Draft some tests for the Neighbors class

Author: Marius1311

tests/conftest.py

@@ -12,3 +12,11 @@ def sample_distances():
 def sample_indices():
     # 3 samples, 2 neighbors each
     return np.array([[0, 1], [1, 2], [2, 0]], dtype=np.int64)
+
+
+@pytest.fixture
+def small_data():
+    # 5 points in 2D, easy to check neighbors
+    x = np.array([[0, 0], [1, 0], [0, 1], [1, 1], [2, 2]], dtype=np.float64)
+    y = np.array([[0, 0], [1, 0], [0, 1], [1, 1], [2, 2]], dtype=np.float64)
+    return x, y

tests/test_neighbors.py

@@ -0,0 +1,34 @@
+import numpy as np
+
+from cellmapper.knn import Neighbors
+
+
+def assert_adjacency_equal(neigh1, neigh2, attrs=("xx", "yy", "xy", "yx")):
+    for attr in attrs:
+        # Direct attribute access instead of getattr for Ruff compliance
+        adj1 = getattr(neigh1, attr).boolean_adjacency().toarray()
+        adj2 = getattr(neigh2, attr).boolean_adjacency().toarray()
+        assert np.allclose(adj1, adj2), f"Adjacency matrices ({attr}) differ between methods"
+
+
+def test_neighbors_sklearn_vs_pynndescent(small_data):
+    x, y = small_data
+    n_neighbors = 3
+    # sklearn
+    neigh_skl = Neighbors(x, y)
+    neigh_skl.compute_neighbors(n_neighbors=n_neighbors, method="sklearn")
+    # pynndescent
+    neigh_pynn = Neighbors(x, y)
+    neigh_pynn.compute_neighbors(n_neighbors=n_neighbors, method="pynndescent")
+    # Compare adjacency matrices
+    assert_adjacency_equal(neigh_skl, neigh_pynn)
+
+
+def test_neighbors_repr(small_data):
+    x, y = small_data
+    neigh = Neighbors(x, y)
+    r = repr(neigh)
+    assert "Neighbors(" in r and "xrep_shape" in r and "yrep_shape" in r
+    neigh.compute_neighbors(n_neighbors=2, method="sklearn")
+    r2 = repr(neigh)
+    assert "xx=True" in r2 and "yy=True" in r2 and "xy=True" in r2 and "yx=True" in r2

tests/test_neighbors_results.py

@@ -29,13 +29,15 @@ def test_knn_graph_distances(sample_distances, sample_indices):
     assert np.allclose(mat.diagonal(), 0)
 
 
-def test_knn_graph_connectivities_gaussian(sample_distances, sample_indices):
+@pytest.mark.parametrize("kernel", ["gaussian", "scarches", "random", "inverse_distance"])
+def test_knn_graph_connectivities_kernels(sample_distances, sample_indices, kernel):
     nr = NeighborsResults(distances=sample_distances, indices=sample_indices)
-    mat = nr.knn_graph_connectivities(kernel="gaussian")
+    mat = nr.knn_graph_connectivities(kernel=kernel)
     assert isinstance(mat, csr_matrix)
     assert mat.shape == (3, 3)
-    # All values should be in (0, 1]
-    assert np.all((mat.data > 0) & (mat.data <= 1))
+    # All values should be > 0 for all kernels except 'random', which can be 0
+    if kernel != "random":
+        assert np.all(mat.data > 0)
 
 
 def test_boolean_adjacency(sample_distances, sample_indices):