add test to workflow

Author: FEO\preus

.github/workflows/build-package.yml

@@ -7,9 +7,6 @@ jobs:
   download_releases:
     runs-on: ubuntu-latest
     steps:
-      - name: checkout repo
-        uses: actions/checkout@v4
-
       - name: download release asset from deterministicGaussianSampling
         uses: robinraju/release-downloader@v1
         with:
@@ -18,6 +15,92 @@ jobs:
           fileName: '*.zip'
           out-file-path: 'deterministicGaussianSamplingLibs'
           token: ${{ secrets.GITHUB_TOKEN }}
+    
+      - name: ppload downloaded libs as artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: deterministicGaussianSampling-libs
+          path: deterministicGaussianSamplingLibs
+
+  test:
+    needs: download_releases
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest] #, macos-latest]
+        python-version: [3.9, 3.11]
+    runs-on: ${{ matrix.os }}
+
+    steps:
+      - name: checkout repo
+        uses: actions/checkout@v4
+
+      - name: set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: get deterministicGaussianSampling libs
+        uses: actions/download-artifact@v4
+        with:
+          name: deterministicGaussianSampling-libs
+          path: external_libs
+
+      - name: extract ZIP on Unix
+        if: ${{ runner.os == 'Linux' }}
+        run: |
+            unzip external_libs/linux.zip -d external_libs_unpacked
+            mkdir -p src/deterministic_gaussian_sampling/lib/linux/bin
+            cp -r external_libs_unpacked/lib/* src/deterministic_gaussian_sampling/lib/linux/bin
+
+      - name: extract ZIP on Windows
+        if: ${{ runner.os == 'Windows' }}
+        shell: pwsh
+        run: |
+            Expand-Archive -Path external_libs/windows.zip -DestinationPath external_libs_unpacked -Force
+            New-Item -ItemType Directory -Path src\deterministic_gaussian_sampling\lib\windows\bin -Force
+            Copy-Item external_libs_unpacked\bin\* src\deterministic_gaussian_sampling\lib\windows\bin -Recurse -Force
+
+      - name: install package
+        run: pip install -e .
+
+      - name: run tests
+        run: pytest -v
+
+
+
+name: Download OS‑specific release and test
+
+on:
+  push:
+  workflow_dispatch:
+
+jobs:
+  test:
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest]
+    runs-on: ${{ matrix.os }}
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Download release asset
+        uses: robinraju/release-downloader@v1
+        with:
+          repository: 'OWNER/REPO'                # replace with actual
+          latest: true
+          fileName: ${{ matrix.os == 'windows-latest' && 'windows.zip' || 'linux.zip' }}
+          out-file-path: external_libs
+          token: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract archive
+        if: runner.os != 'Windows'
+        run: unzip external_libs/linux.zip -d external_libs_unpacked
+
+      - name: Extract archive on Windows
+        if: runner.os == 'Windows'
+        shell: pwsh
+        run: Expand-Archive -Path external_libs/windows.zip -DestinationPath external_libs_unpacked -Force
+
+      # ... then run your tests etc.
 
-      - name: list extracted files
-        run: ls -R deterministicGaussianSamplingLibs

src/deterministic_gaussian_sampling/dll_handling/dynamic_dll_loading.py

@@ -1,5 +1,6 @@
-from pathlib import Path
 import ctypes
+import platform
+from pathlib import Path
 
 import deterministic_gaussian_sampling.type_wrapper.ctypes_wrapper as ctypes_wrapper
 
@@ -113,6 +114,17 @@ def _setup_ctypes_dll(cdll: ctypes.CDLL) -> ctypes.CDLL:
 
 def load_dll() -> ctypes.CDLL:
     package_root = Path(__file__).resolve().parent.parent
-    dll_path = package_root / "lib" / "windows" / "bin" / "libapproxLCD.dll"
 
