abs-hdf5 0.2.0

A Python + C++ toolkit for reading CAD B-Rep data from HDF5, sampling surfaces/curves, and generating blue-noise point clouds.

ABS-HDF5: Geometry processing & blue-noise sampling for HDF5 B-Rep data

Description

ABS-HDF5 is a Python + C++ toolkit that turns the raw B-Rep information stored in HDF5 CAD datasets into immediately useful geometry:

• read curves, surfaces, topology into convenient Python objects
• sample points on faces/edges with exact parametric control
• perform fast Poisson-disk (blue-noise) down-sampling via a native pybind11 C++ extension
• export ready-to-visualise point clouds in PLY or analysis-ready Pickle files

---

Quick install

pip install abs-hdf5        # pre-built wheels for CPython 3.8 – 3.12

Building from source? You need

• Python 3.8 +
• a C++17 compiler (GCC 9 / Clang 12 / MSVC 17.6 or newer)
• CMake ≥ 3.22 and Ninja

pip install .  # inside a git checkout – scikit-build-core will compile abspy

---

Command-line tools

abs-to-ply

Convert one file (or a directory of .hdf5 files) to point-cloud PLY.

abs-to-ply data/Cone.hdf5   out/          -n 5000  -j 8
# ^input file / dir          ^output dir   ^pts/part ^workers

Each face gets uniformly-random samples; Poisson-disk down-sampling keeps only 5000 nearly-evenly-spaced points per part. Normals are written alongside every vertex in ASCII PLY.

abs-to-pickle

Same interface, but saves a per-part *.pkl with a dict:

{'file': 'Cone.hdf5', 'part': 0, 'points': ndarray(N,3), 'normals': ndarray(N,3)}

Useful for machine-learning pipelines that prefer NumPy-pickles.

---

Library use

import abs                       # auto-loads the C++ extension `abspy`

# Read every part stored in a single HDF5 file
parts = abs.read_parts("data/Cylinder.hdf5")

# How to extract 10 000 blue-noise samples on each part (with normals)
def face_normals(part, topo, uv):
    return topo.normal(uv) if topo.is_face() else None

points_per_part, normals_per_part = abs.sample_parts(
        parts, num_samples=10_000, lambda_func=face_normals)

# Poisson-disk down-sample an arbitrary XYZ array to 1 000 pts
idx = abs.poisson_disk_downsample(points_per_part[0], 1000)
sub_pts = points_per_part[0][idx]

See the doc-strings in abs.part_processor, abs.sampler, and abs.shape.Shape for the full low-level API.

Tip: It is recommended to use a virtual environment for isolated installations.

For more usage details, please refer to the Usage page in our documentation.

Documentation

Detailed documentation is available on our website. It includes:

• Installation Guide
• Usage Examples
• API Documentation: Covers modules like Geometry, Topology, Sampling, and Visualization.
• FAQ, Changelog, Contributing Guidelines, and More

Visit our website: better-step.github.io

Contributing

Contributions are welcome! Please follow these guidelines:

1. Fork the Repository: Click the "Fork" button on GitHub.

2. Clone Your Fork Locally:

   git clone https://github.com/better-step/abs.git
   cd abs
   python -m pip install -e .
   pytest
   

3. Create a Branch:

   git checkout -b feature/your-feature-name
   

4. Make Your Changes: Ensure your code follows our coding standards (PEP 8) and include docstrings and tests where applicable.

5. Commit and Push:

   git add .
   git commit -m "Add feature: [description]"
   git push origin feature/your-feature-name
   

6. Submit a Pull Request: Open a pull request on GitHub with a clear description of your changes.

Note: Please also review our Code of Conduct before contributing.

License

• Python bindings, packaging scripts, tests – MIT License (see LICENSE-MIT)
• Embedded C++ core – Mozilla Public License 2.0 (see LICENSE-MPL-2.0)

Acknowledgments

Better Step is developed and maintained by a dedicated team. For a complete list of contributors, please see the Authors page.

---

Happy sampling! – The Better Step maintainers