hvg works

Author: flying-sheep

src/scanpy/preprocessing/_highly_variable_genes.py

@@ -411,7 +411,10 @@ def _subset_genes(
 def _nth_highest(x: NDArray[np.float64] | DaskArray, n: int) -> float | DaskArray:
     x = x[~np.isnan(x)]
     if n > x.size:
-        msg = "`n_top_genes` > number of normalized dispersions, returning all genes with normalized dispersions."
+        msg = (
+            f"`n_top_genes` (={n}) > number of normalized dispersions (={x.size}), "
+            "returning all genes with normalized dispersions."
+        )
         # 5: caller -> 4: `highly_variable_genes` -> 3: `_…_single_batch` -> 2: `_subset_genes` -> 1: here
         warnings.warn(msg, UserWarning, stacklevel=5)
         n = x.size

src/testing/scanpy/_helpers/__init__.py

@@ -101,7 +101,7 @@ def check_rep_results(func, X, *, fields: Iterable[str] = ("layer", "obsm"), **k
 
 
 def _check_check_values_warnings(
-    function, adata, expected_warning, kwargs=MappingProxyType({})
+    function, adata: AnnData, expected_warning: str, kwargs=MappingProxyType({})
 ):
     """Run `function` on `adata` with provided arguments `kwargs` twice.
 
@@ -110,12 +110,14 @@ def _check_check_values_warnings(
     """
     # expecting 0 no-int warnings
     with warnings.catch_warnings(record=True) as record:
+        warnings.filterwarnings("always")
         function(adata.copy(), **kwargs, check_values=False)
     warning_msgs = [w.message.args[0] for w in record]
     assert expected_warning not in warning_msgs
 
     # expecting 1 no-int warning
     with warnings.catch_warnings(record=True) as record:
+        warnings.filterwarnings("always")
         function(adata.copy(), **kwargs, check_values=True)
     warning_msgs = [w.message.args[0] for w in record]
     assert expected_warning in warning_msgs

tests/test_highly_variable_genes.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import itertools
+from contextlib import nullcontext
 from pathlib import Path
 from string import ascii_letters
 from typing import TYPE_CHECKING
@@ -391,11 +392,12 @@ def test_compare_to_upstream(
         sc.pp.log1p(pbmc)
         sc.pp.highly_variable_genes(pbmc, flavor=flavor, **params, inplace=True)
     elif func == "fgd":
-        sc.pp.filter_genes_dispersion(
-            pbmc, flavor=flavor, **params, log=True, subset=False
-        )
+        with pytest.warns(FutureWarning, match=r"sc\.pp\.highly_variable_genes"):
+            sc.pp.filter_genes_dispersion(
+                pbmc, flavor=flavor, **params, log=True, subset=False
+            )
     else:
-        raise AssertionError()
+        pytest.fail(f"Unknown func {func}")
 
     np.testing.assert_array_equal(
         hvg_info["highly_variable"], pbmc.var["highly_variable"]
@@ -505,7 +507,7 @@ def test_seurat_v3_warning():
 
 def test_batches():
     adata = pbmc68k_reduced()
-    adata[:100, :100].X = np.zeros((100, 100))
+    adata.X[:100, :100] = np.zeros((100, 100))
 
     adata.obs["batch"] = ["0" if i < 100 else "1" for i in range(adata.n_obs)]
     adata_1 = adata[adata.obs["batch"] == "0"].copy()
@@ -558,6 +560,7 @@ def test_batches():
     assert np.all(np.isin(colnames, hvg1.columns))
 
 
+@pytest.mark.filterwarnings("ignore:invalid value encountered:RuntimeWarning")
 def test_degenerate_batches():
     adata = AnnData(
         X=np.random.randn(10, 100),
@@ -692,10 +695,17 @@ def test_dask_consistency(adata: AnnData, flavor, batch_key, to_dask):
     adata_dask = adata.copy()
     adata_dask.X = to_dask(adata_dask.X)
 
-    output_mem, output_dask = (
-        sc.pp.highly_variable_genes(ad, flavor=flavor, n_top_genes=15, inplace=False)
-        for ad in [adata, adata_dask]
-    )
+    with (
+        pytest.warns(UserWarning, match="n_top_genes.*normalized dispersions")
+        if flavor == "cell_ranger"
+        else nullcontext()
+    ):
+        output_mem, output_dask = (
+            sc.pp.highly_variable_genes(
+                ad, flavor=flavor, n_top_genes=15, inplace=False
+            )
+            for ad in [adata, adata_dask]
+        )
 
     assert isinstance(output_mem, pd.DataFrame)
     assert isinstance(output_dask, pd.DataFrame)