Merge branch 'dask-sparse-mean-var' of github.com:scverse/scanpy into…

… dask-sparse-mean-var

.azure-pipelines.yml

@@ -4,7 +4,7 @@ trigger:
 
 variables:
   python.version: '3.12'
-  PYTEST_ADDOPTS: '-v --color=yes --durations=0 --nunit-xml=test-data/test-results.xml'
+  PYTEST_ADDOPTS: '-v --color=yes --nunit-xml=test-data/test-results.xml'
   TEST_EXTRA: 'test-full'
   DEPENDENCIES_VERSION: "latest"  # |"pre-release" | "minimum-version"
   TEST_TYPE: "standard" # | "coverage"

docs/release-notes/1.10.0.md

@@ -49,6 +49,7 @@
 * Fix warnings caused by internal usage of `pandas.DataFrame.stack` with `pandas>=2.1` {pr}`2864`{smaller}`I Virshup`
 * {func}`scanpy.get.aggregate` now always returns {class}`numpy.ndarray` {pr}`2893` {smaller}`S Dicks`
 * Removes self from array of neighbors for `use_approx_neighbors = True` in {func}`~scanpy.pp.scrublet` {pr}`2896`{smaller}`S Dicks`
+* Compatibility with scipy 1.13 {pr}`2943` {smaller}`I Virshup`
 
 
 

scanpy/_utils/__init__.py

@@ -280,25 +280,6 @@ def get_igraph_from_adjacency(adjacency, directed=None):
     return g
 
 
-def get_sparse_from_igraph(graph, weight_attr=None):
-    from scipy.sparse import csr_matrix
-
-    edges = graph.get_edgelist()
-    if weight_attr is None:
-        weights = [1] * len(edges)
-    else:
-        weights = graph.es[weight_attr]
-    if not graph.is_directed():
-        edges.extend([(v, u) for u, v in edges])
-        weights.extend(weights)
-    shape = graph.vcount()
-    shape = (shape, shape)
-    if len(edges) > 0:
-        return csr_matrix((weights, zip(*edges)), shape=shape)
-    else:
-        return csr_matrix(shape)
-
-
 # --------------------------------------------------------------------------------
 # Group stuff
 # --------------------------------------------------------------------------------

scanpy/tools/_paga.py

@@ -186,7 +186,7 @@ def _compute_connectivities_v1_2(self):
         n = sum(ns)
         es_inner_cluster = [vc.subgraph(i).ecount() for i in range(len(ns))]
         cg = vc.cluster_graph(combine_edges="sum")
-        inter_es = _utils.get_sparse_from_igraph(cg, weight_attr="weight")
+        inter_es = cg.get_adjacency_sparse(attribute="weight")
         es = np.array(es_inner_cluster) + inter_es.sum(axis=1).A1
         inter_es = inter_es + inter_es.T  # \epsilon_i + \epsilon_j
         connectivities = inter_es.copy()
@@ -220,7 +220,7 @@ def _compute_connectivities_v1_0(self):
         )
         ns = vc.sizes()
         cg = vc.cluster_graph(combine_edges="sum")
-        inter_es = _utils.get_sparse_from_igraph(cg, weight_attr="weight") / 2
+        inter_es = cg.get_adjacency_sparse(attribute="weight") / 2
         connectivities = inter_es.copy()
         inter_es = inter_es.tocoo()
         n_neighbors_sq = self._neighbors.n_neighbors**2
@@ -301,7 +301,7 @@ def compute_transitions(self):
         )
         # set combine_edges to False if you want self loops
         cg_full = vc.cluster_graph(combine_edges="sum")
-        transitions = _utils.get_sparse_from_igraph(cg_full, weight_attr="weight")
+        transitions = cg_full.get_adjacency_sparse(attribute="weight")
         transitions = transitions - transitions.T
         transitions_conf = transitions.copy()
         transitions = transitions.tocoo()
@@ -343,7 +343,7 @@ def compute_transitions_old(self):
             g_bool, membership=self._adata.obs[self._groups_key].cat.codes.values
         )
         cg_bool = vc_bool.cluster_graph(combine_edges="sum")  # collapsed version
-        transitions = _utils.get_sparse_from_igraph(cg_bool, weight_attr="weight")
+        transitions = cg_bool.get_adjacency_sparse(attribute="weight")
         total_n = self._neighbors.n_neighbors * np.array(vc_bool.sizes())
         transitions_ttest = transitions.copy()
         transitions_confidence = transitions.copy()