Refactor degree sequence functions

[cxzhong]

Refactor degree sequence generation to use generators, reducing memory usage. Fix #41187

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[cxzhong]

Can you help me check if the result is true? Thank you @maxale

[maxale]

These changes look good to me, but with respect to refactoring, I'd also suggest to change list to tuple to enable easy hashability. Thanks!

[cxzhong]

sage: sum(1 for _ in DegreeSequences(18))
675759564

The result is right. But it takes 11 minutes. Is it normal? @maxale

[cxzhong]

These changes look good to me, but with respect to refactoring, I'd also suggest to change list to tuple to enable easy hashability. Thanks!

I have used tuple instaed of list

[mantepse]

Looks great!

[mantepse]

Maybe it would make sense to implement a cardinality method using https://arxiv.org/abs/2301.07022, section 5.1. Note that go code is attached to the arxiv submission. I guess that translating the go code to python is a bit too much work, but perhaps one can implement a simple version.

[user202729]

wait. Is returning a tuple instead of a list considered a breaking change? (although tuple is "better" in a way that it's immutable and hashable, as pointed out above)

Looks like the original author is no longer around, so I'll ask a few other frequent contributors. @tscrim @dimpase @saraedum

[cxzhong]

As suggestions I would like to use range instead of this loop

[saraedum]

wait. Is returning a tuple instead of a list considered a breaking change? (although tuple is "better" in a way that it's immutable and hashable, as pointed out above)

Looks like the original author is no longer around, so I'll ask a few other frequent contributors. @tscrim @dimpase @saraedum

I don't think we'd usually consider this a breaking change. I don't think we have a clear policy on this.

[user202729]

okay.

cf. #41038 where Family is considered not compatible with tuple

[mantepse]

Do we really want to include the last commit (which is a slowdown)?

[cxzhong]

Do we really want to include the last commit (which is a slowdown)?

If It does not be slower much than before, I think it is acceptable. @user202729

[mantepse]

Do we really want to include the last commit (which is a slowdown)?

If It does not be slower much than before, I think it is acceptable. @user202729

Yes, but is there any good reason to do it? The other loops in the same file are consistently done with the same syntax, and I don't see any advantage of imitating python style in this particular case.

[cxzhong]

Maybe I revert this

[cxzhong]

Do we really want to include the last commit (which is a slowdown)?

If It does not be slower much than before, I think it is acceptable. @user202729

Yes, but is there any good reason to do it? The other loops in the same file are consistently done with the same syntax, and I don't see any advantage of imitating python style in this particular case.

I have done this. Maybe revert is much faster

[cxzhong]

@mantepse can you help me review this?

[github-actions]

Documentation preview for this PR (built with commit 9adc2e0; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[cxzhong]

@user202729 Can you review this again?

[user202729]

• I'd rather not exposing _DegreeSequenceEnumerator until there's a need. It's an internal class after all. (Although not everyone agree on this point.)
• Is it possible to overflow the unsigned char in any reasonable time? (If yes, given the old code does not support n ≥ 256 either, I think that's best left to future work.)
• some ASCII drawing in comments (boxes with # borders) get misaligned, but that's unimportant.

otherwise looks correct.

[cxzhong]

I think _DegreeSequenceEnumerator is a internal class which is only for DegreeSequence. So I do not think it needs to expose to others

[user202729]

one comment left

• Is it possible to overflow the unsigned char in any reasonable time? (If yes, given the old code does not support n ≥ 256 either, I think that's best left to future work.)

but shouldn't be blocking.