Community call - 2023.09.28

[gdalle]

Agenda

• Taking notes
• Introduction round: people and their projects
• Latest news: new website, GraphsBase.jl
• Organization guidelines (ColPrac?)
• Release Graphs v1.9 together
• Open floor
• Future meetings

See main objectives at the bottom

[gdalle]

Taking notes

One issue per meeting

Introduction round

• Guillaume Dalle: EPFL. Works on graph machine learning and combinatorial optimization. Maintainer of the organization.
• Matt Fishman: Flatiron institute: Works on quantum physics and tensor networks. Needs standardized options graphs with labels and metadata, have developed their own in the meantime.
• Joseph Tindall: Same as Matt.
• Filippos Christou: University of Stuttgart. Works on optical networks. Have developed NestedGraphs.jl and AttributeGraphs.jl
• Aurora Rossi: Université Côte d'Azur. Works on temporal graphs, did GSoC on GraphNeuralNetworks.jl, contributed to Graphs.jl and GraphsOptim.jl.
• Henrik Wolf : Dresden and Copenhagen. Works at the intersection of physics and complex systems, mobility. Interested in MetaGraphs.jl
• Hector Perez: Carnegie Mellon. Works on operations research with RelationalAI.

Miscellaneous

• Recruiting students with GSoC is a good idea

[gdalle]

Latest news

GraphsBase.jl

• will be breaking but the goal is to adapt all algorithms from Graphs.jl
• core idea: vertices can be arbitrary, metadata is supported
• interface design is a hard question and needs feedback
• PRs are open for review (Interface as simple as possible GraphsBase.jl#13)
• will replace at least SimpleWeightedGraphs.jl and MetaGraphs(Next).jl
• should hopefully support multigraphs (see WrappedMultigraphs.jl for inspiration?)
• where to put vertex and edge metadata?
• open issues in the new repo to discuss these things

[gdalle]

Organization guidelines

• ColPrac is a good start
• waiting for review is frustrating on the dev side
• waiting for answers on issues is frustrating on the user side
• some PRs are bad
• some PRs are easy to merge (eg. docs improvement)
• problem of popularity: are there enough PRs to begin with?
• does it matter if PRs and issues just sit there?
• categorizing importance: bug fix > new feature
• should the reviewer understand the algorithm?
• more stringent with other things like tests
• what if there's a bug and the original author has disappeared?
• Test correctness against networkx Graphs.jl#287 could be good for correctness but raises the barrier to contribution a lot
• add a little "please contribute even if you're a beginner"
• leave the easy stuff for new people with "good first issue" (ambiguous wording)
• reviews are valuable even from people without merge rights

[gdalle]

Future meetings

Once per month is a good frequency

[gdalle]

Main objectives in the coming weeks

• Vote on ColPrac in Adopt ColPrac? #10
• Lower the contribution barrier (categorize issues and PRs, invite beginners, review carefully but not too demandingly)
• Give feedback and suggestions for the nascent GraphsBase.jl interface in the repo's issues

[filchristou]

I just realized that there is a community calendar. It might be a good idea to broadcast the event details there.

[gdalle]

I tried but I'm still waiting to hear about who's in charge of it now (the page still says to contact Logan)