Source code for gemseo.core.dependency_graph

# Copyright 2021 IRT Saint Exupéry, https://www.irt-saintexupery.com
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License version 3 as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public License
# along with this program; if not, write to the Free Software Foundation,
# Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
# Contributors:
#    INITIAL AUTHORS - initial API and implementation and/or initial
#                         documentation
#        :author: Charlie Vanaret
"""Graphs of the disciplines dependencies and couplings."""
from __future__ import annotations

import logging
from shutil import move
from typing import Iterator

# graphviz is an optional dependency

try:
    import graphviz
except ImportError:
    graphviz = None
import networkx as nx

from gemseo.core.discipline import MDODiscipline
from pathlib import Path

LOGGER = logging.getLogger(__name__)


[docs]class DependencyGraph: """Graph of dependencies between disciplines. This class can create the sequence of execution of the disciplines. This is done by determining the strongly connected components (scc) of the graph. The disciplines in the components of a scc have the same order as in the passed disciplines passed when instantiating the class. The couplings between the disciplines can also be computed. """ def __init__(self, disciplines): # noqa: D201,D205,D212,D415 """ Args: disciplines: The disciplines to build the graph with. """ self.__graph = self.__create_graph(disciplines) @property def disciplines(self) -> Iterator[MDODiscipline]: """The disciplines used to build the graph.""" return iter(self.__graph.nodes)
[docs] def get_execution_sequence(self): """Compute the execution sequence of the disciplines. Returns: list(set(tuple(set(MDODisciplines)))) """ condensed_graph = self.__create_condensed_graph() execution_sequence = [] while True: leaves = self.__get_leaves(condensed_graph) if not leaves: break parallel_tasks = list( tuple(condensed_graph.nodes[node_id]["members"]) for node_id in leaves ) execution_sequence += [parallel_tasks] condensed_graph.remove_nodes_from(leaves) return list(reversed(execution_sequence))
def __create_condensed_graph(self): """Return the condensed graph.""" # scc = nx.kosaraju_strongly_connected_components(self.__graph) # scc = nx.strongly_connected_components_recursive(self.__graph) # fastest routine scc = nx.strongly_connected_components(self.__graph) return nx.condensation(self.__graph, scc=self.__get_ordered_scc(scc)) def __get_ordered_scc(self, scc): """Return the scc nodes ordered by the initial disciplines. Args: scc: The scc nodes. Yields: List[MDODisciplines]: The ordered scc nodes. """ disciplines = list(self.__graph.nodes) for components in scc: disc_indexes = {} for component in components: index = disciplines.index(component) disc_indexes[index] = component ordered_components = [] for index in sorted(disc_indexes.keys()): ordered_components += [disc_indexes[index]] yield ordered_components
[docs] def get_disciplines_couplings(self): """Return the couplings between the disciplines. Returns: The disciplines couplings, a coupling is composed of a discipline, one of its successor and the sorted variables names. """ couplings = [] for from_disc, to_disc, edge_names in self.__graph.edges(data="io"): couplings += [(from_disc, to_disc, sorted(edge_names))] return couplings
@staticmethod def __create_graph(disciplines): """Create the full graph. The coupled inputs and outputs names are stored as an edge attributes named io. Args: disciplines (list): The disciplines to build the graph with. Returns: networkx.DiGraph: The graph of disciplines. """ # python 2: for consistency with the python 3 version nodes_to_ios = {} for disc in disciplines: nodes_to_ios[disc] = ( set(disc.get_input_data_names()), set(disc.get_output_data_names()), ) graph = nx.DiGraph() graph.add_nodes_from(disciplines) graph_add_edge = graph.add_edge # Create the graph edges by discovering the coupled disciplines for disc_i, (_, outputs_i) in nodes_to_ios.items(): for disc_j, (inputs_j, _) in nodes_to_ios.items(): if disc_i != disc_j: coupled_io = outputs_i & inputs_j if coupled_io: graph_add_edge(disc_i, disc_j, io=coupled_io) return graph @staticmethod def __get_node_name(graph, node): """Return the name of a node for the representation of a graph. Args: graph (networkx.DiGraph): A full or condensed graph. node (int, MDODiscipline): A node of the graph. Returns: str: The name of the node. """ if isinstance(node, MDODiscipline): # not a scc node return str(node) # networkx stores the scc nodes under the members node attribute condensed_discs = graph.nodes[node]["members"] if len(condensed_discs) == 1: # not a scc node in a scc graph return str(list(condensed_discs)[0]) # scc node return "MDA of {}".format(", ".join(map(str, condensed_discs))) @staticmethod def __get_scc_edge_names(graph, node_from, node_to=None): """Return the names of an edge in a condensed graph. Args: graph (networkx.DiGraph): A condensed graph. node_from (int): The predecessor node in the graph. node_to (int, optional): The successor node in the graph. Returns: set(str): The names of the edge. """ output_names = set() for disc in graph.nodes[node_from]["members"]: output_names.update(disc.get_output_data_names()) if node_to is None: return output_names input_names = set() for disc in graph.nodes[node_to]["members"]: input_names.update(disc.get_input_data_names()) return output_names & input_names def __write_graph(self, graph, file_path): """Write the representation of a graph. Args: graph (networkx.DiGraph): A graph. file_path (str): A path to the file. """ if graphviz is None: LOGGER.warning("Cannot write graph: graphviz cannot be imported.") return viz_graph = graphviz.Digraph(name="Dependency graph") # add edges for node_from, node_to, edge_names in graph.edges(data="io"): name_from = self.__get_node_name(graph, node_from) name_to = self.__get_node_name(graph, node_to) if not isinstance(node_to, MDODiscipline): # a scc edge edge_names = self.__get_scc_edge_names(graph, node_from, node_to) if edge_names is not None: label = ",".join(sorted(edge_names)) else: label = None viz_graph.edge(name_from, name_to, label=label) # add leaves with an invisible node so an edge with the edge names are visible for node_from in self.__get_leaves(graph): if isinstance(node_from, MDODiscipline): edge_names = node_from.get_output_data_names() node_name = str(node_from) else: # a scc edge edge_names = self.__get_scc_edge_names(graph, node_from) node_name = self.__get_node_name(graph, node_from) if not edge_names: continue dummy_node_name = f"_{node_from}" viz_graph.node(dummy_node_name, style="invis", shape="point") viz_graph.edge( node_name, dummy_node_name, label=",".join(sorted(edge_names)) ) # write the dot and target files path = Path(file_path) viz_graph.render(str(path.with_suffix("")), format=path.suffix[1:], view=False) # rename the dot file left by graphviz with the proper extension path_with_dot = path.with_suffix(".dot") if path_with_dot.exists(): path_with_dot.unlink() move(str(path.with_suffix("")), str(path_with_dot))
[docs] def write_full_graph(self, file_path): """Write a representation of the full graph. Args: file_path (str): A path to the file. """ self.__write_graph(self.__graph, file_path)
export_initial_graph = write_full_graph
[docs] def write_condensed_graph(self, file_path): """Write a representation of the condensed graph. Args: file_path (str): A path to the file. """ self.__write_graph(self.__create_condensed_graph(), file_path)
export_reduced_graph = write_condensed_graph @staticmethod def __get_leaves(graph): """Return the leaf nodes of a graph. Args: graph (networkx.DiGraph): A graph. """ return [n for n in graph.nodes if graph.out_degree(n) == 0]