Merge pull request #2715 from Robbybp/incidence-graph

[contrib] Refactor `incidence_analysis` to cache a graph instead of a matrix

pyomo/contrib/incidence_analysis/__init__.py

@@ -1,7 +1,10 @@
 from .triangularize import block_triangularize
 from .matching import maximum_matching
-from .interface import IncidenceGraphInterface
+from .interface import (
+    IncidenceGraphInterface,
+    get_bipartite_incidence_graph,
+)
 from .util import (
-        generate_strongly_connected_components,
-        solve_strongly_connected_components,
-        )
+    generate_strongly_connected_components,
+    solve_strongly_connected_components,
+)

pyomo/contrib/incidence_analysis/common/dulmage_mendelsohn.py

@@ -27,8 +27,7 @@
 
 
 def _get_projected_digraph(bg, matching, top_nodes):
-    """
-    """
+    """ """
     digraph = DiGraph()
     digraph.add_nodes_from(top_nodes)
     for n in top_nodes:
@@ -45,8 +44,7 @@ def _get_projected_digraph(bg, matching, top_nodes):
 
 
 def _get_reachable_from(digraph, sources):
-    """
-    """
+    """ """
     _filter = set()
     reachable = []
     for node in sources:
@@ -96,14 +94,17 @@ def dulmage_mendelsohn(bg, top_nodes=None, matching=None):
     t_other = [t for t in top_nodes if t not in _filter]
     b_other = [b for b in bot_nodes if b not in _filter]
 
-    return ((
-        t_unmatched,
-        t_reachable,
-        t_matched_with_reachable,
-        t_other,
-        ), (
-        b_unmatched,
-        b_reachable,
-        b_matched_with_reachable,
-        b_other,
-        ))
+    return (
+        (
+            t_unmatched,
+            t_reachable,
+            t_matched_with_reachable,
+            t_other,
+        ),
+        (
+            b_unmatched,
+            b_reachable,
+            b_matched_with_reachable,
+            b_other,
+        ),
+    )

pyomo/contrib/incidence_analysis/connected.py

@@ -13,8 +13,7 @@
 
 
 def get_independent_submatrices(matrix):
-    """
-    """
+    """ """
     nxc = nx.algorithms.components
     nxb = nx.algorithms.bipartite
     from_biadjacency_matrix = nxb.matrix.from_biadjacency_matrix
@@ -27,11 +26,10 @@ def get_independent_submatrices(matrix):
     # nodes have values in [N, N+M-1].
     # We could also check the "bipartite" attribute of each node...
     row_blocks = [
-        sorted([node for node in comp if node < N])
-        for comp in connected_components
+        sorted([node for node in comp if node < N]) for comp in connected_components
     ]
     col_blocks = [
-        sorted([node-N for node in comp if node >= N])
+        sorted([node - N for node in comp if node >= N])
         for comp in connected_components
     ]
     return row_blocks, col_blocks

pyomo/contrib/incidence_analysis/dulmage_mendelsohn.py

@@ -13,20 +13,22 @@
 from pyomo.common.dependencies import networkx as nx
 from pyomo.contrib.incidence_analysis.common.dulmage_mendelsohn import (
     dulmage_mendelsohn as dm_nx,
-    )
+)
+
 """
 This module imports the general Dulmage-Mendelsohn-on-a-graph function
 from "common" and implements an interface for coo_matrix-like objects.
 """
 
 RowPartition = namedtuple(
-        "RowPartition",
-        ["unmatched", "overconstrained", "underconstrained", "square"],
-        )
+    "RowPartition",
+    ["unmatched", "overconstrained", "underconstrained", "square"],
+)
 ColPartition = namedtuple(
-        "ColPartition",
-        ["unmatched", "underconstrained", "overconstrained", "square"],
-        )
+    "ColPartition",
+    ["unmatched", "underconstrained", "overconstrained", "square"],
+)
+
 
 def dulmage_mendelsohn(matrix_or_graph, top_nodes=None, matching=None):
     """
@@ -45,9 +47,9 @@ def dulmage_mendelsohn(matrix_or_graph, top_nodes=None, matching=None):
         graph = matrix_or_graph
         if top_nodes is None:
             raise ValueError(
-                    "top_nodes must be specified if a graph is provided,"
-                    "\notherwise the result is ambiguous."
-                    )
+                "top_nodes must be specified if a graph is provided,"
+                "\notherwise the result is ambiguous."
+            )
         partition = dm_nx(graph, top_nodes=top_nodes, matching=matching)
         # RowPartition and ColPartition do not make sense for a general graph.
         # However, here we assume that this graph comes from a Pyomo model,
@@ -74,17 +76,17 @@ def dulmage_mendelsohn(matrix_or_graph, top_nodes=None, matching=None):
         # Matrix rows have bipartite=0, columns have bipartite=1
         bg = from_biadjacency_matrix(matrix)
         row_partition, col_partition = dm_nx(
-                bg,
-                top_nodes=list(range(M)),
-                matching=matching,
-                )
+            bg,
+            top_nodes=list(range(M)),
+            matching=matching,
+        )
 
         partition = (
-                row_partition,
-                tuple([n - M for n in subset] for subset in col_partition)
-                # Column nodes have values in [M, M+N-1]. Apply the offset
-                # to get values corresponding to indices in user's matrix.
-                )
+            row_partition,
+            tuple([n - M for n in subset] for subset in col_partition)
+            # Column nodes have values in [M, M+N-1]. Apply the offset
+            # to get values corresponding to indices in user's matrix.
+        )
 
     partition = (RowPartition(*partition[0]), ColPartition(*partition[1]))
     return partition