GSoC projects

[pca006132]

List of mentors:

1. @elalish
2. @pca006132
3. @zalo

Potential Projects:

• Better mesh ID handling (mesh id handling #678) (@pca006132)
• Add minGap function (Add minGap function #186) (@pca006132)
• Fix vert properties and improve documentation on vert properties (vertProperties behavior #536 How does Manifold::SetProperties and properties in general work? #661)
• SDF batch LevelSet (SDF updates #655)
• Add more fuzz tests (more fuzz tests #667) (@pca006132)
• Mesh UV map generation (Mesh UVs #683)
• Improve mesh simplification (Mesh Simplification is incomplete after one pass #675)

Some are probably too hard for GSoC. We can probably add some js/python projects as well to attract beginners.

[elalish]

Great starter list, thanks! We can remove refine to length - I'm about halfway done with that now. A few others I've thought of:

• Add undo, redo, format, and share buttons to ManifoldCAD.org.
• Better make-manifold function with hole filling.
• Refine to precision - make heuristic for number of divisions of each edge based on its tangents.

[elalish]

Actually, @pca006132 already implemented just that: #643 - we just need to hook it up to a button.

[pca006132]

Not sure if the manifoldcad.org button project will be too trivial for GSoC. Maybe a good fit for short (75h)?

[elalish]

Yeah, and honestly I might do it before summer anyway.

[pca006132]

@elalish thinking if it is a good idea to combine the mesh ID thing and improving mesh simplification. They are related, and combining them can free one slot (considering we can have at most 2 slots).

[elalish]

Sure, sounds good.

[pca006132]

Another project: optimize quickhull implementation. I think it should be possible to look into the code, find places where there are bottlenecks and do something with it. E.g. parallelization.

[pca006132]

copying @zalo's comment on zulip:

I believe the VHACD library switched from QuickHull to Julio Jerez (of Newton Physics)'s Convex Hull algorithm because it's much faster and more numerically stable:
https://github.com/kmammou/v-hacd/blob/master/include/VHACD.h#L1988-L3489

[pca006132]

@zalo would you like to mentor this? I think you are more knowledgeable than me about this.

[zalo]

I have some idea about the pitfalls in common convex hull algorithms that I might be able to provide PR review/supervision for if anyone is interested in tackling speed improvements 👍

[pca006132]

@elalish are we using the metallic menger sponge as logo?

[elalish]

Indeed - and it looks like @t-paul already grabbed it.

[t-paul]

If you would like to see anything updated on the website part, you can do that via PR or just let me know. For the repository with the project proposals you should have direct access as all the other mentors. We'll have to see how the discussion goes regarding the presentation, but I hope we can keep the project list there. I don't think it's viable to host a separate shared application just for presenting the projects, nobody has time for that. As mentioned in the Zulip chat, I can see an additional nicer view-only front-end as entry page replacing the current very static github-page at https://opencax.github.io/.

[t-paul]

Initial update for 2024 at https://opencax.github.io/ - we still need to collect the definitive list of projects to take part this year and probably dust of some of the sub-pages.

[elalish]

When should we start putting our projects into this issue list? What do you think is a good number?

[pca006132]

I think we probably want to claim the projects first, for the potential mentor thing, and then the potential mentor can put the project into the issue list.

Should we add a communication channel? It seems that everyone else have something other than github discussion. Maybe we want to use discord/zulip? I am OK with any of them.

For project home page, we probably want to use https://github.com/elalish/manifold instead of manifoldcad page?

[elalish]

I think that's just because those projects are older than Github discussions are. I prefer to have everything in one place. ManifoldCAD does link to the github, plus it's a bit more exciting from a newbie perspective.

@t-paul I read something about Jan 22nd - when are we trying to get things ready by?

[t-paul]

Jan 22nd the org applications open, that means Sean will apply as BRL-CAD umbrella org to GSoC, I would assume we need to have all ducks in a row when the application deadline happens at Feb 6th (https://developers.google.com/open-source/gsoc/timeline). After that the Google project admins will likely start looking at the referenced presentations.

Following what Sean said, a good number for proposals might be 3 - 5. With 3 mentors, the suggestion would be taking on at most 2 projects. Ideally every project would have 2 mentors.

As for communication channel, it seems fine to just keep Github discussions, it's indeed only usable after most of the projects found their places. When thinking about some place to chat and IRC seems not attractive enough, I'd also point to Matrix as potential option. The closed nature of Discord feels not ideal for open source projects as it makes it difficult to get data out. Even Github makes it very easy to get pretty much all data out via a backup - I do that once a month for the most central OpenSCAD repos. I have not seen that for Discord and history shows policies will change at some point.

[pca006132]

@elalish @zalo There are some discussions going on in the BRL-CAD zulip (https://brlcad.zulipchat.com/#narrow/stream/111975-Google-Summer-of-Code/topic/manifold), it would be better if you can join the discussion over there. Zulip provides email discussion by default so you will not miss the messages there even if you are not online.