def test_three_hops(self):
g = create_test_graph(is_directed=True)
bfw = DirectedBreadthFirstNeighbours(g)
graph_nodes = g.node_ids_to_ilocs(list(g.nodes()))
for _ in range(50):
node = random.choice(graph_nodes)
in_size = [random.randint(0, 2) for _ in range(3)]
out_size = [random.randint(0, 2) for _ in range(3)]
subgraph = bfw.run(nodes=[node],
n=1,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == 1
assert len(subgraph[0]) == 15 # 2**(num_hops+1) - 1 = 15
assert len(subgraph[0][0]) == 1
assert len(subgraph[0][1]) == in_size[0]
assert len(subgraph[0][2]) == out_size[0]
assert len(subgraph[0][3]) == in_size[0] * in_size[1]
assert len(subgraph[0][4]) == in_size[0] * out_size[1]
assert len(subgraph[0][5]) == out_size[0] * in_size[1]
assert len(subgraph[0][6]) == out_size[0] * out_size[1]
assert len(subgraph[0][7]) == in_size[0] * in_size[1] * in_size[2]
assert len(subgraph[0][8]) == in_size[0] * in_size[1] * out_size[2]
assert len(subgraph[0][9]) == in_size[0] * out_size[1] * in_size[2]
assert len(
subgraph[0][10]) == in_size[0] * out_size[1] * out_size[2]
assert len(
subgraph[0][11]) == out_size[0] * in_size[1] * in_size[2]
assert len(
subgraph[0][12]) == out_size[0] * in_size[1] * out_size[2]
assert len(
subgraph[0][13]) == out_size[0] * out_size[1] * in_size[2]
assert len(
subgraph[0][14]) == out_size[0] * out_size[1] * out_size[2]
def test_two_hops(self):
g = create_simple_graph()
bfw = DirectedBreadthFirstNeighbours(g)
# - The following case should be [[["root"], [None], [child*2], [None], [None*2], ["root"*2], [grandchild*4]]]
in_size = [1, 1]
out_size = [2, 2]
subgraph = bfw.run(nodes=["root"],
n=1,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == 1
assert len(subgraph[0]) == 7
assert len(subgraph[0][1]) == in_size[0]
assert subgraph[0][1][0] is None
assert len(subgraph[0][2]) == out_size[0]
for child in subgraph[0][2]:
assert child in ["0", 1, 2]
assert len(subgraph[0][3]) == in_size[0] * in_size[1]
assert subgraph[0][3][0] is None
assert len(subgraph[0][4]) == in_size[0] * out_size[1]
for child in subgraph[0][4]:
assert child is None
assert len(subgraph[0][5]) == out_size[0] * in_size[1]
for parent in subgraph[0][5]:
assert parent == "root"
assert len(subgraph[0][6]) == out_size[0] * out_size[1]
for grandchild in subgraph[0][6]:
assert grandchild in [None, "c1.1", "c2.1", "c2.2"]
for idx, child in enumerate(subgraph[0][2]):
grandchildren = subgraph[0][6][(2 * idx):(2 * idx + 2)]
if child == "0":
for grandchild in grandchildren:
assert grandchild is None
elif child == 1:
for grandchild in grandchildren:
assert grandchild == "c1.1"
else: # child == 2
for grandchild in grandchildren:
assert grandchild in ["c2.1", "c2.2"]
# - Check structure size for multiple start nodes
# - For each start node, should be [[[node], [in], [out], [in.in], [in.out], [out.in], [out.out]]]
nodes = list(g.nodes())
in_size = [2, 3]
out_size = [4, 5]
subgraph = bfw.run(nodes=nodes,
n=1,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == len(nodes)
for node_graph in subgraph:
assert len(node_graph) == 7
assert len(node_graph[0]) == 1 # 1 start node
assert len(node_graph[1]) == in_size[0]
assert len(node_graph[2]) == out_size[0]
assert len(node_graph[3]) == in_size[0] * in_size[1]
assert len(node_graph[4]) == in_size[0] * out_size[1]
assert len(node_graph[5]) == out_size[0] * in_size[1]
assert len(node_graph[6]) == out_size[0] * out_size[1]
def test_benchmark_bfs_walk(self, benchmark):
g = example_graph_random(n_nodes=100, n_edges=500, is_directed=True)
bfw = DirectedBreadthFirstNeighbours(g)
nodes = np.arange(0, 50)
n = 5
in_size = [5, 5]
out_size = [5, 5]
benchmark(lambda: bfw.run(
nodes=nodes, n=n, in_size=in_size, out_size=out_size))
def test_benchmark_bfs_walk(self, benchmark):
g = create_test_graph(is_directed=True)
bfw = DirectedBreadthFirstNeighbours(g)
nodes = ["0"]
n = 5
in_size = [5, 5]
out_size = [5, 5]
benchmark(lambda: bfw.run(
nodes=nodes, n=n, in_size=in_size, out_size=out_size))
def test_weighted(self):
g, checker = weighted_tree(is_directed=True)
bfw = DirectedBreadthFirstNeighbours(g)
walks = bfw.run(nodes=[0],
n=10,
in_size=[20, 20],
out_size=[20, 20],
weighted=True)
checker(node_id for walk in walks for hop in walk for node_id in hop)
def test_weighted_all_zero(self):
edges = pd.DataFrame({
"source": [0, 0],
"target": [1, 2],
"weight": [0.0, 0]
})
g = StellarDiGraph(edges=edges)
bfw = DirectedBreadthFirstNeighbours(g)
