people want tropical matrices, then it seems we need new classes Matrix_semiring and Vector_semiring, possibly inheriting from some ModuleElement_semiring...

[pjbruin]

comment:18

sage -t --long src/sage/structure/element.pyx  # 2 doctests failed

[sagetrac-git]

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

05c7c30

Trac 804: fix type error

[sagetrac-git]

Changed commit from 0fd3d31 to 05c7c30

[lftabera]

Reviewer: Luis Felipe Tabera Alonso

[lftabera]

comment:21

Looks good to me. I also think that tropical matrices should have their own classes.

My copy did not compile against 6.3.beta3 but was probably more an issue with the beta since it failed before attempting to build the sage library. With 6.3.beta5 it works like a charm.

[vbraun]

Changed branch from u/pbruin/804-Matrix_inheritance to 05c7c30