Change csc/csr_matrix to array

docs/source/mods/min-cost-flow.rst

@@ -89,7 +89,7 @@ An example of these inputs with their respective requirements is shown below.
           >>> from gurobi_optimods import datasets
           >>> G, capacities, cost, demands = datasets.simple_graph_scipy()
           >>> G
-          <5x6 sparse matrix of type '<class 'numpy.int64'>'
+          <5x6 sparse array of type '<class 'numpy.int64'>'
                   with 7 stored elements in COOrdinate format>
           >>> print(G)
             (0, 1)        1
@@ -207,7 +207,7 @@ formats.
           >>> obj
           31.0
           >>> sol
-          <5x6 sparse matrix of type '<class 'numpy.float64'>'
+          <5x6 sparse array of type '<class 'numpy.float64'>'
                   with 5 stored elements in COOrdinate format>
           >>> print(sol)
             (0, 1)        1.0

docs/source/mods/qubo.rst

@@ -109,7 +109,7 @@ definition as a SciPy sparse matrix in a comment).
     # weights = [-3, 2, -1, -2, 3]
     # row = [1, 2, 0, 0, 1]
     # col = [1, 2, 1, 2, 2]
-    # Q = sp.coo_matrix((weights, (row, col)), shape=(3, 3))
+    # Q = sp.coo_array((weights, (row, col)), shape=(3, 3))
 
     result = solve_qubo(Q)
 

pyproject.toml

@@ -23,7 +23,7 @@ classifiers = [
     "Operating System :: OS Independent",
 ]
 dependencies = [
-    "gurobipy>=10.0.1",
+    "gurobipy>=10.0.3",
     "gurobipy-pandas>=1.0.0",
     "numpy",
     "pandas",

src/gurobi_optimods/bipartite_matching.py

@@ -194,7 +194,7 @@ def _maximum_bipartite_matching_scipy(adjacency, nodes1, nodes2, create_env):
         indptr = np.arange(0, 2 * from_arc.shape[0] + 2, 2)
         ones = np.ones(from_arc.shape)
         data = np.column_stack((ones * -1.0, ones)).reshape(-1, order="C")
-        A = sp.csc_matrix((data, indices, indptr))
+        A = sp.csc_array((data, indices, indptr))
 
         # Solve model with gurobi, return cost and flows
         x = model.addMVar(A.shape[1], lb=0, ub=capacity)
@@ -215,5 +215,5 @@ def _maximum_bipartite_matching_scipy(adjacency, nodes1, nodes2, create_env):
 
     # Return undirected, unweighted adjacency matrix
     arg = (np.ones(from_arc_result.shape), (from_arc_result, to_arc_result))
-    matching = sp.coo_matrix(arg, dtype=float, shape=G.shape)
+    matching = sp.coo_array(arg, dtype=float, shape=G.shape)
     return matching + matching.T

src/gurobi_optimods/datasets.py

@@ -106,7 +106,7 @@ def _convert_pandas_to_scipy(
     edge_data, node_data, capacity=True, cost=True, demand=True
 ):
     """
-    Convert from a pandas DataFrame to several scipy.sparse.coo_matrix contain
+    Convert from a pandas DataFrame to several scipy.sparse.coo_array contain
     the graph structure, the capacity and cost values per edge, and the demand
     values per node.
     """
@@ -116,17 +116,17 @@ def _convert_pandas_to_scipy(
     a1 = np.array([c[1] for c in coords])
 
     data = np.ones(len(coords), dtype=np.int64)
-    G = sp.coo_matrix((data, (a0, a1)))
+    G = sp.coo_array((data, (a0, a1)))
 
     costs = None
     if cost:
         data = edge_data["cost"].values
-        costs = sp.coo_matrix((data, (a0, a1)))
+        costs = sp.coo_array((data, (a0, a1)))
 
     cap = None
     if capacity:
         data = edge_data["capacity"].values
-        cap = sp.coo_matrix((data, (a0, a1)))
+        cap = sp.coo_array((data, (a0, a1)))
 
     dem = None
     if demand:

src/gurobi_optimods/min_cost_flow.py

@@ -132,7 +132,7 @@ def min_cost_flow_scipy(
     ones = np.ones(edge_source.shape)
     data = np.column_stack((ones * -1.0, ones)).reshape(-1, order="C")
 
-    A = sp.csc_matrix((data, indices, indptr))
+    A = sp.csc_array((data, indices, indptr))
 
     logger.info("Solving min-cost flow with {0} nodes and {1} edges".format(*A.shape))
 
@@ -149,7 +149,7 @@ def min_cost_flow_scipy(
         edge_source_result = edge_source[select]
         edge_target_result = edge_target[select]
         arg = (x.X[select], (edge_source_result, edge_target_result))
-        return model.ObjVal, sp.coo_matrix(arg, dtype=float, shape=G.shape)
+        return model.ObjVal, sp.coo_array(arg, dtype=float, shape=G.shape)
 
 
 @optimod()

tests/test_bipartite_matching.py

@@ -137,6 +137,17 @@ def test_random_csr_matrix(self):
         self.assertEqual(matching.shape, adjacency.shape)
         self.assert_is_unweighted_matching(matching)
 
+    def test_random_csr_array(self):
+        # Property test for matchings on random graphs
+        adjacency, nodes1, nodes2 = random_bipartite(n1=8, n2=7, p=0.5, seed=98634)
+
+        matching = maximum_bipartite_matching(sp.csr_array(adjacency), nodes1, nodes2)
+
+        self.assertIsInstance(matching, sp.spmatrix)
+        self.assertIsNot(matching, adjacency)
+        self.assertEqual(matching.shape, adjacency.shape)
+        self.assert_is_unweighted_matching(matching)
+
     def test_random_csc_array(self):
         # Property test for matchings on random graphs
         adjacency, nodes1, nodes2 = random_bipartite(n1=5, n2=2, p=0.5, seed=34687)