walks = bfw.run(nodes=[0],
n=10,
in_size=[20, 20],
out_size=[20, 20],
weighted=True)
assert len(walks) == 10
for walk in walks:
assert len(walk) == 7
assert walk[0] == [0]
for hop in walk:
np.testing.assert_array_equal(hop[1:], -1)
def test_one_hop(self, tree_graph):
bfw = DirectedBreadthFirstNeighbours(tree_graph)
# - The following case should be [[["root"], [None], [child]]]
nodes = tree_graph.node_ids_to_ilocs(["root"])
subgraph = bfw.run(nodes=nodes, n=1, in_size=[1], out_size=[1])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][0]) == 1
assert subgraph[0][0][0] == tree_graph.node_ids_to_ilocs(["root"])[0]
assert len(subgraph[0][1]) == 1
assert subgraph[0][1][0] == -1
assert len(subgraph[0][2]) == 1
assert subgraph[0][2][0] in tree_graph.node_ids_to_ilocs(["0", 1, 2])
# - The following case should be [[["root"], [None], []]]
subgraph = bfw.run(nodes=nodes, n=1, in_size=[1], out_size=[0])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][2]) == 0
# - The following case should be [[["root"], [], []]]
subgraph = bfw.run(nodes=nodes, n=1, in_size=[0], out_size=[0])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][1]) == 0
# - The following case should be [[["root"], [None, None, None], [child, child, child, child]]]
subgraph = bfw.run(nodes=nodes, n=1, in_size=[3], out_size=[4])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][1]) == 3
for child in subgraph[0][1]:
assert child == -1
assert len(subgraph[0][2]) == 4
for child in subgraph[0][2]:
assert child in tree_graph.node_ids_to_ilocs(["0", 1, 2])
def test_one_hop(self):
g = create_simple_graph()
bfw = DirectedBreadthFirstNeighbours(g)
# - The following case should be [[["root"], [None], [child]]]
subgraph = bfw.run(nodes=["root"], n=1, in_size=[1], out_size=[1])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][0]) == 1
assert subgraph[0][0][0] == "root"
assert len(subgraph[0][1]) == 1
assert subgraph[0][1][0] is None
assert len(subgraph[0][2]) == 1
assert subgraph[0][2][0] in ["0", 1, 2]
# - The following case should be [[["root"], [None], []]]
subgraph = bfw.run(nodes=["root"], n=1, in_size=[1], out_size=[0])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][2]) == 0
# - The following case should be [[["root"], [], []]]
subgraph = bfw.run(nodes=["root"], n=1, in_size=[0], out_size=[0])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][1]) == 0
# - The following case should be [[["root"], [None, None, None], [child, child, child, child]]]
subgraph = bfw.run(nodes=["root"], n=1, in_size=[3], out_size=[4])
assert len(subgraph) == 1
assert len(subgraph[0]) == 3
assert len(subgraph[0][1]) == 3
for child in subgraph[0][1]:
assert child is None
assert len(subgraph[0][2]) == 4
for child in subgraph[0][2]:
assert child in ["0", 1, 2]
def test_parameter_checking(self):
g = create_test_graph(is_directed=True)
bfw = DirectedBreadthFirstNeighbours(g)
nodes = g.node_ids_to_ilocs(["0", 1])
n = 1
in_size = [1]
out_size = [1]
with pytest.raises(ValueError):
# nodes should be a list of node ids even for a single node
bfw.run(nodes=None, n=n, in_size=in_size, out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=0, n=n, in_size=in_size, out_size=out_size)
# n has to be positive integer
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=-1, in_size=in_size, out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=10.1, in_size=in_size, out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=0, in_size=in_size, out_size=out_size)
# sizes have to be list of non-negative integers
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=0, out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=in_size, out_size=[-5])
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=-1, in_size=[2.4], out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=in_size, out_size=(1, 2))
# sizes have to have same length
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=[1, 2], out_size=[3])
# okay to have zero sized samples
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=in_size, out_size=[0, 0])
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=in_size, out_size=[1, 0])
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=in_size, out_size=[0, 5])
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=[0, 0], out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=[1, 0], out_size=out_size)
with pytest.raises(ValueError):
bfw.run(nodes=nodes, n=n, in_size=[0, 5], out_size=out_size)
