pyCERR/pyproject.toml at main · cerr/pyCERR · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
[build-system]
requires = ["setuptools>=64", "setuptools_scm>=8"]
build-backend = "setuptools.build_meta"

[project]
name = "pycerr"
dynamic = ["version"]
authors = [
  { name="Aditya P. Apte", email="aptea@mskcc.org" },
  { name="Aditi Iyer", email="iyera@mskcc.org" },
  { name="Eve LoCastro", email="locastre@mskcc.org" },
  { name="Joseph O. Deasy", email="deasyj@mskcc.org" }
]
description = "A Computational Environment for Radiological Research"
readme = "README.md"
requires-python = ">=3.8"
classifiers = [
    "Programming Language :: Python :: 3",
    "License :: OSI Approved :: GNU General Public License (GPL)",
    "Operating System :: OS Independent",
]
dependencies = [
    'scipy',
    'pydicom',
    # Decoders for compressed DICOM pixel data (JPEG Lossless / JPEG-LS /
    # JPEG 2000). pydicom needs one of these plugins to decode `pixel_array`
    # for compressed transfer syntaxes; without it, reading such scans raises
    # "Unable to decompress ... pixel data". python-gdcm is an alternative
    # single-package decoder (also adds RLE): `pip install python-gdcm`.
    'pylibjpeg',
    'pylibjpeg-libjpeg',
    'pylibjpeg-openjpeg',
    'python-dateutil',
    'h5py',
    'pandas',
    'SimpleITK',
    'nibabel',
    'scikit-learn',
    'scikit-image',
    'PyWavelets',
    'shapelysmooth',
    'matplotlib >= 3.5',
    'ipywidgets',
    'surface-distance-based-measures',
    'itk-strain',
    # JIT-compiled, multi-threaded kernels for the urOMT solver
    # (cerr.uromt.kernels): the trilinear gather/scatter/derivative operators
    # that dominate its Gauss-Newton CG matvec. ~90x on the derivative operator
    # versus the vectorized numpy form.
    'numba',
    # Desktop Qt viewer (cerr.viewer.pycerr_gui): PyQt5 is the Qt binding;
    # pyvista/pyvistaqt drive the GPU-accelerated 3D view (vtk is pulled in
    # transitively by pyvista).
    'PyQt5',
    'pyvista',
    'pyvistaqt'
]

[project.urls]
"Homepage" = "https://github.com/cerr/pyCERR"
"Bug Tracker" = "https://github.com/cerr/pyCERR/issues"

[project.optional-dependencies]
napari = [
    "napari[all]",
    "napari-animation"
]
# NOTE: model_installer is only distributed as a git repo, and PyPI rejects
# direct URL (git+) dependencies in package metadata. So it is NOT declared as
# an extra here; install it alongside pyCERR when needed with:
#   pip install "model_installer @ git+https://github.com/cerr/model_installer.git"
ants = [
    "antspyx",
]
# Optional GPU acceleration for the urOMT solver (cerr.uromt.gpu), enabled with
# the `gpu` setting. Without it urOMT runs on the CPU (numba, multi-threaded)
# and a `gpu=1` request warns and falls back.
#
# This extra targets a CUDA 13.x driver. For CUDA 12.x install `cupy-cuda12x`
# plus the `nvidia-*-cu12` library wheels instead; for CUDA 11.x,
# `cupy-cuda11x`. Check what your driver supports with:
#   python -c "import ctypes;v=ctypes.c_int();ctypes.WinDLL('nvcuda.dll').cuDriverGetVersion(ctypes.byref(v));print(v.value)"
#
# The nvidia-* wheels are NOT optional on a machine that also has an older CUDA
# toolkit installed: cupy resolves its libraries through cuda-pathfinder, which
# prefers site-packages over CUDA_PATH/PATH, so shipping them here is what stops
# a stale toolkit's NVRTC being loaded (see cerr/uromt/gpu.py).
gpu = [
    "cupy-cuda13x",
    "nvidia-cuda-nvrtc",
    "nvidia-cuda-runtime",
    "nvidia-cufft",
    "nvidia-cublas",
    "nvidia-cusparse",
    "nvidia-cusolver",
    "nvidia-curand",
]

[tool.setuptools]
include-package-data = true

[tool.setuptools.exclude-package-data]
# Test-only datasets (mirrors MANIFEST.in) - keep them out of the wheel.
# The notebook-facing datasets (radiomics_phantom_dicom, sample_ct,
# radiomics_settings, convolutional_filter_settings) are retained.
"cerr.datasets" = [
    "ibsi_radiomics_dicom/*",
    "ibsi_radiomics_dicom/**/*",
    "reference_values_for_tests/*",
    "reference_values_for_tests/**/*",
    # head_and_neck is downloaded on demand (fetch_sample_data); keep it out of
    # the wheel. dosimetric_model_test_data stays bundled (NTCP notebook).
    "sample_ct/head_and_neck/*",
    "sample_ct/head_and_neck/**/*",
]

[tool.setuptools_scm]
version_file = "cerr/_version.py"
# Drop the +g<sha> local segment so dev builds are uploadable to TestPyPI
# (PyPI rejects local versions). A clean tag still yields e.g. 0.7.0.
local_scheme = "no-local-version"

[tool.ruff.lint]
select = ["NPY201"]