def random_data(request):
vertices_num, regex_str = request.param
edges_num = vertices_num * (vertices_num - 1) // 5
v_from = [random.randint(0, vertices_num) for _ in range(edges_num)]
v_to = [random.randint(0, vertices_num) for _ in range(edges_num)]
values = [random.choice(['a', 'b', 'c', 'd']) for _ in range(edges_num)]
edges_list = zip(v_from, values, v_to)
graph = BMGraph.from_edges_list(edges_list)
regex = BMGraph.from_regex_string(regex_str)
return graph, regex
def test_hellings_2():
test_path = os.path.join(os.getcwd(), 'tests/data/cfpq/test2')
graph = BMGraph.from_edges_file(os.path.join(test_path, 'graph.txt'))
grammar = GrammarAlgos.from_grammar_file(
os.path.join(test_path, 'grammar.txt'))
adj_matrix = GrammarAlgos.Hellings(grammar, graph)
expected = {(0, 1), (3, 3)}
actual = set(BMGraph.get_reachable_vertices(adj_matrix))
assert expected == actual
def test_cfpq_matrix_1():
test_path = os.path.join(os.getcwd(), 'tests/data/cfpq/test1')
graph = BMGraph.from_edges_file(os.path.join(test_path, 'graph.txt'))
grammar = GrammarAlgos.from_grammar_file(
os.path.join(test_path, 'grammar.txt'))
adj_matrix = GrammarAlgos.cfpq_matrix_multiplication(grammar, graph)
expected = {(0, 2), (0, 3), (1, 2), (1, 3), (2, 2), (2, 3)}
actual = set(BMGraph.get_reachable_vertices(adj_matrix))
assert expected == actual
def test_cfpq_tensor_3():
test_path = os.path.join(os.getcwd(), 'tests/data/cfpq/test3')
graph = BMGraph.from_edges_file(
os.path.join(test_path, 'graph.txt'))
grammar = GrammarAlgos.from_grammar_file(
os.path.join(test_path, 'grammar.txt'))
adj_matrix = GrammarAlgos.cfpq_tensor_product(grammar, graph)
expected = {(0, 0), (1, 1), (0, 2), (3, 3), (2, 2)}
actual = set(BMGraph.get_reachable_vertices(adj_matrix))
assert expected == actual
def test_intersection_1():
test_path = os.path.join(os.getcwd(), 'tests/data/test1')
graph = BMGraph.from_edges_file(os.path.join(test_path, 'graph.txt'))
regex = BMGraph.from_regex_file(os.path.join(test_path, 'regex.txt'))
intersection = graph.intersect(regex)
ans = intersection.to_automaton()
a = Symbol('a')
b = Symbol('b')
assert ans.accepts([])
assert ans.accepts([a])
assert ans.accepts([a, a, a])
assert not ans.accepts([b])
def main():
parser = ArgumentParser(description="""Intersect graph and regex,
and show reachability of graph vertices"""
)
parser.add_argument(
'path_to_graph',
help='Path to graph represented in \'from value to\' tuples')
parser.add_argument('path_to_regex', help='Path to regex')
parser.add_argument(
'--from',
dest='vertices_from',
help='Optional: source vertices from graph in \'v1 v2 v3 ...\' form')
parser.add_argument(
'--to',
dest='vertices_to',
help=
'Optional: destination vertices from graph in \'v1 v2 v3 ...\' form')
args = parser.parse_args()
graph = BMGraph.from_edges_file(args.path_to_graph)
regex = BMGraph.from_regex_file(args.path_to_regex)
if args.vertices_from is not None:
vertices = read_vertices(args.vertices_from)
graph.start_states = vertices
if args.vertices_to is not None:
vertices = read_vertices(args.vertices_to)
graph.final_states = vertices
intersection = graph.intersect(regex)
print('Edges for each label:')
for (value, matrix) in intersection.matrices.items():
print('{} has {} edges'.format(value, matrix.nvals))
closure = intersection.transitive_closure()
print('Reachable vertices:')
reachable = BMGraph.get_reachable_vertices(closure)
for (v_from, v_to) in reachable:
if v_from in intersection.start_states and v_to in intersection.final_states:
print('{} -> {}'.format(v_from // regex.states_amount,
v_to // regex.states_amount))
def cfpq_tensor_product(grammar: CFG, graph: BMGraph):
res = graph.dup()
rfa = BMGraph()
rfa_heads = dict()
rfa.states_amount = sum(
[len(prod.body) + 1 for prod in grammar.productions])
rfa.states = set(range(rfa.states_amount))
index = 0
for prod in grammar.productions:
start_state = index
final_state = index + len(prod.body)
rfa.start_states.add(start_state)
rfa.final_states.add(final_state)
rfa_heads[(start_state, final_state)] = prod.head.value
for var in prod.body:
matrix = rfa.matrices.get(
var.value,
Matrix.sparse(BOOL, rfa.states_amount, rfa.states_amount))
matrix[index, index + 1] = True
rfa.matrices[var.value] = matrix
index += 1
index += 1
for prod in grammar.productions:
if len(prod.body) == 0:
matrix = Matrix.sparse(BOOL, graph.states_amount,
graph.states_amount)
for i in range(graph.states_amount):
matrix[i, i] = True
