def QueryCollapsedGraphs(dataset, request_type, fuzzy):
"""
Find motifs in the graphs with collapsed node sequences. Saves the motifs to
file
@params dataset: dataset to find motifs in the collapsed sequences
@params request_type: request for this particular set of traces
@params fuzzy: can this motif be fuzzy?
"""
# create the directory structure
if not os.path.exists('motifs'):
os.mkdir('motifs')
if not os.path.exists('motifs/subgraphs'):
os.mkdir('motifs/subgraphs')
if not os.path.exists('motifs/subgraphs/{}'.format(dataset)):
os.mkdir('motifs/subgraphs/{}'.format(dataset))
# read the frequent subgraphs for this dataset/request type
subgraphs = IdentifyFrequentSubgraphs(dataset, request_type, True, fuzzy)
# read all of the traces
traces = ReadTraces(dataset, request_type, None)
for iv, trace in enumerate(traces):
# start statistics
start_time = time.time()
# write the motifs to disk
if fuzzy: output_filename = 'motifs/subgraphs/{}/{}-motifs-fuzzy-collapsed-complete.motifs'.format(dataset, trace.base_id)
else: output_filename = 'motifs/subgraphs/{}/{}-motifs-collapsed-complete.motifs'.format(dataset, trace.base_id)
if os.path.exists(output_filename): continue
# reduced nodes to nodes is a funciton to go from the reduced node space to the original
graph, reduced_nodes_to_nodes = ConvertCollapsedGraph2GraphTool(trace, fuzzy)
motifs = []
for motif_index, subgraph in enumerate(subgraphs):
motif = ConvertSubGraph2GraphTool(subgraph)
# use graph tool to find all motif occurrences
vertex_maps = gt.subgraph_isomorphism(motif, graph, vertex_label=(motif.vp.label, graph.vp.label))
# go through all of the found motif patterns
for vertex_map in vertex_maps:
nodes = []
# add all of the nodes that belong to each vertex (collapsed nodes)
for reduced_node in vertex_map:
for node in reduced_nodes_to_nodes[reduced_node]:
assert (not trace.nodes[node] in nodes)
nodes.append(trace.nodes[node])
motifs.append(Motif(trace, nodes, motif_index))
WriteMotifs(output_filename, motifs)
# print statistics
print ('Mined traces for {} in {:0.2f} seconds.'.format(trace.base_id, time.time() - start_time))
def QueryTraces(dataset, request_type):
"""
Find all occurrences for each motif for this dataset/request_type comboination.
@params dataset: the dataset to mine for frequent sub graphs
@params request_type: the request type for this set of traces
"""
# create the directory structure
if not os.path.exists('motifs'):
os.mkdir('motifs')
if not os.path.exists('motifs/subgraphs'):
os.mkdir('motifs/subgraphs')
if not os.path.exists('motifs/subgraphs/{}'.format(dataset)):
os.mkdir('motifs/subgraphs/{}'.format(dataset))
# start statistics
start_time = time.time()
# read the frequent subgraphs for this dataset/request type
subgraphs = IdentifyFrequentSubgraphs(dataset, request_type, False, False)
# read all of the traces and mine the graph
traces = ReadTraces(dataset, request_type, None)
for iv, trace in enumerate(traces):
# start statistics
start_time = time.time()
# skip over the file if it already exists
output_filename = 'motifs/subgraphs/{}/{}-motifs-complete.motifs'.format(dataset, trace.base_id)
if os.path.exists(output_filename): continue
graph = ConvertTrace2GraphTool(dataset, trace)
motifs = []
for motif_index, subgraph in enumerate(subgraphs):
motif = ConvertSubGraph2GraphTool(subgraph)
# use graph tool to find all motif occurrences
vertex_maps = gt.subgraph_isomorphism(motif, graph, vertex_label=(motif.vp.label, graph.vp.label))
# go through all of the found motif patterns
for vertex_map in vertex_maps:
nodes = []
for node in vertex_map:
nodes.append(trace.nodes[node])
motifs.append(Motif(trace, nodes, motif_index))
# write the motifs to disk
WriteMotifs(output_filename, motifs)
# print statistics
print ('Mined traces for {} in {:0.2f} seconds.'.format(trace.base_id, time.time() - start_time))
def assert_graph_equal(expect, other):
"""
Test if the graph is equal to the given one.
