extend solve_right to all cases; implement solve_left

[williamstein]

A.solve_right only worked for A nonsingular, and there was no solve_left. Now A.solve_right should work for any A and there is a solve_left.

Component: linear algebra

Issue created by migration from https://trac.sagemath.org/ticket/2581

[williamstein]

Attachment: sage-2581.patch.gz

[williamstein]

comment:2

WARNING: def restrict_codomain(self, V) is not done yet (it's just a copy of restrict_domain). I'll create a patch later tonight that correctly implements this.

[williamstein]

this provides the extra restrict_codomain functionality.

[craigcitro]

comment:3

Attachment: sage-2581_part2-restrict_codomain.patch.gz

Looks great. Merge this right away!

[sagetrac-mabshoff]

comment:4

Merged both patches in Sage 2.11.alpha0

[ncalexan]

comment:5

An unexpected problem: solve_right is defined in matrix_integer_dense and the versions don't provide the same functionality. This could happen more places.

Is there any hope for a solution_space_right that handles under-determined systems?

[williamstein]

comment:6

An unexpected problem: solve_right is defined in
matrix_integer_dense and the versions don't provide
the same functionality. This could happen more places.

Good point -- I had meant to address that but fell asleep and forgot.
A patch will be forthcoming.

Is there any hope for a solution_space_right that handles under-determined systems?

Not in this patch. Make a trac ticket and somebody will get to it.

[sagetrac-mabshoff]

comment:7

For the record: This ticket was reopened after sage-2581.patch and sage-2581_part2-restrict_codomain.patch had been merged. So any additional patch should make clear how it should be applied.

Cheers,

Michael

[williamstein]

comment:8

I'm attaching a patch that:

1. addresses Nick's concern about derived classes overloading solve_right.

2. I found that Clement recently introduced a solve right for sparse modn matrices for some reason, and that it had several bugs. I've fixed those as well in this patch.

[williamstein]

Attachment: sage-2581_part3.patch.gz

[ncalexan]

comment:10

I think this looks good and should be applied. I think that means part3 should be applied since the other two already have been.

[sagetrac-mabshoff]

comment:11

Merged sage-2581_part3.patch in Sage 2.11.alpha0.