-    return _setup_ctypes_dll(ctypes.CDLL(dll_path))
+    system = platform.system()
+    if system == "Windows":
+        dll_rel = Path("lib") / "windows" / "bin" / "libapproxLCD.dll"
+    elif system == "Linux":
+        dll_rel = Path("lib") / "linux" / "bin" / "libapproxLCD.so"
+    elif system == "Darwin":
+        dll_rel = Path("lib") / "macos" / "bin" / "libapproxLCD.dylib"
+    else:
+        raise RuntimeError(f"Unsupported OS: {system}")
+    
+    dll_path = package_root / dll_rel
+
+    return _setup_ctypes_dll(ctypes.CDLL(dll_path))

tests/test_dirac_to_dirac.py

@@ -0,0 +1,79 @@
+import deterministic_gaussian_sampling
+import numpy as np
+from scipy import stats
+import pytest
+
+def compare_stats(full: np.ndarray, reduced: np.ndarray, rtol=1e-1, atol=1e-2):
+    """
+    Compare the statistics of two datasets.
+    Raises AssertionError if they differ more than allowed tolerances.
+    """
+    assert full.ndim == 2 and reduced.ndim == 2
+    assert full.shape[1] == reduced.shape[1]  # same dimensionality
+
+    # Compute mean, variance, std
+    stats_full = {
+        'mean': full.mean(axis=0),
+        'var': full.var(axis=0, ddof=1),
+        'std': full.std(axis=0, ddof=1)
+    }
+    stats_red = {
+        'mean': reduced.mean(axis=0),
+        'var': reduced.var(axis=0, ddof=1),
+        'std': reduced.std(axis=0, ddof=1)
+    }
+
+    # Assert that means and stds are approximately equal
+    np.testing.assert_allclose(stats_full['mean'], stats_red['mean'], rtol=rtol, atol=atol)
+    np.testing.assert_allclose(stats_full['std'], stats_red['std'], rtol=rtol, atol=atol)
+
+    # Optionally: per-dimension Welch's t-test
+    for dim in range(full.shape[1]):
+        t_stat, p_val = stats.ttest_ind(full[:, dim], reduced[:, dim], equal_var=False)
+        # Expect p-value > 0.05 (no significant difference)
+        assert p_val > 0.05, f"Significant difference in dimension {dim}: p={p_val:.3f}"
+
+
+@pytest.mark.parametrize("M,L,N", [
+    (100, 3, 2),
+    (100, 5, 8),
+    (250, 5, 3),
+    (250, 20, 9),
+    (500, 5, 3),
+    # (500, 75, 20),
+    # (750, 5, 3),
+    # (750, 50, 12),
+    # (1000, 100, 3),
+    # (1000, 100, 15),
+])
+def test_dirac_approximation_simple(M, L, N, seed=42):
+    np.random.seed(seed)
+    x = np.random.rand(L, N)
+    y = np.random.rand(M, N)
+    d2dApprox = deterministic_gaussian_sampling.DiracToDiracApproximation()
+    result = d2dApprox.approximate_double(y, M, L, N, x)
+
+    # Compare statistics between original and reduced sets
+    compare_stats(x, result.x)
+
+@pytest.mark.parametrize("M,L,N", [
+    (100, 3, 2),
+    (100, 5, 8),
+    (250, 5, 3),
+    (250, 20, 9),
+    (500, 5, 3),
+    # (500, 75, 20),
+    # (750, 5, 3),
+    # (750, 50, 12),
+    # (1000, 100, 3),
+    # (1000, 100, 15),
+])
+def test_dirac_approximation_threaded(M, L, N, seed=42):
+    np.random.seed(seed)
+    x = np.random.rand(L, N)
+    y = np.random.rand(M, N)
+    d2dApprox = deterministic_gaussian_sampling.DiracToDiracApproximation()
+    result = d2dApprox.approximate_thread_double(y, M, L, N, x)
+
+    # Compare statistics between original and reduced sets
+    compare_stats(x, result.x)