Cleanup - test against different versions of libraries and setups

HISTORY.rst

@@ -32,6 +32,7 @@ API and Bug Fixes:
 * ``get_interaction_matrix`` and ``get_interaction_pvals`` take the graph's
   CSR arrays (``neigh_indptr``, ``neigh_indices``) instead of a numba ``List``.
 * ``get_edges`` no longer needs to deduplicates in a quadratic loop.
+* Removed ``UndefinedException`` from ``resistance_matrix``.
 
 1.5.0 (2026-11-01)
 ------------------

docs/release_notes/1.5.1.rst

@@ -30,4 +30,5 @@ and aligned to other the LR data.
   ``stlearn.tl.cci.edges`` are removed.
 * ``get_interaction_matrix`` and ``get_interaction_pvals`` take the graph's
   CSR arrays (``neigh_indptr``, ``neigh_indices``) instead of a numba ``List``.
-* ``get_edges`` no longer needs to deduplicates in a quadratic loop.
+* ``get_edges`` no longer needs to deduplicates in a quadratic loop.
+* Removed ``UndefinedException`` from ``resistance_matrix``.

pyproject.toml

@@ -17,33 +17,39 @@ dependencies = [
     "anndata>=0.12.2",
     "bokeh>=3.7.0,<4.0",
     "click>=8.2.0,<9.0",
+    "dask>=2025.12.0",
+    "geopandas>=1.0.0,<2.0",
+    "h5py>=3.11",
     "igraph>=1.0.0",
+    "imageio>=2.37.0,<3.0",
     "leidenalg>=0.11.0",
     "matplotlib>=3.11,<4.0",
+    "natsort",
+    "networkx>=3.0",
     "numba>=0.64.0",
     "numpy>=2.4.0",
+    "openpyxl>=3.1",
+    "pandas>=2.3.0",
     "pillow>=13.0.0,<14.0",
+    "pooch[progress]>=1.8.1",
+    "requests>=2.32.4",
     "scanpy>=1.12.0",
     "scikit-image>=0.25.0",
-    "pandas>=2.3.0",
-    "imageio>=2.37.0,<3.0",
-    "scipy>=1.17.1,<2.0",
     "scikit-learn>=1.8.0,<2.0",
     "scikit-misc>=0.5.1",
-    "zarr>=3.1.6,<4.0",
-    "dask>=2025.12.0",
+    "scipy>=1.17.1,<2.0",
+    "shapely>=2.0.6,<3.0",
     "spatialdata>=0.7.3",
     "spatialdata-io>=0.6.0",
     "spatialdata-plot>=0.3.0",
+    "statsmodels>=0.14.5",
     "torch>=2.5.0",
     "torchvision>=0.20.0",
-    "geopandas>=1.0.0,<2.0",
-    "shapely>=2.0.6,<3.0",
-    "pooch[progress]>=1.8.1",
-    "requests>=2.32.4",
-    "tifffile>=2026.5.2",
+    "tqdm",
     "umap-learn>=0.5.12",
-    "openpyxl>=3.1",
+    "tifffile>=2026.5.2",
+    "xarray>=2024.10.0",
+    "zarr>=3.1.6,<4.0",
 ]
 
 keywords = ["stlearn"]
@@ -65,7 +71,9 @@ dev = [
     "black>=26.10",
     "ruff>=0.17.3",
     "mypy>=2.0",
-    "pytest>=7.0",
+    "pytest>=8.0",
+    "pytest-cov>=5",
+    "psutil>=5.9",
     "tox>=4.0",
     "ghp-import>=2.1.0",
     "sphinx>=4.0",
@@ -123,12 +131,8 @@ select = [
     "RUF", # ruff-specific rules
 ]
 ignore = [
-    "E722", # bare except — you've already accepted these
-    "F811", # redefinition — accepted
     "N803", "N806", # naming: lowercase args/vars — too invasive to fix
-    "N818", # exception naming
-    "PD011", # .values vs .to_numpy() — too noisy, mostly false positives
-    "B008", # function calls in default args — used by typer/click
+    "PD011", # .values vs .to_numpy()
     "SIM108", # ternary suggestions can hurt readability
 ]
 

stlearn/spatial/trajectory/utils.py

@@ -296,18 +296,6 @@ def laplacian_matrix(A, normalized=False):
     return L
 
 
-# """
-# **********
-# Exceptions
-# **********
-# Custom exceptions for NetComp.
-# """
-
-
-class UndefinedException(Exception):
-    """Raised when matrix to be returned is undefined"""
-
-
 # """
 # **********
 # Resistance
@@ -329,6 +317,11 @@ def resistance_matrix(A, check_connected=True):
     -------
     R : NumPy matrix
        Matrix of pairwise resistances between nodes.
+    Raises
+    ------
+    ValueError
+        If check_connected is True and the graph is not connected, since the
+        resistance matrix is undefined in that case.
     Notes
     -----
     Uses formula for resistance matrix R in terms of Moore-Penrose of
@@ -358,9 +351,7 @@ def resistance_matrix(A, check_connected=True):
         else:
             G = nx.from_numpy_array(A)
         if not nx.is_connected(G):
-            raise UndefinedException(
-                "Graph is not connected. Resistance matrix is undefined."
-            )
+            raise ValueError("Graph is not connected. Resistance matrix is undefined.")
     L = laplacian_matrix(A)
     with contextlib.suppress(AttributeError):
         L = L.todense()

