CadQuery –- A Python parametric CAD scripting framework based on OCCT • Hacker News Board

Re: CadQuery –- A Python parametric CAD scripting framework based on OCCT

#41

Post by dr_zoidberg » Mon, Feb 07, 2022, 1:20 PM UTC

bb88 wrote: post #30

Already__Taken wrote: post #7

I tried this a couple of months ago, I found what it's trying to do very pleasant to use but the tooling around it is very annoying. The up-to-date version 2 isn't on pypi, conda is thoroughly unpleasant to have as an environment. There seems to be a drive to using it inside it's own editor https://github.com/CadQuery/CQ-editor while nice seems like wasted engineering effort in developing an editor when the language…

Yeah, I think their approach to blaming PyPi for not being able to distribute their 1GB standard install is really putting the blame on the wrong people. https://github.com/CadQuery/cadquery/issues/153#issuecomment... As an example of an approach that doesn't break Pypi, nltk distributes through Pypi but has an download() method for installation. https://pypi.org/project/nltk/ https://www.nltk.org/data.html The fact…

1GB is huge for PyPI. In this kind of situations, is there a good way to provide a PyPI install while providing a secondary channel to download the massive assests? Say that I was distributing a project that relies on a few large binaries, and still want it to be pip-installable, what would be the best way to do that?

Re: CadQuery –- A Python parametric CAD scripting framework based on OCCT

#42

Post by ramboldio » Mon, Feb 07, 2022, 1:34 PM UTC

pmayrgundter wrote: post #32

samwillis wrote: post #22

Earlier quoted context omitted.

That confirms the impression I had, I guess they are just so complex that building a new open source one from scratch would be Herculean effort. Guess OCCT is the only option for now. The itch I would love to scratch is an online real-time collaborative parametric cad app using CRDTs. Current thinking is combining OpenCascade compiled to WASM [0] with Yjs[1], but frankly don’t have the time. A kind of Figma but for 3…

Doing just this. Starting with BIM/IFCjs but have near-term plans for integrating a STEP-compatible engine soon. Thanks for the opencascade/wasm link.. looks like a nice option! github.com/buildrs/Share

"building a new open source one from scratch would be Herculean effort."

how are you tackling this?

Re: CadQuery –- A Python parametric CAD scripting framework based on OCCT

#43

Post by pmayrgundter » Mon, Feb 07, 2022, 5:11 PM UTC

ramboldio wrote: post #42

pmayrgundter wrote: post #32

Earlier quoted context omitted.

Doing just this. Starting with BIM/IFCjs but have near-term plans for integrating a STEP-compatible engine soon. Thanks for the opencascade/wasm link.. looks like a nice option! github.com/buildrs/Share

"building a new open source one from scratch would be Herculean effort." how are you tackling this?

Agree it's a big effort, so approaching as open as we can to attract cooperation. We also have a new company that has taken seed funding. I'm a systems engineer/data scientist, working with architects and 3d folks who know the state of the art well.

The architecture will tend towards platform + apps, to help address the scale problem. We're looking to partner/port the geometry kernel and focus on workflow and integration with open standards (IFC,STEP,gltf,Collada).

Initial UI is in react+three.js (soon react-three-fiber) with git for file-based versioning underneath. Also interested in recent discussions on HN about CRDTs on Matrix for collaboration. So kind of a CRDT real-time google docs capability on top of a persistent file based rep on git. This part is very fun coding :)

Just getting started! Appreciate any feedback, esp on our github/discord :)

Re: CadQuery –- A Python parametric CAD scripting framework based on OCCT

#44

Post by defterGoose » Mon, Feb 07, 2022, 7:47 PM UTC

ur-whale wrote: post #36

defterGoose wrote: post #25

Earlier quoted context omitted.

AFAIK there isn't really another reasonable way of rendering other than having render meshes attached to everything (due to GPU architecture), but they shouldn't be getting used for any calculations so it also shouldn't matter very much.

> it also shouldn't matter very much. There is usually a point at which you need to take your model out of the CAD out of the CAD tool. For example to 3D print or to render. And at that point, fine control on the meshing is rather important. Sure, if you have a manufacturing tool that can directly handle NURBS, great. That's certainly not the majority of the the hobbyist / prosumer market. And the actual pros ... the…

> So, yes: proper conversion to a polygonal mesh with fine control matters a great deal.

Sure, but the output conversion tools shouldn't be built around the render mesh. Any CAD tool worth it's salt should be able to take the best underlying representation of the model and give you an .stl with arbitrary precision. Every time I've ever exported a mesh from a NURBS modeller, I've gotten a dialog that lets you choose parameters.

I'm not so experienced with standalone renderers (I usually only need the built-in ones), but the mesh that the CAD program uses to render should basically only be used for that purpose. Most programs shouldn't even give you access to it unless you open it up and look and the guts.

Re: CadQuery –- A Python parametric CAD scripting framework based on OCCT

#45

Post by eggy » Tue, Feb 08, 2022, 2:04 AM UTC

I started parametric geometry or CAD with AutoLISP, but you were tied to AutoCAD. Then I found Blender in the late 90's and early 2000's, and used Python to do some cool stuff. FreeCAD was my next attempt at using the engine without the GUI, but most of my daytime work in engineering is on Autodesk Fusion or Inventor, or in Rhino. Then I found Rosetta[1], a cool attempt at creating a geometry grammar and a backend for popular CAD programs like AutoCAD or Rhino. You code in Rosetta, a Racket DSL/library, and it renders in the CAD backend. I actually used it to generate a spherical truss dome that I could analyze structurally, change parameters, and reload in the analysis. I was almost at the point of doing my own version of generative design assist that Autodesk now has working. I started to think about writing a backend for FreeCAD, because I prefer Lisps over python for creative coding. I'll have to check CadQuery out now that it has moved from FreeCAD to OCCT. I still like writing AutoLISP for fun. Still not a fan of python for no good reason other than the biased reason that Lisp makes me happy, and python doesn't. CadQuery's Jupyter interface seems like a good sandbox to experiment with if the speed is good for iterative or historical design.

[1] http://web.ist.utl.pt/antonio.menezes.leitao/Rosetta/index.h...