agnimhd 0.1.0__tar.gz

agnimhd-0.1.0/.zenodo.json

@@ -0,0 +1,13 @@
+{
+  "title": "agnimhd: a differentiable finite-n ideal MHD stability solver",
+  "upload_type": "software",
+  "description": "AGNI computes the most unstable finite-n ideal MHD mode of a 3D equilibrium: it discretizes the energy principle on a straight-field-line grid and returns the squared growth rate of the most unstable mode and its derivative with respect to the equilibrium's parameters, for optimization through equilibrium codes such as DESC.",
+  "creators": [
+    {"name": "Gaur, Rahul"}
+  ],
+  "license": "MIT",
+  "keywords": ["plasma physics", "MHD stability", "stellarator", "tokamak", "JAX", "differentiable"],
+  "related_identifiers": [
+    {"identifier": "10.48550/arXiv.2608.01750", "relation": "isSupplementTo", "resource_type": "publication-preprint", "scheme": "doi"}
+  ]
+}

agnimhd-0.1.0/CITATION.cff

@@ -0,0 +1,26 @@
+cff-version: 1.2.0
+message: "If you use agnimhd, please cite the AGNI paper."
+title: "agnimhd"
+repository-code: "https://github.com/rahulgaur104/AGNI"
+license: MIT
+authors:
+  - family-names: Gaur
+    given-names: Rahul
+preferred-citation:
+  type: article
+  title: "AGNI: A differentiable MHD stability solver & optimizer for magnetic confinement fusion devices"
+  authors:
+    - family-names: Gaur
+      given-names: Rahul
+    - family-names: Patil
+      given-names: Sanket
+    - family-names: Gupta
+      given-names: Prateek
+    - family-names: Patch
+      given-names: Djin
+    - family-names: Qian
+      given-names: Tony
+  year: 2026
+  journal: "arXiv preprint arXiv:2608.01750"
+  doi: "10.48550/arXiv.2608.01750"
+  url: "https://arxiv.org/abs/2608.01750"

agnimhd-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Rahul Gaur
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

agnimhd-0.1.0/MANIFEST.in

@@ -0,0 +1,12 @@
+# The sdist carries the package and what describes it; the test data, examples,
+# benchmarks, figures and the paper stay in the repository.
+include README.md LICENSE CITATION.cff pyproject.toml
+graft src
+prune tests
+prune examples
+prune benchmarks
+prune docs
+prune tools
+global-exclude __pycache__ *.py[cod]
+prune .github
+exclude .flake8 .gitignore .pre-commit-config.yaml .test_durations mkdocs.yml

