Mesa Development Session Notes (December 2021)

[tpike3]

Tasks to Complete by Next Dev Session (January 2022):

Interested in helping develop Mesa, come to the next dev session

**Pull Request Review: ** (request #, person, difficulty, category)
1091 tpike3 Easy Docs COMPLETE
1090 tpike3 Easy Docs COMPLETE
1087 jackiekazil Significant Backward compatibility COMPLETE
1084 jackiekazil Significant Backward compatibility COMPLETE
1083 danielxu05 Medium Checks
1075 tpike3 Easy Check Fail COMPLETE
1057 tpike3 Easy Example merge
1035 rht TBD Checks
998 majdal Easy Docs
991 jackiekazilTBD validate
924 rht and tpike3 significant validate/backward compatibility MERGED--Moved to 1108 for docs update and deprecation warning

Long Term Projects

Scalability:

Cython implementation (rht, tpike3, danielxu5)

• Best Approach?: build a separate Cython based Mesa or build a conversion module
• GPUs: Can Cython approach also allow for GPU usage

Education

• Talk with Santa Fe Institute about building a Complexity Explorer based tutorial (tpike3)

[rht]

spaCy has a note on how it uses Cython: https://spacy.io/api/cython. The quickest win we can get is maybe to cythonize the Grid class:

• Faster than using Numpy ndarray on small array size. See https://stackoverflow.com/questions/6559463/why-is-numpy-array-so-slow
• The GIL can be disabled

One drawback with using Cython is that it won't work in PyPy.

[jackiekazil]

@tpike3 Ty for posting these!

[jackiekazil]

#991 needs a conflict resolution. I can either do that OR someone else can and I can evaluate. I don't think I can should be doing it and merging. Not sure if @Corvince has bandwidth these days -- @Corvince, please comment on the ticket if you do.

[Corvince]

Hey there, thanks for pinging me, I wouldn't have noticed otherwise.
Actually it's a bit better these days for me so hopefully I can fix the conflicts the next days

[jackiekazil]

<3 Awesome! Let me know if you need any help!

[tpike3]

SFI Complexity Explorer is excited about a Mesa tutorial. I will fill out the attached questionnaire and share. Next step will be to have a Zoom call and start talking specifics.

ComplexityExplorer Online Course Proposal.txt

[EwoutH]

There are currently at least 3 PRs open that would reduce backward compatibility. I would propose:

1. Merging as much as possible bugfixes and improvements that don't break backwards compatibility
2. Tagging and deploying a Mesa 0.8.10 release
3. Merging the things that do break backwards compatibility
4. Tagging a 0.9.0 release

GitHub Milestones are a great way to keep track which issues and PRs are scheduled for which release. Creating Milestones for 0.8.10 and 0.9.0 could be very useful.

[rht]

Would be great if @EwoutH's PRs that don't break compatibility can be reviewed and merged soon. Otherwise, there is this psychological barrier that any contribution one make take weeks to be merged.

Can be merged IMO:

• Convert all text strings to f-strings #1099
• Update intro_tutorial.rst #1097
• Docs: Add back some comments in space.py #1091
• Docs: space.py: Fix single case of neighbor spelled as neighbour #1090

[tpike3]

Concur and done!

[tpike3]

I filled out the complexity explorer questionnaire (attached). Please feel free to provide feedback, I will send out before the new year.

Complexity Explorer Proposal.pdf
(changed to pdf here too)

[rht]

The file is in Windows line endings. I tried to convert it to Unix using dos2unix to no avail. I think it is better if you format it as pdf.

[tpike3]

Arrgh...please see attached .pdf
Complexity Explorer Proposal.pdf

[rht]

• Typo: should be "agent-based model" instead of "agent based model"
• For bullet point 12, I think an assessment for the part 2 of the course should be a merged pull request? There are plenty of NetLogo examples that can be ported to Mesa. See https://github.com/NetLogo/models/tree/master/Sample%20Models

[rht]

See also #134.

[jackiekazil]

LGTM.
@rht is right on the separate repo. We need to prioritize next dev sprint.

[tpike3]

Interesting, so to makes sure I am understanding..... part 1 of the tutorial could be building an model (such as from nlogo examples) and then part 2 is running through a merge request to add to Mesa model library>

Then their assessment is building their own example model and doing a merge into a Mesa examples folder?

I like it.....that is a rock solid assessment.

(I will also fix the typo)

[rht]

OK that makes sense.

[rht]

IMO #134 needs to be resolved before you can submit the proposal. My vote is to make a separate repo for the new example contributions.

[tpike3]

Roger, concur, I think a separate repo is the common consensus. It will have to be discussed as the next dev meeting.

However, I think we can probably submit the proposal before the next dev meeting as the details don't have to be flushed out until we record (I am thinking the earliest is June/July 2022) . I do however need to update various questions for more robust assessments in which students build an independent example and then do a pull request to the example folder.

[tpike3]

Copy of submitted proposal to SFI.

Mesa Complexity Explorer Proposal.pdf

[tpike3]

SFI has accepted and is starting to work their support piece.

[jackiekazil]

Help me out -- where are we with merges and what do we need to for the next release? (I lost track with the holiday.)

[tpike3]

The minimum I am tracking for the next release is the update to docs for BatchRunner which @jonahaag is working on.

The other one we may want to consider is the contribution by @rht that improves performance by optimizing empties.

My understanding is once we have a consensus self.empties it should be pretty quick

[jackiekazil]

So I should hold off until those 2 things come thru?

[tpike3]

I would say yes!