This is intended mostly for testing purposes.
"""
__tracebackhide__ = True
assert expect.num_vertices() == other.num_vertices(),\
"Number of vertices differ expected %d, found %d\n%s\n%s\n\n%s\n%s" % (
expect.num_vertices(), other.num_vertices(),
dump_vertices(expect), dump_edges(expect),
dump_vertices(other), dump_edges(other))
n_vertices = expect.num_vertices()
# check if the graphs are isomorphic
# there must exist an isomorphism for which the vertex
# data matches, so we check all of them.
isomaps = subgraph_isomorphism(expect, other, subgraph=False)
assert len(isomaps) > 0, "Graph topology differ:\n%s\n%s\n\n%s\n%s" % (
dump_vertices(expect), dump_edges(expect), dump_vertices(other),
dump_edges(other))
errors = []
for isomap in isomaps:
try:
# check that vertices match
for v in expect.vertices():
assert_vertex_equal(expect, v, other, isomap[v])
for v_out in v.out_neighbours():
e = expect.edge(v, v_out)
f = other.edge(isomap[v], isomap[v_out])
assert_edge_equal(expect, e, other, f)
except AssertionError as ex:
errors.append((v, ex))
continue
else:
break
else:
# all isomaps raised AssertionErrors
msg = "Graphs do not have an isomorphism for which vertices match.\n"
msg += "Failures for each isomorphism:\n"
for e, iso in zip(errors, isomaps):
msg += "Isomap: %s\n" % [iso[i] for i in range(n_vertices)]
msg += "vertex %d: %s\n\n" % e
msg += "Expected graph:\n%s\n%s\n\n" % (dump_vertices(expect),
dump_edges(expect))
msg += "Found graph:\n%s\n%s\n\n" % (dump_vertices(other),
dump_edges(other))
pytest.fail(msg)
def index_query_graphs(q_gs, f_sgs):
fvectors = np.zeros((len(q_gs), len(f_sgs)))
fv = 0
for f_sg in f_sgs:
qv = 0
for q_g in q_gs:
# subgraph isomorphisms return list of maps, i.e we check for empty if not a subgraph
# check subgraph parameter in function below
if len(
gt.subgraph_isomorphism(
f_sg,
q_g,
max_n=1,
vertex_label=(f_sg.vertex_properties["molecule"],
q_g.vertex_properties["molecule"]),
edge_label=(f_sg.edge_properties["bond"],
q_g.edge_properties["bond"]),
induced=False,
subgraph=True,
generator=False)) > 0:
fvectors[qv][fv] = 1
qv += 1
fv += 1
return fvectors
query_fvectors = index_query_graphs(qgs, fgs)
database_graphs = get_database_graph_objects(outfile)
mapping = get_map(mapfile)
result = {}
for i in range(query_fvectors.shape[0]):
result[i] = []
for j in range(fvectors.shape[0]):
if (np.all(query_fvectors[i] == np.logical_and(
fvectors[j], query_fvectors[i]))):
x = gt.subgraph_isomorphism(
qgs[i],
database_graphs[j],
max_n=1,
vertex_label=(
qgs[i].vertex_properties['molecule'],
database_graphs[j].vertex_properties['molecule']),
edge_label=(qgs[i].edge_properties['bond'],
database_graphs[j].edge_properties['bond']),
induced=False,
subgraph=True,
generator=False)
result[i].append(str(mapping[str(j)]))
qs = len(result)
with open(output_filename, 'w') as f:
for i in range(qs):
f.write('\t'.join(result[i]) + '\n')