Improve matrix_modn_sparse methods for converting to dense and taking submatrices

[tornaria]

Here I'll post some improvements that are needed for #10733 but are of independent interest.

Depends on #17837

Component: linear algebra

Author: Gonzalo Tornaria

Branch/Commit: public/matrix/improve_modn_sparse_methods-10734 @ d3f8b5f

Reviewer: William Stein, Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/10734

[williamstein]

comment:2

This ticket introduces a bunch of new functions, but doesn't include any documentation or doctests for them.

[tornaria]

comment:3

The new patch adds documentation to all the functions created in the first patch.

[williamstein]

comment:4

Great!

[jdemeyer]

Reviewer: William Stein

[jdemeyer]

comment:6

Patches should be created using hg export tip and not hg diff or any other method. Also make sure you have a file $HOME/.hgrc as explained in the developer's manual.

[tornaria]

comment:7

Replying to @jdemeyer:

Patches should be created using hg export tip and not hg diff or any other method. Also make sure you have a file $HOME/.hgrc as explained in the developer's manual.

Is it just this

# User Gonzalo Tornaria <[email protected]>

what the patches are missing?

This is a misfeature of mercurial (queues), because it doesn't set the user by default in the patches. Very silly.

Do you need me to reupload all the patches again, or can you just use

-u "Gonzalo Tornaria <[email protected]>"

as an option when you run hg import? Reuploading 7 files to trac gets boring (is there an easier way to a batch upload to trac? from the command line?).

[as a general comment; I'd prefer if the requirements are stated in terms of what the patches need to include, rather than dictating how one must use mercurial --
I find mercurial inconvenient to use without using queues, because the history cannot be rewritten and I don't know how to effectively use branches -- I've been spoiled by git, I guess]

[tornaria]

Attachment: trac_10734.dense_matrix_and_submatrix.gz

For matrix_modn_sparse: implement optimized versions of submatrix(), dense_matrix(), add methods dense_submatrix() and set_block_unsafe()

[tornaria]

Attachment: trac_10734.02-documentation.gz

Add documentation for the functions added in #10734

[tornaria]

refactor code to support p=2 case (part 1)

[tornaria]

comment:8

Attachment: trac_10734.03-referee_refactor_to_support_p2.patch.gz

I've reuploaded the two patches (hopefully) with the right format (username and date).

Also, I moved part of the referee patch in #10733 here, otherwise some doctests would fail until that ticket is merged.

There's no new code at all, so I assume the positive review is still valid.

[tornaria]

comment:9

apply trac_10734.dense_matrix_and_submatrix, trac_10734.02-documentation, trac_10734.03-referee_refactor_to_support_p2.patch

[fchapoton]

comment:10

Maybe you should add .patch at the end of the names of your patches. Otherwise they are not recognised as such, it seems.

[loefflerd]

Apply only this patch. Patch against 5.0.beta7

[loefflerd]

comment:11

Attachment: trac_10734-dense_matrix_and_submatrix-folded.patch.gz

I qfolded Gonzalo's patches into a single patch, and made sure it has ".patch" at the end of its name, as Frederic suggests. But with this (or, equivalently, Gonzalo's original three patches) installed I get some doctest failures in sage/matrix/matrix_modn_sparse -- any ideas what's causing that?

[tscrim]

comment:17

Git branch and also has the fixes for the sparse conversion by making the matrix_modn_dense_template inherit from matrix_mod_dense

---

New commits:

c0a74b9	#10734: Matrix_modn_sparse speed improvements
8bc6935	Made modn dense template inherit from mod_dense.

[tscrim]

Changed reviewer from William Stein to William Stein, Travis Scrimshaw

[tscrim]

Branch: public/matrix/improve_modn_sparse_methods-10734

[tscrim]

Commit: 8bc6935

[fchapoton]

comment:18

needs rebase

[sagetrac-git]

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

d3f8b5f

Merge branch 'public/matrix/improve_modn_sparse_methods-10734' of trac.sagemath.org:sage into public/matrix/improve_modn_sparse_methods-10734

[sagetrac-git]

Changed commit from 8bc6935 to d3f8b5f

[tscrim]

comment:20

Rebased to 6.4.rc1.

[videlec]

comment:21

Hello,

What is the purpose of introducing the class Matrix_mod_dense?

Vincent

[videlec]

comment:22

Ticket #17837 proposes to remove matrix_modn_dense (it is unused since a long time and much slower than matrix_modn_dense_float or matrix_modn_dense_double using LinBox). Is there anybody working on this ticket that can say something about it? Otherwise I will put #17837 in positive review (and consequently this one in needs work).

Vincent

[jdemeyer]

Dependencies: #17837