res.matrices[prod.head] = matrix
is_changing = True
while is_changing:
is_changing = False
intersection = rfa.intersect(res)
closure = intersection.transitive_closure()
for i, j, _ in zip(*closure.to_lists()):
rfa_from, rfa_to = i // res.states_amount, j // res.states_amount
graph_from, graph_to = i % res.states_amount, j % res.states_amount
if (rfa_from, rfa_to) not in rfa_heads:
continue
var = rfa_heads[(rfa_from, rfa_to)]
matrix = res.matrices.get(
var,
Matrix.sparse(BOOL, graph.states_amount,
graph.states_amount))
if matrix.get(graph_from, graph_to) is None:
is_changing = True
matrix[graph_from, graph_to] = True
res.matrices[var] = matrix
return res.matrices.get(
grammar.start_symbol,
Matrix.sparse(BOOL, graph.states_amount, graph.states_amount))
def closure_benchmark():
path_to_data = os.path.join(os.getcwd(), 'benchmarks/refinedDataForRPQ')
tests = ['LUBM1.9M']
for test in tests:
path_to_test = os.path.join(path_to_data, test)
graph = BMGraph.from_edges_file('{}/{}.txt'.format(path_to_test, test))
output = open('{}/output.csv'.format(path_to_test), 'w+')
regex_dir = os.path.join(path_to_test, 'regexes')
for filename in os.listdir(regex_dir):
closure = None
res = None
regex_str = os.path.join(regex_dir, filename)
regex = BMGraph.from_regex_file(regex_str, False)
sum = 0
for _ in range(5):
start = time.monotonic()
res = graph.intersect(regex)
closure = transitive_closure_sq(res)
end = time.monotonic()
sum += end - start
sq_pairs = closure.nvals
sq_time = sum / 5
start = time.monotonic()
for (value, matrix) in res.matrices.items():
v, n = value, matrix.nvals
end = time.monotonic()
sq_pairs_time = end - start
sum = 0
for _ in range(5):
start = time.monotonic()
res = graph.intersect(regex)
closure = transitive_closure_mp(res)
end = time.monotonic()
sum += end - start
mp_pairs = closure.nvals
mp_time = sum / 5
start = time.monotonic()
for (value, matrix) in res.matrices.items():
v, n = value, matrix.nvals
end = time.monotonic()
mp_pairs_time = end - start
if (sq_pairs != mp_pairs):
print('{}, {}, {}'.format(sq_pairs, mp_pairs, filename))
assert sq_pairs == mp_pairs
output.write(
'{:s},{:s},{:d},{:.3f},{:.3f},{:d},{:.3f},{:.3f}\r\n'.format(
test, filename, sq_pairs, sq_time, sq_pairs_time, mp_pairs,
mp_time, mp_pairs_time))
output.close()
def cfpq_benchmark():
path_to_data = os.path.join(os.getcwd(), 'benchmarks/dataForCFPQ')
tests = ['FullGraph', 'MemoryAliases', 'WorstCase']
for test in tests:
path_to_test = os.path.join(path_to_data, test)
output = open('{}/output.csv'.format(path_to_test), 'w+')
output.write('test,graph,grammar,algo_name,algo_time\r\n')
graph_dir = os.path.join(path_to_test, 'graphs')
# for graph_name in sorted(os.listdir(graph_dir), key=lambda s: int(s.split('_')[1])):
for graph_name in os.listdir(graph_dir):
graph_path = os.path.join(graph_dir, graph_name)
graph = BMGraph.from_edges_file(graph_path)
grammar_dir = os.path.join(path_to_test, 'grammars')
for grammar_name in os.listdir(grammar_dir):
grammar_path = os.path.join(grammar_dir, grammar_name)
grammar = GrammarAlgos.from_grammar_file(grammar_path)
start = time.monotonic()
hellings_res = GrammarAlgos.Hellings(grammar, graph)
end = time.monotonic()
algo_name = 'hellings'
algo_time = end - start
res_str = '{:s},{:s},{:s},{:s},{:.3f}\r\n'.format(
test, graph_name, grammar_name, algo_name, algo_time)
print(res_str)
output.write(res_str)
start = time.monotonic()
mult_res = GrammarAlgos.cfpq_matrix_multiplication(
grammar, graph)
end = time.monotonic()
algo_name = 'mult'
algo_time = end - start
res_str = '{:s},{:s},{:s},{:s},{:.3f}\r\n'.format(
test, graph_name, grammar_name, algo_name, algo_time)
print(res_str)
output.write(res_str)
start = time.monotonic()
tensor_res = GrammarAlgos.cfpq_tensor_product(grammar, graph)
end = time.monotonic()
algo_name = 'tensor'
algo_time = end - start
res_str = '{:s},{:s},{:s},{:s},{:.3f}\r\n'.format(
test, graph_name, grammar_name, algo_name, algo_time)
print(res_str)
output.write(res_str)
wcnf = GrammarAlgos.to_wcnf(grammar)
start = time.monotonic()
tensor_wcnf_res = GrammarAlgos.cfpq_tensor_product(wcnf, graph)
end = time.monotonic()
algo_name = 'tensor_wcnf'
algo_time = end - start
res_str = '{:s},{:s},{:s},{:s},{:.3f}\r\n'.format(
test, graph_name, grammar_name, algo_name, algo_time)
print(res_str)
output.write(res_str)
assert (hellings_res == mult_res)
assert (mult_res == tensor_res)
assert (tensor_res == tensor_wcnf_res)
output.close()