Nucleide

Nucleide is a modern Rust toolkit for nuclear-engineering data, measurement, and workflow glue: legacy transport-code I/O, nuclide identification, materials, CRAM depletion, enrichment analytics, point kinetics, gamma-ray spectroscopy, tritium transport, variance reduction, UQ sampling, CSG translation, and code-card emission — exposed through a typed Python API.

The project is a fresh Rust implementation of capabilities pioneered by PyNE, focused on memory safety, fast builds, and pip install-able wheels. Scope is intentionally narrow today and will expand as more parsers and workflow pieces land.

Installation

pip install nucleide

Prebuilt wheels cover Linux, macOS, and Windows for Python >= 3.10 (abi3). To build from source instead, see the Development section below.

Why

Nuclear-engineering workflows spend most of their time moving data between codes rather than solving transport itself. The established tooling for that glue layer carries a heavy build chain (CMake + Fortran + Cython) and hand-written parsers that are hard to extend and harder to embed. Nucleide rebuilds the high-value subset in memory-safe Rust with one-command pip install wheels, keeping Python as the user-facing API.

Features

Out of scope

Transport solvers, Fortran discrete-ordinates ports, ENSDF evaluators, MOAB-dependent meshing, and GUIs. Nucleide complements transport codes; it does not replace them.

Layout

nucleide/
├── crates/
│   ├── nuclei/        # nuclide ids, naming conventions, physical data
│   ├── material/      # compositions, mixing, libraries, XML export
│   ├── mcnp-io/       # xsdir/meshtal/SSW/MCTAL/PTRAC/WWINP
│   ├── csg-xlate/     # scoped MCNP CSG -> OpenMC/Serpent/PHITS translation
│   ├── mcpl-io/       # MCPL interchange read/write + SSW conversion
│   ├── serpent-io/    # res/dep/det readers
│   ├── fluka-io/      # usrbin reader, material cards
│   ├── alara-io/      # ALARA deck/flux/libs/output/photon/schedule glue (no solver)
│   ├── cccc-io/       # ISOTXS/RTFLUX text-subset parsers + PARTISN writer (no solver)
│   ├── fispact-io/    # FISPACT-II inventory output parser (output-only)
│   ├── origen-io/     # scoped ORIGEN TAPE5/6/9 readers
│   ├── r2s/           # scoped R2S workflow builder (photon tags + spectra)
│   ├── vr-tools/      # MAGIC weight windows, source sampling
│   ├── enrichment/    # cascades, SWU
│   ├── depletion/     # CRAM + chain files
│   ├── kinetics/      # prescribed-reactivity point kinetics + inhour
│   ├── tritium/       # 1D diffusion-trapping kernel + permeation checks
│   ├── spectroscopy/  # smoothing, counting, calibration, X-ray, SPE
│   ├── emit/          # five-dialect card emission + mass-drift reports
│   └── linalg/        # isolation facade over the linear-algebra backend
├── bindings/python/   # PyO3 crate -> nucleide._internal
├── python/nucleide/   # typed pure-Python facade (maturin mixed layout)
├── fixtures/          # golden-byte test data
├── validation/        # cross-code validation harness vs PyNE/OpenMC
└── tests/             # Python-side tests

Development

git clone https://github.com/nukehub-dev/nucleide.git
cd nucleide

# Rust side
cargo test                       # workspace unit tests
cargo clippy --all-targets -- -D warnings

# Python side (needs: rustup, pip install maturin)
pip install maturin pytest pytest-cov ruff mypy
maturin develop                  # build + install into current venv
pytest tests/

# Lint / type-check / format the Python surface
ruff format python tests
ruff check python tests
mypy                             # strict type-check against .pyi stubs

# Rust coverage (needs llvm-tools-preview component)
cargo llvm-cov --workspace       # or --lcov for CI upload

Tooling

Wheels are built with maturin (PyO3 mixed layout). One wheel serves all Python >= 3.10 via abi3 — the same stack used by pydantic-core, polars, and ruff.

Validation strategy

1. Parsers are validated against golden-byte fixtures in fixtures/; parser output must match recorded snapshots before any release.
2. Numeric kernels (CRAM, cascade solving) are checked against published analytic vectors and cross-code results on shared inputs; the runnable cross-code harness in validation/ compares Nucleide against PyNE and OpenMC and commits its measured results.
3. Behavioral compatibility with legacy tool output is asserted wherever a fixture exists, so downstream workflows see identical data.

Criterion benchmarks (cargo bench) cover the numeric kernels and parsers.

Citing

If you use Nucleide in research, see CITATION.cff and the JOSS paper draft in paper.md.

Status

Pre-alpha. APIs may change without notice.

Documentation

Additional tutorials, reference pages, and developer guides live in the docs/ tree.

Acknowledgments

Nucleide is a fresh Rust implementation of workflow-glue capabilities pioneered by PyNE ("Python for Nuclear Engineering", BSD-3-Clause). Some reference data and golden test fixtures — notably the DOE/PNNL Materials Compendium — are vendored directly from PyNE; see fixtures/data/MaterialsCompendium.LICENSE for its terms.

License

BSD-2-Clause.