agnimhd-0.1.0/PKG-INFO

@@ -0,0 +1,148 @@
+Metadata-Version: 2.4
+Name: agnimhd
+Version: 0.1.0
+Summary: AGNI: differentiable finite-n ideal MHD stability solver
+Author-email: Rahul Gaur <rgaur@terpmail.umd.edu>
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/rahulgaur104/AGNI
+Project-URL: Issues, https://github.com/rahulgaur104/AGNI/issues
+Project-URL: Paper, https://arxiv.org/abs/2608.01750
+Project-URL: Documentation, https://rahulgaur104.github.io/AGNI/
+Keywords: plasma,mhd,stability,stellarator,tokamak,jax
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Physics
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: jax>=0.4.30
+Requires-Dist: numpy>=1.24
+Requires-Dist: scipy>=1.10
+Requires-Dist: matfree>=0.6.2
+Provides-Extra: hdf5
+Requires-Dist: h5py>=3.0; extra == "hdf5"
+Provides-Extra: plot
+Requires-Dist: matplotlib>=3.7; extra == "plot"
+Provides-Extra: multigpu
+Requires-Dist: jaxmg; extra == "multigpu"
+Provides-Extra: test
+Requires-Dist: pytest>=7.0; extra == "test"
+Requires-Dist: pytest-cov>=4.0; extra == "test"
+Requires-Dist: pytest-split>=0.8; extra == "test"
+Requires-Dist: pytest-mpl>=0.16; extra == "test"
+Provides-Extra: benchmark
+Requires-Dist: pytest-benchmark>=4.0; extra == "benchmark"
+Provides-Extra: dev
+Requires-Dist: agnimhd[hdf5,plot,test]; extra == "dev"
+Requires-Dist: black>=24.0; extra == "dev"
+Requires-Dist: flake8>=7.0; extra == "dev"
+Requires-Dist: isort>=5.13; extra == "dev"
+Requires-Dist: pre-commit>=3.5; extra == "dev"
+Provides-Extra: docs
+Requires-Dist: mkdocs>=1.5; extra == "docs"
+Requires-Dist: mkdocs-material>=9.5; extra == "docs"
+Requires-Dist: mkdocstrings[python]>=0.24; extra == "docs"
+Requires-Dist: mkdocs-jupyter>=0.24; extra == "docs"
+Provides-Extra: notebooks
+Requires-Dist: nbmake>=1.5; extra == "notebooks"
+Requires-Dist: ipykernel>=6; extra == "notebooks"
+Dynamic: license-file
+
+# agnimhd
+
+[![CI](https://github.com/rahulgaur104/AGNI/actions/workflows/ci.yml/badge.svg)](https://github.com/rahulgaur104/AGNI/actions/workflows/ci.yml)
+[![docs](https://github.com/rahulgaur104/AGNI/actions/workflows/docs.yml/badge.svg)](https://rahulgaur104.github.io/AGNI/)
+[![PyPI](https://img.shields.io/pypi/v/agnimhd)](https://pypi.org/project/agnimhd/)
+
+AGNI is a differentiable finite-n ideal MHD stability solver. It discretizes the
+ideal MHD energy principle on a straight-field-line grid and returns the squared
+growth rate of the most unstable mode (positive: unstable) and its gradient. It
+computes the stability objective; it does not run an optimization itself.
+
+The package is under active development, and the API, the file format and the
+numerics change without notice. For a stable version use the AGNI
+implementation inside DESC,
+[PR #1893](https://github.com/PlasmaControl/DESC/pull/1893), branch
+`rg/AGNI_var`.
+
+## Quick start
+
+Stability of a DESC equilibrium (needs DESC installed):
+
+```bash
