(fix): fix igraph on windows (#3745)

Author: flying-sheep

.github/workflows/ci.yml

@@ -102,7 +102,7 @@ jobs:
           file: test-data/test-results.xml
 
       - name: Publish debug artifacts
-        if: matrix.env.test-type == 'coverage'
+        if: ${{ !cancelled() }}
         uses: actions/upload-artifact@v4
         with:
           name: debug-data-${{ matrix.env.name }}

docs/release-notes/3745.bugfix.md

@@ -0,0 +1 @@
+Fix {func}`~scanpy.tl.leiden` with igraph backend on Windows {smaller}`P Angerer`

src/scanpy/_utils/random.py

@@ -38,14 +38,20 @@
 
 
 class _RNGIgraph:
-    """Random number generator for ipgraph so global seed is not changed.
+    """Random number generator for igraph so global seed is not changed.
 
     See :func:`igraph.set_random_number_generator` for the requirements.
     """
 
     def __init__(self, random_state: int | np.random.RandomState = 0) -> None:
         self._rng = check_random_state(random_state)
 
+    def getrandbits(self, k: int) -> int:
+        return self._rng.tomaxint() & ((1 << k) - 1)
+
+    def randint(self, a: int, b: int) -> int:
+        return self._rng.randint(a, b + 1)
+
     def __getattr__(self, attr: str):
         return getattr(self._rng, "normal" if attr == "gauss" else attr)
 

tests/_images/heatmap_var_as_dict/expected.png

@@ -115,7 +115,7 @@ def test_pbmc3k(image_comparer):  # noqa: PLR0915
     sc.tl.leiden(
         adata,
         resolution=0.9,
-        random_state=0,
+        random_state=1,
         directed=False,
         n_iterations=2,
         flavor="igraph",
@@ -130,22 +130,18 @@ def test_pbmc3k(image_comparer):  # noqa: PLR0915
     # Due to incosistency with our test runner vs local, these clusters need to
     # be pre-annotated as the numbers for each cluster are not consistent.
     marker_genes = [
-        "RP11-18H21.1",
-        "GZMK",
-        "CD79A",
-        "FCGR3A",
-        "GNLY",
-        "S100A8",
-        "FCER1A",
-        "PPBP",
+        *["RP11-18H21.1", "GZMK", "CD79A", "FCGR3A"],
+        *["GNLY", "S100A8", "FCER1A", "PPBP"],
     ]
-    new_labels = ["0", "1", "2", "3", "4", "5", "6", "7"]
     data_df = adata[:, marker_genes].to_df()
     data_df["leiden"] = adata.obs["leiden"]
     max_idxs = data_df.groupby("leiden", observed=True).mean().idxmax()
-    leiden_relabel = {}
-    for marker_gene, new_label in zip(marker_genes, new_labels, strict=True):
-        leiden_relabel[max_idxs[marker_gene]] = new_label
+    assert not max_idxs[marker_genes][
+        max_idxs[marker_genes].duplicated(keep=False)
+    ].tolist(), "Not all marker genes are unique per cluster"
+    leiden_relabel = {
+        max_idxs[marker_gene]: str(i) for i, marker_gene in enumerate(marker_genes)
+    }
     adata.obs["leiden_old"] = adata.obs["leiden"].copy()
     adata.rename_categories(
         "leiden", [leiden_relabel[key] for key in sorted(leiden_relabel.keys())]
@@ -175,14 +171,8 @@ def test_pbmc3k(image_comparer):  # noqa: PLR0915
     # save_and_compare_images('rank_genes_groups_4')
 
     new_cluster_names = [
-        "CD4 T cells",
-        "CD8 T cells",
-        "B cells",
-        "NK cells",
-        "FCGR3A+ Monocytes",
-        "CD14+ Monocytes",
-        "Dendritic cells",
-        "Megakaryocytes",
+        *["CD4 T cells", "CD8 T cells", "B cells", "NK cells"],
+        *["FCGR3A+ Monocytes", "CD14+ Monocytes", "Dendritic cells", "Megakaryocytes"],
     ]
     adata.rename_categories("leiden", new_cluster_names)
 

tests/notebooks/test_pbmc3k.py

@@ -2,6 +2,7 @@
 
 from functools import partial
 
+import pandas as pd
 import pytest
 from sklearn.metrics.cluster import normalized_mutual_info_score
 
@@ -71,8 +72,8 @@ def test_leiden_random_state(adata_neighbors, flavor):
         directed=is_leiden_alg,
         n_iterations=n_iterations,
     )
-    assert (adata_1.obs["leiden"] == adata_1_again.obs["leiden"]).all()
-    assert (adata_2.obs["leiden"] != adata_1_again.obs["leiden"]).any()
+    pd.testing.assert_series_equal(adata_1.obs["leiden"], adata_1_again.obs["leiden"])
+    assert not adata_2.obs["leiden"].equals(adata_1_again.obs["leiden"])
 