stlearn/tl/cci/analysis.py

@@ -547,7 +547,7 @@ def run_cci(
         cols_present = np.all(
             [cell_type in adata.uns[uns_key] for cell_type in all_set],
         )
-        rows_present = np.all(adata.uns[uns_key].index.values == adata.obs_names.values)
+        rows_present = adata.uns[uns_key].index.equals(adata.obs_names)
         msg = f"Cell type scores misformatted in adata.uns[{uns_key}]:\n"
         if not cols_present or not rows_present:
             if not cols_present:
@@ -578,7 +578,7 @@ def run_cci(
     count_col = "n_spots_sig" if sig_spots else "n_spots"
     col = "lr_sig_scores" if sig_spots else "lr_scores"
     best_lr_indices = np.flatnonzero(lr_summary[count_col].to_numpy() > min_spots)
-    best_lrs = lr_summary.index.values[best_lr_indices]
+    best_lrs = lr_summary.index.to_numpy()[best_lr_indices]
     lr_genes = genes_of(best_lrs)
     if len(lr_genes) == 0:
         raise ValueError(

stlearn/tl/cci/gridding.py

@@ -169,7 +169,7 @@ def grid(
     grid_expr = pd.DataFrame(
         grid_expr,
         index=[f"grid_{i}" for i in range(n_squares)],
-        columns=adata.var_names.values.astype(str),
+        columns=adata.var_names.to_numpy().astype(str),
     )
     grid_data = AnnData(grid_expr)
     grid_data.obs["imagecol"] = grid_coords[:, 0]
@@ -184,7 +184,7 @@ def grid(
     if use_label is not None and cell_info is not None and cell_set is not None:
         grid_data.uns[use_label] = pd.DataFrame(
             cell_info,
-            index=grid_data.obs_names.values.astype(str),
+            index=grid_data.obs_names.to_numpy().astype(str),
             columns=cell_set,
         )
         max_indices = np.apply_along_axis(np.argmax, 1, cell_info)

stlearn/tl/cci/het.py

@@ -123,7 +123,7 @@ def get_data_for_counting(adata, use_label, mix_mode, all_set):
         col_order = [cols.index(cell_type) for cell_type in all_set]
         cell_data = cell_props.iloc[:, col_order].to_numpy(dtype=np.float64)
     else:
-        cell_labels = adata.obs.loc[:, obs_key].values
+        cell_labels = adata.obs[obs_key].to_numpy().astype(str)
         cell_data = np.zeros((len(cell_labels), len(all_set)), dtype=np.float64)
         for i, cell_type in enumerate(all_set):
             cell_data[:, i] = (

stlearn/tl/cci/ligand_receptors.py

@@ -148,7 +148,7 @@ def __init__(self, frame: pd.DataFrame, *, species: str, source: str = "DataFram
             examples = frame.loc[with_underscore, ["ligand", "receptor"]].head(3)
             raise ValueError(
                 f"Gene symbols in {source} contain '_', which separates ligand from "
-                f"receptor in 'L_R' pair names; e.g. {examples.values.tolist()}",
+                f"receptor in 'L_R' pair names; e.g. {examples.to_numpy().tolist()}",
             )
         frame = _merge_repeats(frame)
         self._set(frame, species.lower(), _labels(frame))

stlearn/tl/cci/perm_utils.py

@@ -229,7 +229,7 @@ def get_lr_features(lr_expr: pd.DataFrame | AnnData, lrs: LRPairs, quantiles):
         # Column by column below, so CSC; dense stays as it is.
         expr = sparse.csc_matrix(expr) if sparse.issparse(expr) else np.asarray(expr)
     else:
-        genes, expr = lr_expr.columns.values, lr_expr.values
+        genes, expr = lr_expr.columns.to_numpy(), lr_expr.to_numpy()
 
     # Determining indices of LR pairs #
     l_indices, r_indices = [], []

stlearn/tl/cci/permutation.py

@@ -93,10 +93,10 @@ def perform_spot_testing(
     lr_feats = get_lr_features(adata[:, lr_genes], lrs, quantiles)
     l_quants = lr_feats.loc[
         names, [col for col in lr_feats.columns if "L_" in col]
-    ].values
+    ].to_numpy()
     r_quants = lr_feats.loc[
         names, [col for col in lr_feats.columns if "R_" in col]
-    ].values
+    ].to_numpy()
     candidate_quants = gene_quantiles(
         adata, genes, quantiles, expression=scorer.columns
     )