+agnimhd solve my_equilibrium.h5 --res 24,12,8
+```
+
+prints `gamma^2` for each toroidal mode family `n = x + k NFP`, all solved on
+one field period, and names the most unstable.
+
+From Python:
+
+```python
+import agnimhd
+
+basis = agnimhd.Basis(24, 12, 8)                 # n_zeta = 8 per field period
+eq, diffmat = agnimhd.from_desc("my_equilibrium.h5", basis, family=0)
+gamma2 = agnimhd.growth_rate(eq, diffmat)        # gamma^2 = -lambda > 0: unstable
+```
+
+`growth_rate` is solve mode and is not differentiable. The gradient comes from
+optimize mode, `jax.grad(agnimhd.growth_rate_of)(params, equilibrium_map,
+diffmat)`, with `equilibrium_map` the equilibrium code's map from its parameters
+to an `EquilibriumData`.
+
+Matrices too large for one GPU: `SolverConfig(eigensolver="dense_mg")` splits the
+dense matrix over all GPUs of a node with JAXMg (tested with `jaxmg==0.0.9`);
+182,784 unknowns in 8.5 minutes on four A100s. See the multi-GPU page of the
+documentation.
+
+Two shipped cases run without DESC: `python examples/cross_sections.py`. Other
+codes build an `EquilibriumData` from arrays, or load a saved one with
+`EquilibriumData.load("eq.npz")`. See the
+[documentation](https://rahulgaur104.github.io/AGNI/).
+
+## Install
+
+```bash
+pip install agnimhd               # jax, numpy, scipy, matfree
+pip install "agnimhd[hdf5,test]"  # HDF5 files and the test suite
+```
+
+Python 3.12. DESC is optional and only needed for `from_desc`. GPU, DESC and
+multi-GPU set-ups: [Installation](https://rahulgaur104.github.io/AGNI/installation/).
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pre-commit install
+pytest tests -q          # about 25 min on one CPU
+```
+
+Releasing: tag `vX.Y.Z` on `master` and publish a GitHub release of it. The
+`release` workflow builds the package from the tag and uploads it to PyPI, and
+Zenodo archives the release with a DOI. The version comes from the tag
+(`setuptools_scm`); nothing in the source is edited.
+
+## Reference
+
+R. Gaur, S. Patil, P. Gupta, D. Patch, T. Qian, *AGNI: A differentiable MHD
+stability solver & optimizer for magnetic confinement fusion devices*,
+[arXiv:2608.01750](https://arxiv.org/abs/2608.01750) (2026),
+doi:[10.48550/arXiv.2608.01750](https://doi.org/10.48550/arXiv.2608.01750).
+PDF of v3 in this repository: [docs/paper/agni_arxiv_2608.01750v3.pdf](docs/paper/agni_arxiv_2608.01750v3.pdf).
+
+```bibtex
+@article{gaur2026agni,
+  title   = {{AGNI}: A differentiable {MHD} stability solver \& optimizer for magnetic confinement fusion devices},
+  author  = {Gaur, Rahul and Patil, Sanket and Gupta, Prateek and Patch, Djin and Qian, Tony},
+  journal = {arXiv preprint arXiv:2608.01750},
+  year    = {2026},
+  doi     = {10.48550/arXiv.2608.01750}
+}
+```
+
+Originally developed inside [DESC](https://github.com/PlasmaControl/DESC)
+([PR #1893](https://github.com/PlasmaControl/DESC/pull/1893), on the
+differentiation matrices of PR #1789). MIT license.

