Hanno Braun

This month's sponsor update just went out! Summary here: www.fornjot.app/blog/sponsor...

The latest sponsor update just went out. Some bad news in this one! Summary here: www.fornjot.app/blog/sponsor...

This month's sponsor update just went out, which includes an overview over my current 4-phase roadmap. Summary here: www.fornjot.app/blog/sponsor...

After the last sponsor update was so late, it's extra-nice to already have a finished draft for the next one. Unless something unexpected gets in the way, I can do the last round of edits and publish tomorrow!

Using triangle meshes as a uniform intermediate representation is a core idea behind this experimental code after all! Doing it this way makes adding new kinds of surfaces simpler.

This also surfaced an architectural problem: The operation that projects points into surfaces needs to be implemented per-surface. Which was reasonable so far, because the latest experiment only supports one kind of surface. But I'd rather do this projection based on the surface's triangle mesh.

I made some progress on those buggy faces I showed yesterday. The problem is that so far, surfaces were curved along one axis at most, while the surfaces of the faces I posted are curved along both axes. And it turns out I missed a few spots that operate based on wrong assumptions now.

My art collection of buggy faces. No idea yet what exactly is going wrong here 😂

crates.io: Rust Package Registry

Last time this happened, I hacked something together using Plotters (crates.io/crates/plott...). Which worked well! But I figured that for a permanent solution, adding a new feature to the existing infrastructure would be better. Much more simple!

I'm working on a bug that manifests as a panic during triangulation. Since the debug viewer could previously only display triangle meshes (the product of triangulation), that kind of issue was hard to debug.

Fornjot's debug viewer can now display faces before they are triangulated!

This month's sponsor update finally went out! Summary here: www.fornjot.app/blog/sponsor...

I'm very sorry, but the sponsor update is going to be late this month. I should be able to get it out on Monday or Tuesday!

Recently, working on Fornjot has been fairly productive, with some good progress towards more advanced geometry. Today though... github.com/Smithay/wayl... 😄

No, it doesn't do NURBS. The geometry it can handle is still super-basic as of now, though the situation is slowly improving.

So now the latest experiment supports arcs. Which puts it roughly at parity with the mainline code in regards to what kind of geometry can be supported! Lots more to do, but things are coming along nicely.

Turns out the bug wasn't even in the brand-new and experimental code I wrote. That was working flawlessly! It was super-old code in the math library, for projecting a point onto a circle. It assumed that the circle was in the x-y plane, regardless of how it was actually oriented.

Now that is much better!

This is not what I expected to see when adding support for arcs 😂

My latest sponsor update just went out! Summary here: www.fornjot.app/blog/sponsor...

Two open windows in front of a desktop background, one showing an intermediate state from the construction of a 3D model, the other showing the final result.

The new, simplified approach to debugging models is working great! You can now display any intermediate state graphically. It's basically `println!`, but for triangle meshes.

Next up, I plan to improve the model viewer API, to make it easier to debug code both within Fornjot, but also code built on top. So far, the viewer has been pretty rigid, built for displaying the final model and not intermediate steps.

So far, the experiments have stood on their own, with no code re-use from the mainline code. (I wanted to make sure I can act freely, without limitations due to legacy decisions.) This is starting to change now, as I'm figuring out the direction in which I want to take this.

I've also replaced the experimental model viewer and export code with its equivalents from the mainline. Some of the mainline code (especially around the viewer) was simplified a lot to make this happen. More steps in the right direction!

I've started to merge back some of the new code from the ongoing experiments back into the mainline code. Just some low-level math code so far. Nothing that's core to the CAD modeling. But improvements from the experiments are starting to trickle back!

It's unfortunate that Fornjot always seems to fall by the wayside when I have something going on that prevents me from working at full capacity. It is what it is, unfortunately. But I'm always trying to find ways to improve in that regard.

fornjot/experiments/2024-12-09 at main · hannobraun/fornjot Early-stage b-rep CAD kernel, written in the Rust programming language. - hannobraun/fornjot

But I'm back now! I documented [the latest experiment](github.com/hannobraun/f...) to conclude it, and I'm excited to move on to the next one. Good stuff to come, hopefully.

I've been on a bit of an involuntary break from Fornjot development. First due to the usual start-of-month stuff (mostly bookkeeping and writing the sponsor update), then because I was sick.

My latest sponsor update just went out! It covers a big breakthrough and a setback on sponsorship income, among other topics. Summary here: www.fornjot.app/blog/sponsor...

There's still some weird stuff going on. I have no idea why some of those top triangles have a different shade of green. But PrusaSlicer says the exported 3MF file has no errors, so it might just be my dodgy rendering code 🤷