Categories for Weyl character rings and weight rings (was: pickling fails in WeightRing)

[nthiery]

This ticket refactors the code of WeylCharacterRing and WeightRing to use categories and (combinatorial) free modules. Along the way, it adds a couple features (Dan: please list them here), and solves a pickling issue which was caught by #7921:

sage: A2 = WeylCharacterRing(['A',2])
sage: a2 = WeightRing(A2)
sage: TestSuite(a2).run()
Failure in _test_element_pickling:
Traceback (most recent call last):
...
AssertionError: 2*a2(0,0,0) != 2*a2(0,0,0)

Indeed:

sage: x = a2.an_element()
sage: x == loads(dumps(x))
False

Apply attachment: trac_7922-rebased-4.7.alpha3.patch

Remove the following pickles from the pickle jar:

_class__sage_combinat_root_system_weyl_characters_WeightRing__.*
_class__sage_combinat_root_system_weyl_characters_WeylCharacterRing_class__.*
_class__sage_combinat_root_system_weyl_characters_WeylCharacter__.*

Copy the contents of
attachment: trac_7922-new_pickles.tar.gz into data/ext_code/pickle_jar.

CC: @dwbump @sagetrac-sage-combinat

Component: combinatorics

Author: Daniel Bump, Nicolas M. Thiéry

Reviewer: Nicolas M. Thiéry, Dan Bump

Merged: sage-4.7.1.alpha2

Issue created by migration from https://trac.sagemath.org/ticket/7922

[dwbump]

comment:1

Regarding speed, testing a large number of branching rules show that the patch is a substantial speedup over the old code with a caveat: the old code has an option cache=true for WeylCharacterRings. If this option (which is not the default) is selected for every WeylCharacterRing, then the old code is faster. Typical times:

Old Code	48 seconds
New Code	25 seconds
Old Code, cache=true	18 seconds

Since I made the method _irr_weights a cached method, the caching done in the new code is approximately equivalent to the caching in the old code, so at the moment I don't see any way to improve the situation.

[dwbump]

Reviewer: nthiery

[dwbump]

comment:2

Revised and reposted the patch in view of Nicolas' comment to use _from_dict(coerce=True).

[dwbump]

Converts WeylCharacterRings and WeightRings to use category framework

[dwbump]

comment:3

Attachment: trac_7922.patch.gz

I uploaded a revised version of the patch. The only change is in classical_crystals.py, where the revision of WeylCharacterRings necessitated a revision.

[dwbump]

comment:4

Requires rebuilding the pickle jar.

[dwbump]

Attachment: trac_7922-rebased-4.6.1.gz

#7922: WeylCharacters inherit from CombinatorialFreemodule (substantial speedup)

[dwbump]

comment:5

Since #9838 was merged in sage-4.6.1.alpha1, this patch needed rebasing.

I therefore posted trac_7922-rebased-4.6.1.

[dwbump]

7922: thematic tutorial revision

[dwbump]

Work Issues: pickle jar will have to be rebuilt

[dwbump]

comment:7

Attachment: trac_7922-doc.patch.gz

This patch slightly conflicts with #8442 which was merged. So I'm posting a second patch trac_7922-doc.patch which revises the thematic tutorial.

[dwbump]

Attachment: trac_7922-rebased-4.7.alpha1.patch.gz

#7922: Revision of Weyl Character Rings

[dwbump]

comment:8

I have posted trac_7922-rebased-4.7.alpha1.patch, which
addresses many of the comments in the reviewer patch:

http://combinat.sagemath.org/patches/file/tip/trac_7922-review-nt.patch

Those changes that are not addressed are discussed here:

http://groups.google.com/group/sage-combinat-devel/browse_thread/thread/277e146862632d72

[dwbump]

#7922: revised pickle jar

[dwbump]

comment:9

Attachment: pickle_jar.tar.bz2.gz

[dwbump]

Changed work issues from pickle jar will have to be rebuilt to none

[dwbump]

#7922: revision of Weyl Character Rings

[dwbump]

Attachment: trac_7922-rebased-4.7.alpha2.patch.gz

#7922: replacement pickles

[nthiery]

comment:13

Attachment: trac_7922-new_pickles.tar.gz

Hi Dan,

I just did a final proofreading, fixed a couple typos, updated coerce_to_sl in the doctests of the thematic tutorials, and removed some trailing white space and tabs.

For me it is now all good to go. Please check my reviewer's patch on the sage-combinat patch server. If it's ok with you, you can fold/upload and set a positive review here on my behalf.

http://combinat.sagemath.org/patches/file/tip/trac_7922-final-review-nt.patch

Cheers,
Nicolas

[nthiery]

Changed reviewer from nthiery to Nicolas M. Thiéry, Dan Bump

[nthiery]

Author: Dan Bump, Nicolas M. Thiéry

[dwbump]

#7922: Categories for Weyl character rings and weight rings

[nthiery]

comment:15

Attachment: trac_7922-rebased-4.7.alpha3.patch.gz

Hi Dan,

I just checked out your final changes on the Sage-Combinat queue (trac_7922_alpha3-changes.patch), and it looks all good. So the final rebased patch you just posted (and which includes the above) is good to go.

Positive review!

[jdemeyer]

Merged: sage-4.7.1.alpha2

[jdemeyer]

comment:18

This needs a few small fixes:

1. at various places doc/en/thematic_tutorials/lie/weyl_character_ring.rst, the indentation is inconsistent. It should be 4 spaces but in the newly added examples it is more or less random.
2. in doc/en/thematic_tutorials/lie/weyl_character_ring.rst, the test on line 250 takes a very long time but is not marked as such.

Please add an additional patch fixing these issues.

[nthiery]

comment:20

Replying to @jdemeyer:

This needs a few small fixes:

1. at various places doc/en/thematic_tutorials/lie/weyl_character_ring.rst, the indentation is inconsistent. It should be 4 spaces but in the newly added examples it is more or less random.
2. in doc/en/thematic_tutorials/lie/weyl_character_ring.rst, the test on line 250 takes a very long time but is not marked as such.

Please add an additional patch fixing these issues.

Fixing the indentation is now #14678. The very long time is not necessary anymore, most likely thanks to the optimizations in #13391 (the example takes 0.22s on my machine).

[fchapoton]

Changed author from Dan Bump, Nicolas M. Thiéry to Daniel Bump, Nicolas M. Thiéry