agnimhd-0.1.0/README.md

@@ -0,0 +1,97 @@
+# agnimhd
+
+[![CI](https://github.com/rahulgaur104/AGNI/actions/workflows/ci.yml/badge.svg)](https://github.com/rahulgaur104/AGNI/actions/workflows/ci.yml)
+[![docs](https://github.com/rahulgaur104/AGNI/actions/workflows/docs.yml/badge.svg)](https://rahulgaur104.github.io/AGNI/)
+[![PyPI](https://img.shields.io/pypi/v/agnimhd)](https://pypi.org/project/agnimhd/)
+
+AGNI is a differentiable finite-n ideal MHD stability solver. It discretizes the
+ideal MHD energy principle on a straight-field-line grid and returns the squared
+growth rate of the most unstable mode (positive: unstable) and its gradient. It
+computes the stability objective; it does not run an optimization itself.
+
+The package is under active development, and the API, the file format and the
+numerics change without notice. For a stable version use the AGNI
+implementation inside DESC,
+[PR #1893](https://github.com/PlasmaControl/DESC/pull/1893), branch
+`rg/AGNI_var`.
+
+## Quick start
+
+Stability of a DESC equilibrium (needs DESC installed):
+
+```bash
+agnimhd solve my_equilibrium.h5 --res 24,12,8
+```
+
+prints `gamma^2` for each toroidal mode family `n = x + k NFP`, all solved on
+one field period, and names the most unstable.
+
+From Python:
+
+```python
+import agnimhd
+
+basis = agnimhd.Basis(24, 12, 8)                 # n_zeta = 8 per field period
+eq, diffmat = agnimhd.from_desc("my_equilibrium.h5", basis, family=0)
+gamma2 = agnimhd.growth_rate(eq, diffmat)        # gamma^2 = -lambda > 0: unstable
+```
+
+`growth_rate` is solve mode and is not differentiable. The gradient comes from
+optimize mode, `jax.grad(agnimhd.growth_rate_of)(params, equilibrium_map,
+diffmat)`, with `equilibrium_map` the equilibrium code's map from its parameters
+to an `EquilibriumData`.
+
+Matrices too large for one GPU: `SolverConfig(eigensolver="dense_mg")` splits the
+dense matrix over all GPUs of a node with JAXMg (tested with `jaxmg==0.0.9`);
+182,784 unknowns in 8.5 minutes on four A100s. See the multi-GPU page of the
+documentation.
+
+Two shipped cases run without DESC: `python examples/cross_sections.py`. Other
+codes build an `EquilibriumData` from arrays, or load a saved one with
+`EquilibriumData.load("eq.npz")`. See the
+[documentation](https://rahulgaur104.github.io/AGNI/).
+
+## Install
+
+```bash
+pip install agnimhd               # jax, numpy, scipy, matfree
+pip install "agnimhd[hdf5,test]"  # HDF5 files and the test suite
+```
+
+Python 3.12. DESC is optional and only needed for `from_desc`. GPU, DESC and
+multi-GPU set-ups: [Installation](https://rahulgaur104.github.io/AGNI/installation/).
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pre-commit install
+pytest tests -q          # about 25 min on one CPU
+```
+
+Releasing: tag `vX.Y.Z` on `master` and publish a GitHub release of it. The
+`release` workflow builds the package from the tag and uploads it to PyPI, and
+Zenodo archives the release with a DOI. The version comes from the tag
+(`setuptools_scm`); nothing in the source is edited.
+
+## Reference
+
+R. Gaur, S. Patil, P. Gupta, D. Patch, T. Qian, *AGNI: A differentiable MHD
+stability solver & optimizer for magnetic confinement fusion devices*,
+[arXiv:2608.01750](https://arxiv.org/abs/2608.01750) (2026),
+doi:[10.48550/arXiv.2608.01750](https://doi.org/10.48550/arXiv.2608.01750).
+PDF of v3 in this repository: [docs/paper/agni_arxiv_2608.01750v3.pdf](docs/paper/agni_arxiv_2608.01750v3.pdf).
+
+```bibtex
+@article{gaur2026agni,
+  title   = {{AGNI}: A differentiable {MHD} stability solver \& optimizer for magnetic confinement fusion devices},
+  author  = {Gaur, Rahul and Patil, Sanket and Gupta, Prateek and Patch, Djin and Qian, Tony},
+  journal = {arXiv preprint arXiv:2608.01750},
+  year    = {2026},
+  doi     = {10.48550/arXiv.2608.01750}
+}
+```
+
+Originally developed inside [DESC](https://github.com/PlasmaControl/DESC)
+([PR #1893](https://github.com/PlasmaControl/DESC/pull/1893), on the
+differentiation matrices of PR #1789). MIT license.