 
 @needs.igraph

tests/test_clustering.py

@@ -28,6 +28,7 @@
 
 if TYPE_CHECKING:
     from collections.abc import Callable
+    from typing import Any
 
     from matplotlib.axes import Axes
 
@@ -58,71 +59,53 @@ def test_highest_expr_genes(image_comparer, col, layer):
 
 
 @needs.leidenalg
-def test_heatmap(image_comparer):
+@pytest.mark.parametrize(
+    ("params", "key"),
+    [
+        pytest.param({}, "heatmap", id="default"),
+        pytest.param(
+            dict(swap_axes=True, figsize=(10, 3), cmap="YlGnBu"),
+            "heatmap_swap_axes",
+            id="swap",
+        ),
+        pytest.param(
+            dict(
+                groupby="numeric_value",
+                num_categories=4,
+                figsize=(4.5, 5),
+                dendrogram=False,
+            ),
+            "heatmap2",
+            id="numeric",
+        ),
+        pytest.param(
+            dict(standard_scale="var", layer="test"),
+            "heatmap_std_scale_var",
+            id="std_scale=var",
+        ),
+        pytest.param(
+            dict(standard_scale="obs"),
+            "heatmap_std_scale_obs",
+            id="std_scale=obs",
+        ),
+    ],
+)
+def test_heatmap(image_comparer, params: dict[str, Any], key: str) -> None:
     save_and_compare_images = partial(image_comparer, ROOT, tol=15)
 
     adata = krumsiek11()
-    sc.pl.heatmap(
-        adata, adata.var_names, "cell_type", use_raw=False, show=False, dendrogram=True
-    )
-    save_and_compare_images("heatmap")
-
-    # test swap axes
-    sc.pl.heatmap(
-        adata,
-        adata.var_names,
-        "cell_type",
-        use_raw=False,
-        show=False,
-        dendrogram=True,
-        swap_axes=True,
-        figsize=(10, 3),
-        cmap="YlGnBu",
-    )
-    save_and_compare_images("heatmap_swap_axes")
-
-    # test heatmap numeric column():
-
-    # set as numeric column the vales for the first gene on the matrix
     adata.obs["numeric_value"] = adata.X[:, 0]
-    sc.pl.heatmap(
-        adata,
-        adata.var_names,
-        "numeric_value",
-        use_raw=False,
-        num_categories=4,
-        figsize=(4.5, 5),
-        show=False,
-    )
-    save_and_compare_images("heatmap2")
-
-    # test var/obs standardization and layer
     adata.layers["test"] = -1 * adata.X.copy()
-    sc.pl.heatmap(
-        adata,
-        adata.var_names,
-        "cell_type",
-        use_raw=False,
-        dendrogram=True,
-        show=False,
-        standard_scale="var",
-        layer="test",
-    )
-    save_and_compare_images("heatmap_std_scale_var")
 
-    # test standard_scale_obs
-    sc.pl.heatmap(
-        adata,
-        adata.var_names,
-        "cell_type",
-        use_raw=False,
-        dendrogram=True,
-        show=False,
-        standard_scale="obs",
-    )
-    save_and_compare_images("heatmap_std_scale_obs")
+    params = dict(groupby="cell_type", dendrogram=True) | params
+    sc.pl.heatmap(adata, adata.var_names, **params, use_raw=False, show=False)
+    save_and_compare_images(key)
+
+
+@needs.leidenalg
+def test_heatmap_var_as_dict(image_comparer) -> None:
+    save_and_compare_images = partial(image_comparer, ROOT, tol=15)
 
-    # test var_names as dict
     pbmc = pbmc68k_reduced()
     sc.tl.leiden(
         pbmc,
@@ -154,8 +137,13 @@ def test_heatmap(image_comparer):
     )
     save_and_compare_images("heatmap_var_as_dict")
 
-    # test that plot elements are well aligned
-    # small
+
+@needs.leidenalg
+@pytest.mark.parametrize("swap_axes", [True, False])
+def test_heatmap_alignment(*, image_comparer, swap_axes: bool) -> None:
+    """Test that plot elements are well aligned."""
+    save_and_compare_images = partial(image_comparer, ROOT, tol=15)
+
     a = AnnData(
         np.array([[0, 0.3, 0.5], [1, 1.3, 1.5], [2, 2.3, 2.5]]),
         obs={"foo": ["a", "b", "c"]},
@@ -166,21 +154,11 @@ def test_heatmap(image_comparer):
         a,
         var_names=a.var_names,
         groupby="foo",
-        swap_axes=True,
-        figsize=(4, 4),
-        show=False,
-    )
-    save_and_compare_images("heatmap_small_swap_alignment")
-
-    sc.pl.heatmap(
-        a,
-        var_names=a.var_names,
-        groupby="foo",
-        swap_axes=False,
+        swap_axes=swap_axes,
         figsize=(4, 4),
         show=False,
     )
-    save_and_compare_images("heatmap_small_alignment")
+    save_and_compare_images(f"heatmap_small{'_swap' if swap_axes else ''}_alignment")
 
 
 @pytest.mark.skipif(