test(types): Test SpatialImage API type inference

Author: effigies

pyproject.toml

@@ -76,6 +76,7 @@ test = [
   "pytest-httpserver >=1.0.7",
   "pytest-xdist >=3.5",
   "coverage[toml]>=7.2",
+  "pytest-mypy-testing>=0.1.3",
 ]
 # Remaining: Simpler to centralize in tox
 dev = ["tox"]

tests/ruff.toml

@@ -0,0 +1 @@
+line-length = 200

tests/typing/test_spatialimage_api.py

@@ -0,0 +1,80 @@
+import typing as ty
+
+import numpy as np
+import pytest
+
+from nibabel import AnalyzeImage, Spm99AnalyzeImage, Spm2AnalyzeImage, Nifti1Image, Nifti2Image, MGHImage
+from nibabel.spatialimages import SpatialImage
+
+if ty.TYPE_CHECKING:
+    from typing import reveal_type
+else:
+
+    def reveal_type(x: ty.Any) -> None:
+        pass
+
+
+@pytest.mark.mypy_testing
+def test_affine_tracking() -> None:
+    img_with_affine = SpatialImage(np.empty((5, 5, 5)), np.eye(4))
+    img_without_affine = SpatialImage(np.empty((5, 5, 5)), None)
+
+    reveal_type(img_with_affine)  # R: nibabel.spatialimages.SpatialImage[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(img_without_affine)  # R: nibabel.spatialimages.SpatialImage[None]
+
+
+@pytest.mark.mypy_testing
+def test_SpatialImageAPI() -> None:
+    img = SpatialImage(np.empty((5, 5, 5)), np.eye(4))
+
+    # Affine
+    reveal_type(img.affine)  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+    reveal_type(SpatialImage(np.empty((5, 5, 5)), None).affine)  # R: None
+
+    # Data
+    reveal_type(img.dataobj)  # R: nibabel.arrayproxy.ArrayLike
+    reveal_type(img.get_fdata())  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+    reveal_type(img.get_fdata(dtype=np.float32))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.floating[numpy._typing._nbit_base._32Bit]]]
+    reveal_type(img.get_fdata(dtype=np.float64))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+    reveal_type(img.get_fdata(dtype=np.dtype(np.float32)))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.floating[numpy._typing._nbit_base._32Bit]]]
+    reveal_type(img.get_fdata(dtype=np.dtype(np.float64)))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+    reveal_type(img.get_fdata(dtype=np.dtype("f4")))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.floating[numpy._typing._nbit_base._32Bit]]]
+    reveal_type(img.get_fdata(dtype=np.dtype("f8")))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+    reveal_type(img.get_fdata(dtype="f4"))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.floating[numpy._typing._nbit_base._32Bit]]]
+    reveal_type(img.get_fdata(dtype="f8"))  # R: numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]
+
+    # Indirect header
+    reveal_type(img.shape)  # R: builtins.tuple[builtins.int, ...]
+    reveal_type(img.ndim)  # R: builtins.int
+
+    # SpatialHeader fields
+    reveal_type(img.header.get_data_dtype())  # R: numpy.dtype[Any]
+    reveal_type(img.header.get_data_shape())  # R: builtins.tuple[builtins.int, ...]
+    reveal_type(img.header.get_zooms())  # R: builtins.tuple[builtins.float, ...]
+
+
+@pytest.mark.mypy_testing
+def test_image_and_header_types() -> None:
+    analyze_img = AnalyzeImage(np.empty((5, 5, 5)), np.eye(4))
+    reveal_type(analyze_img)  # R: nibabel.analyze.AnalyzeImage[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(analyze_img.header)  # R: nibabel.analyze.AnalyzeHeader
+
+    spm99_img = Spm99AnalyzeImage(np.empty((5, 5, 5)), np.eye(4))
+    reveal_type(spm99_img)  # R: nibabel.spm99analyze.Spm99AnalyzeImage[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(spm99_img.header)  # R: nibabel.spm99analyze.Spm99AnalyzeHeader
+
+    spm2_img = Spm2AnalyzeImage(np.empty((5, 5, 5)), np.eye(4))
+    reveal_type(spm2_img)  # R: nibabel.spm2analyze.Spm2AnalyzeImage[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(spm2_img.header)  # R: nibabel.spm2analyze.Spm2AnalyzeHeader
+
+    ni1_img = Nifti1Image(np.empty((5, 5, 5)), np.eye(4))
+    reveal_type(ni1_img)  # R: nibabel.nifti1.Nifti1Image[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(ni1_img.header)  # R: nibabel.nifti1.Nifti1Header
+
+    ni2_img = Nifti2Image(np.empty((5, 5, 5)), np.eye(4))
+    reveal_type(ni2_img)  # R: nibabel.nifti2.Nifti2Image[numpy.ndarray[builtins.tuple[builtins.int, ...], numpy.dtype[numpy.float64]]]
+    reveal_type(ni2_img.header)  # R: nibabel.nifti2.Nifti2Header
+
+    mgh_img = MGHImage(np.empty((5, 5, 5), dtype=np.float32), np.eye(4))
+    reveal_type(mgh_img)  # R: nibabel.freesurfer.mghformat.MGHImage
+    reveal_type(mgh_img.header)  # R: nibabel.freesurfer.mghformat.MGHHeader

tox.ini

@@ -17,6 +17,7 @@ envlist =
   doctest
   style
   typecheck
+  type-inference
 skip_missing_interpreters = true
 
 # Configuration that allows us to split tests across GitHub runners effectively
@@ -179,6 +180,18 @@ skip_install = true
 commands =
   mypy nibabel
 
+[testenv:type-inference]
+description = Check type inference
+labels = test
+deps =
+  pytest-mypy-testing @ git+https://github.com/effigies/pytest-mypy-testing@rf/global-mypy-run
+commands =
+  python -m pytest \
+    --cov tests --cov nibabel --cov-report xml:cov.xml \
+    --junitxml test-results.xml \
+    --durations=20 --durations-min=1.0 \
+    tests/ {posargs:-n auto}
+
 [testenv:build{,-strict}]
 labels =
   check