agnimhd-0.1.0/pyproject.toml

@@ -0,0 +1,131 @@
+[build-system]
+requires = ["setuptools>=77", "setuptools_scm>=8"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "agnimhd"
+dynamic = ["version"]  # from the git tag, by setuptools_scm (as DESC uses versioneer)
+description = "AGNI: differentiable finite-n ideal MHD stability solver"
+readme = "README.md"
+# jax 0.11 declares Requires-Python >= 3.12, and 3.12 is the only interpreter
+# this package has been run on. Claiming 3.10 support would mean shipping
+# against an older jax nobody has tested.
+requires-python = ">=3.12"
+license = "MIT"
+license-files = ["LICENSE"]
+authors = [{ name = "Rahul Gaur", email = "rgaur@terpmail.umd.edu" }]
+keywords = ["plasma", "mhd", "stability", "stellarator", "tokamak", "jax"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Science/Research",
+    "Programming Language :: Python :: 3.12",
+    "Topic :: Scientific/Engineering :: Physics",
+]
+
+# EXACTLY these four. DESC appears nowhere -- not as a dependency, not as an
+# optional extra, not in the test extra. The dependency runs the other way:
+# DESC installs agnimhd. See docs/interface.md.
+#
+# `matfree` is REQUIRED, not optional: `from matfree import decomp, eig`
+# supplies the Lanczos tridiagonalization and the shift-invert eigensolve that
+# every matrix-free path uses. It was undeclared in the original DESC pull
+# request, so a clean install raised ModuleNotFoundError at solve time rather
+# than at install time.
+#
+# The 0.6.2 FLOOR IS A CORRECTNESS BOUND, not a preference. Before matfree PR
+# #288 the Arnoldi/Lanczos recurrence orthonormalized with the bilinear form
+# `Q.T @ Q` instead of `Q.conj().T @ Q`. On a real symmetric operator the two
+# coincide and every 3D case is unaffected; on the complex Hermitian operator
+# that `AssemblyConfig(axisym=True)` builds they do not, and the returned Ritz
+# vectors are wrong while the Ritz VALUES stay plausible -- so nothing raises
+# and only the Rayleigh residual shows it. Measured on the shipped QH case at
+# one zeta plane: lam_R = +9.713e-02 against a dense -2.660e-03, residual
+# 1.14e+03. Against 0.6.2 the same solve returns -2.660490917e-03.
+dependencies = [
+    "jax>=0.4.30",
+    "numpy>=1.24",
+    "scipy>=1.10",
+    "matfree>=0.6.2",
+]
+
+[project.optional-dependencies]
+# HDF5 serialization only. The native EquilibriumData format is .npz and needs
+# nothing beyond numpy; save_hdf5/load_hdf5 raise an ImportError naming h5py.
+hdf5 = ["h5py>=3.0"]
+# Plotting only. agnimhd.plotting's array-returning functions need nothing
+# beyond the four required dependencies; only the plot_* functions import
+# matplotlib, and they raise a named ImportError if it is absent.
+plot = ["matplotlib>=3.7"]
+# eigensolver="dense_mg": the dense matrix split over several GPUs.
+multigpu = ["jaxmg"]
+# pytest-split is REQUIRED, not optional: ci.yml runs the suite as four
+# `--splits 4 --group N` jobs, and without it every one of them errors on an
+# unrecognized argument.
+test = ["pytest>=7.0", "pytest-cov>=4.0", "pytest-split>=0.8", "pytest-mpl>=0.16"]
+benchmark = ["pytest-benchmark>=4.0"]
+dev = [
+    "agnimhd[test,hdf5,plot]",
+    "black>=24.0",
+    "flake8>=7.0",
+    "isort>=5.13",
+    "pre-commit>=3.5",
+]
+docs = [
+    "mkdocs>=1.5",
+    "mkdocs-material>=9.5",
+    "mkdocstrings[python]>=0.24",
+    "mkdocs-jupyter>=0.24",
+]
+notebooks = ["nbmake>=1.5", "ipykernel>=6"]
+
+[project.scripts]
+agnimhd = "agnimhd.cli:main"
+
+[project.urls]
+Homepage = "https://github.com/rahulgaur104/AGNI"
+Issues = "https://github.com/rahulgaur104/AGNI/issues"
+Paper = "https://arxiv.org/abs/2608.01750"
+Documentation = "https://rahulgaur104.github.io/AGNI/"
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[tool.setuptools_scm]
+# A release is a tag vX.Y.Z on master; between tags the version is X.Y.(Z+1).devN+g<sha>.
+version_file = "src/agnimhd/_version.py"
+
+[tool.setuptools.package-data]
+agnimhd = ["py.typed"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+markers = [
+    "slow: takes more than ~60 s. NOT deselected in CI -- the four-way\n    pytest-split matrix in ci.yml absorbs it. Deselecting would hide the\n    solver tests, which are the slow ones and the ones worth running.",
+    "regression: pins a measured number from a documented run.",
+]
+filterwarnings = ["error::DeprecationWarning:agnimhd.*"]
+
+[tool.coverage.run]
+# A PACKAGE name, not a file path. Handing coverage a .py path silently
+# measures nothing and reports "No data to report".
+source = ["agnimhd"]
+branch = true
+
+[tool.coverage.report]
+exclude_lines = [
+    "pragma: no cover",
+    "raise NotImplementedError",
+    "if TYPE_CHECKING:",
+]
+
+[tool.black]
+line-length = 88
+target-version = ["py310"]
+
+[tool.isort]
+profile = "black"
+line_length = 88
+
+[tool.flake8]
+max-line-length = 88
+extend-ignore = ["E203", "W503"]

