cascadio

cascadio

A Python library which uses OpenCASCADE to convert STEP files to a GLB file which can quickly be loaded by trimesh and other libraries.

The primary effort here is build and packaging using the wonderful work done recently on scikit-build-core and cibuildwheel. The goal is to produce wheels that don't require users to build OpenCASCADE themselves.

This is not intended to be a full binding of OpenCASCADE like OCP or PythonOCC. Rather it is intended to be an easy minimal way to load boundary representation files into a triangulated scene in Python. There are a few options for loading STEP geometry in the open-source ecosystem: GMSH, FreeCAD, etc. However nearly all of them use OpenCASCADE under the hood as it is pretty much the only open-source BREP kernel.

Install

The primary goal of this project is building wheels so vanilla pip can be used:

pip install cascadio

Currently this works on non-MUSL flavors of Linux, Windows x64, and MacOS x64+ARM. You can check PyPi for current platforms.

:warning: :warning: :warning: :warning:

PyPI has a size limit, and each release of this is large! We will not be keeping every release on PyPi (i.e. if we run out of space we delete versions) so be very careful pinning the version!

We'll keep the following versions as "LTS" style releases on PyPi:

pip install cascadio==0.0.13

Motivation

A lot of analysis can be done on triangulated surface meshes that doesn't need the analytical surfaces from a STEP or BREP file.

Contributing

Developed on Linux which should build wheels locally with docker:

# this doesn't cache the OCCT build unfortunately.
# It would be nice if it did! You could do it by building OCCT
# in the manylinux images and then passing the new tag to CIBW
CIBW_BUILD="cp312-manylinux_x86_64" cibuildwheel --platform linux

Or, if you want to develop that will only work in your local environment for development:

# just run the `before-all` from pyproject.toml which is approximatly:
cd upstream/OCCT
cmake -G Ninja -DCMAKE_BUILD_TYPE=Release \
      -DUSE_RAPIDJSON:BOOL="ON" \
      -D3RDPARTY_RAPIDJSON_INCLUDE_DIR="../rapidjson/include" .
ninja
mv lin64/gcc/lib .

Then pip install . will build and install locally. Make sure to point LD_LIBRARY_PATH=upstream/OCCT/lin64/gcc/lib or wherever you put the libraries.

Future Work

Pull requests welcome!

• Add passable parameters for options included in the RWGLTF writer.
• use in-memory data for input and output, i.e. stepReader.ReadStream() instead of a file name. Ideally the Python function signature would be:
  • convert_to_glb(data: bytes, file_type: str, **parameters) -> bytes
  • Currently using file names because it's easier.
• Support IGES
  • Investigate using OpenCASCADE "Advanced Data Exchange" for Parasolid .x_b/.x_t and JT .jt support.