Merge branch 'main' into tjlane/consistent-map-indices

test/conftest.py

@@ -1,5 +1,6 @@
 from pathlib import Path
 
+import numpy as np
 import pytest
 
 from meteor.testing import check_test_file_exists
@@ -25,3 +26,8 @@ def testing_mtz_file(data_dir: Path) -> Path:
     path = data_dir / "scaled-test-data.mtz"
     check_test_file_exists(path)
     return path
+
+
+@pytest.fixture(scope="session")
+def np_rng() -> np.random.Generator:
+    return np.random.default_rng(seed=0)

test/unit/conftest.py

@@ -14,8 +14,6 @@
 SPACE_GROUP = gemmi.find_spacegroup_by_name("P1")
 CARBON1_POSITION = (5.0, 5.0, 5.0)
 
-NP_RNG = np.random.default_rng()
-
 
 @pytest.fixture
 def tv_denoise_result_source_data() -> dict:
@@ -39,10 +37,10 @@ def test_map_columns() -> MapColumns:
     )
 
 
-def single_atom_map_coefficients(*, noise_sigma: float) -> Map:
+def single_atom_map_coefficients(*, noise_sigma: float, np_rng: np.random.Generator) -> Map:
     density_map = single_carbon_density(CARBON1_POSITION, SPACE_GROUP, UNIT_CELL, RESOLUTION)
     density_array = np.array(density_map.grid)
-    grid_values = density_array + noise_sigma * NP_RNG.normal(size=density_array.shape)
+    grid_values = density_array + noise_sigma * np_rng.normal(size=density_array.shape)
     ccp4_map = numpy_array_to_map(grid_values, spacegroup=SPACE_GROUP, cell=UNIT_CELL)
 
     map_coefficients = Map.from_ccp4_map(ccp4_map=ccp4_map, high_resolution_limit=RESOLUTION)
@@ -60,22 +58,22 @@ def ccp4_map() -> gemmi.Ccp4Map:
 
 
 @pytest.fixture
-def noise_free_map() -> Map:
-    return single_atom_map_coefficients(noise_sigma=0.0)
+def noise_free_map(np_rng: np.random.Generator) -> Map:
+    return single_atom_map_coefficients(noise_sigma=0.0, np_rng=np_rng)
 
 
 @pytest.fixture
-def noisy_map() -> Map:
-    return single_atom_map_coefficients(noise_sigma=0.03)
+def noisy_map(np_rng: np.random.Generator) -> Map:
+    return single_atom_map_coefficients(noise_sigma=0.03, np_rng=np_rng)
 
 
 @pytest.fixture
-def very_noisy_map() -> Map:
-    return single_atom_map_coefficients(noise_sigma=1.0)
+def very_noisy_map(np_rng: np.random.Generator) -> Map:
+    return single_atom_map_coefficients(noise_sigma=1.0, np_rng=np_rng)
 
 
 @pytest.fixture
-def random_difference_map(test_map_columns: MapColumns) -> Map:
+def random_difference_map(test_map_columns: MapColumns, np_rng: np.random.Generator) -> Map:
     hall = rs.utils.generate_reciprocal_asu(UNIT_CELL, SPACE_GROUP, RESOLUTION, anomalous=False)
     sigma = 1.0
 
@@ -87,8 +85,8 @@ def random_difference_map(test_map_columns: MapColumns) -> Map:
             "H": h,
             "K": k,
             "L": l,
-            test_map_columns.amplitude: sigma * NP_RNG.normal(size=number_of_reflections),
-            test_map_columns.phase: NP_RNG.uniform(-180, 180, size=number_of_reflections),
+            test_map_columns.amplitude: sigma * np_rng.normal(size=number_of_reflections),
+            test_map_columns.phase: np_rng.uniform(-180, 180, size=number_of_reflections),
         },
         spacegroup=SPACE_GROUP,
         cell=UNIT_CELL,

test/unit/test_iterative.py

@@ -16,8 +16,6 @@
 from meteor.tv import TvDenoiseResult
 from meteor.validate import negentropy
 
-NP_RNG = np.random.default_rng()
-
 
 def map_norm(map1: gemmi.Ccp4Map, map2: gemmi.Ccp4Map) -> float:
     diff = np.array(map1.grid) - np.array(map2.grid)
@@ -46,10 +44,10 @@ def normalized_rms(x: np.ndarray, y: np.ndarray) -> float:
 
 
 @pytest.mark.parametrize("scalar", [0.01, 1.0, 2.0, 100.0])
-def test_projected_derivative(scalar: float) -> None:
+def test_projected_derivative(scalar: float, np_rng: np.random.Generator) -> None:
     n = 16
-    native = NP_RNG.normal(size=n) + 1j * NP_RNG.normal(size=n)
-    derivative = NP_RNG.normal(size=n) + 1j * NP_RNG.normal(size=n)
+    native = np_rng.normal(size=n) + 1j * np_rng.normal(size=n)
+    derivative = np_rng.normal(size=n) + 1j * np_rng.normal(size=n)
     difference = derivative - native
 
     # ensure the projection removes a scalar multiple of the native & difference
@@ -63,13 +61,13 @@ def test_projected_derivative(scalar: float) -> None:
     np.testing.assert_allclose(proj_derivative, derivative)
 
 
-def test_complex_derivative_from_iterative_tv() -> None:
+def test_complex_derivative_from_iterative_tv(np_rng: np.random.Generator) -> None:
     test_image = binary_blobs(length=64)  # type: ignore[no-untyped-call]
 
     constant_image = np.ones_like(test_image) / 2.0
     constant_image_ft = np.fft.fftn(constant_image)
 
-    test_image_noisy = test_image + 0.2 * NP_RNG.normal(size=test_image.shape)
+    test_image_noisy = test_image + 0.2 * np_rng.normal(size=test_image.shape)
     test_image_noisy_ft = np.fft.fftn(test_image_noisy)
 
     denoised_derivative, _ = _complex_derivative_from_iterative_tv(

test/unit/test_utils.py

@@ -8,8 +8,6 @@
 from meteor.rsmap import Map
 from meteor.testing import MapColumns
 
-NP_RNG = np.random.default_rng()
-
 
 def omit_nones_in_list(input_list: list) -> list:
     return [x for x in input_list if x]

test/unit/test_validate.py

@@ -4,26 +4,24 @@
 from meteor import validate
 from meteor.rsmap import Map
 
-NP_RNG = np.random.default_rng()
-
 
 def parabolic_objective(x: float) -> float:
     # has a maximum of y = 0 at x = 1
     return -(x**2) + 2 * x - 1
 
 
-def test_negentropy_gaussian() -> None:
+def test_negentropy_gaussian(np_rng: np.random.Generator) -> None:
     n_samples = 10000
-    samples = NP_RNG.normal(size=n_samples)
+    samples = np_rng.normal(size=n_samples)
     negentropy = validate.negentropy(samples)
 
     # negentropy should be zero for a Gaussian sample
     assert np.abs(negentropy) < 1e-2
 
 
-def test_negentropy_uniform() -> None:
+def test_negentropy_uniform(np_rng: np.random.Generator) -> None:
     n_samples = 1000000
-    samples = NP_RNG.uniform(size=n_samples)
+    samples = np_rng.uniform(size=n_samples)
     negentropy = validate.negentropy(samples)
 
     uniform_negentropy = (1.0 / 2.0) * np.log(np.pi * np.exp(1) / 6.0)