agnimhd-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

agnimhd-0.1.0/src/agnimhd/__init__.py

@@ -0,0 +1,52 @@
+"""AGNI -- Analysis of Global Normal-modes in Ideal MHD.
+
+A differentiable finite-n ideal MHD stability solver.
+
+AGNI solves ideal MHD stability from a **variational principle**: it discretizes
+the energy functional rather than the force operator. That gives a generalized
+symmetric eigenvalue problem ``A x = lambda B x`` with ``B`` (the kinetic/mass
+matrix) symmetric positive definite. ``B`` is Cholesky-factored to reduce this to
+a standard symmetric eigenvalue problem. Everything agnimhd returns is the
+squared growth rate ``gamma^2 = -lambda`` of the most negative eigenvalue
+``lambda``: **positive means unstable**.
+
+The package depends on ``jax``, ``numpy``, ``scipy`` and ``matfree``.
+:func:`from_desc` imports DESC when it is called; nothing else does. Other
+equilibrium codes fill an :class:`EquilibriumData` directly; see
+``docs/interface.md``.
+"""
+
+from .adapters import from_desc
+from .basis import Basis, DiffMat
+from .config import AssemblyConfig, SolverConfig
+from .equilibrium import FORMAT_VERSION, EquilibriumData
+from .objective import (
+    eigenpair,
+    growth_rate,
+    growth_rate_and_grad,
+    growth_rate_of,
+    solve,
+)
+from .sources import load
+
+__all__ = [
+    "AssemblyConfig",
+    "Basis",
+    "DiffMat",
+    "EquilibriumData",
+    "FORMAT_VERSION",
+    "SolverConfig",
+    "eigenpair",
+    "from_desc",
+    "growth_rate",
+    "growth_rate_and_grad",
+    "growth_rate_of",
+    "load",
+    "solve",
+    "__version__",
+]
+
+try:  # written by setuptools_scm at build or install time
+    from ._version import __version__
+except ImportError:  # a bare checkout on sys.path
+    __version__ = "0+unknown"

agnimhd-0.1.0/src/agnimhd/_version.py

@@ -0,0 +1,24 @@
+# file generated by vcs-versioning
+# don't change, don't track in version control
+from __future__ import annotations
+
+__all__ = [
+    "__version__",
+    "__version_tuple__",
+    "version",
+    "version_tuple",
+    "__commit_id__",
+    "commit_id",
+]
+
+version: str
+__version__: str
+__version_tuple__: tuple[int | str, ...]
+version_tuple: tuple[int | str, ...]
+commit_id: str | None
+__commit_id__: str | None
+
+__version__ = version = '0.1.0'
+__version_tuple__ = version_tuple = (0, 1, 0)
+
+__commit_id__ = commit_id = 'g4d8e50872'

agnimhd-0.1.0/src/agnimhd/adapters/__init__.py

@@ -0,0 +1,10 @@
+"""Adapters from equilibrium codes to :class:`agnimhd.EquilibriumData`.
+
+Each adapter imports its equilibrium code lazily, inside the function, so the
+package itself stays free of those dependencies: ``agnimhd`` installs and runs
+without DESC; ``from_desc`` only works when DESC is importable.
+"""
+
+from .desc import from_desc
+
+__all__ = ["from_desc"]