# Is it okay if a zero appears in the same place for both in and out sizes? For now, yes.
subgraph = bfw.run(nodes=nodes, n=n, in_size=[5, 0], out_size=[1, 0])
assert len(subgraph) == len(nodes)
# seed must be positive integer or 0
with pytest.raises(ValueError):
bfw.run(nodes=nodes,
n=n,
in_size=in_size,
out_size=out_size,
seed=-1235)
with pytest.raises(ValueError):
bfw.run(nodes=nodes,
n=n,
in_size=in_size,
out_size=out_size,
seed=10.987665)
with pytest.raises(ValueError):
bfw.run(nodes=nodes,
n=n,
in_size=in_size,
out_size=out_size,
seed=-982.4746)
with pytest.raises(ValueError):
bfw.run(
nodes=nodes,
n=n,
in_size=in_size,
out_size=out_size,
seed="don't be random",
)
# If no root nodes are given, an empty list is returned, which is not an error.
nodes = []
subgraph = bfw.run(nodes=nodes,
n=n,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == 0
def test_two_hops(self, tree_graph):
bfw = DirectedBreadthFirstNeighbours(tree_graph)
# - The following case should be [[["root"], [None], [child*2], [None], [None*2], ["root"*2], [grandchild*4]]]
in_size = [1, 1]
out_size = [2, 2]
node_ilocs = tree_graph.node_ids_to_ilocs(["root"])
subgraph = bfw.run(nodes=node_ilocs,
n=1,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == 1
assert len(subgraph[0]) == 7
assert len(subgraph[0][1]) == in_size[0]
assert subgraph[0][1][0] == -1
assert len(subgraph[0][2]) == out_size[0]
for child in subgraph[0][2]:
assert child in tree_graph.node_ids_to_ilocs(["0", 1, 2])
assert len(subgraph[0][3]) == in_size[0] * in_size[1]
assert subgraph[0][3][0] == -1
assert len(subgraph[0][4]) == in_size[0] * out_size[1]
for child in subgraph[0][4]:
assert child == -1
assert len(subgraph[0][5]) == out_size[0] * in_size[1]
for parent in subgraph[0][5]:
assert parent == tree_graph.node_ids_to_ilocs(["root"])[0]
assert len(subgraph[0][6]) == out_size[0] * out_size[1]
for grandchild in subgraph[0][6]:
assert (grandchild in tree_graph.node_ids_to_ilocs(
["c1.1", "c2.1", "c2.2"]) or grandchild == -1)
for idx, child in enumerate(subgraph[0][2]):
grandchildren = subgraph[0][6][(2 * idx):(2 * idx + 2)]
if child == tree_graph.node_ids_to_ilocs(["0"])[0]:
for grandchild in grandchildren:
assert grandchild == -1
elif child == tree_graph.node_ids_to_ilocs([1])[0]:
for grandchild in grandchildren:
assert grandchild == tree_graph.node_ids_to_ilocs(["c1.1"
])[0]
else: # child == 2
for grandchild in grandchildren:
assert grandchild in tree_graph.node_ids_to_ilocs(
["c2.1", "c2.2"])
# - Check structure size for multiple start nodes
# - For each start node, should be [[[node], [in], [out], [in.in], [in.out], [out.in], [out.out]]]
nodes = tree_graph.node_ids_to_ilocs(list(tree_graph.nodes()))
in_size = [2, 3]
out_size = [4, 5]
subgraph = bfw.run(nodes=nodes,
n=1,
in_size=in_size,
out_size=out_size)
assert len(subgraph) == len(nodes)
for node_graph in subgraph:
assert len(node_graph) == 7
assert len(node_graph[0]) == 1 # 1 start node
assert len(node_graph[1]) == in_size[0]
assert len(node_graph[2]) == out_size[0]
assert len(node_graph[3]) == in_size[0] * in_size[1]
assert len(node_graph[4]) == in_size[0] * out_size[1]
assert len(node_graph[5]) == out_size[0] * in_size[1]
assert len(node_graph[6]) == out_size[0] * out_size[1]
def test_zero_hops(self, tree_graph):
bfw = DirectedBreadthFirstNeighbours(tree_graph)
# By consensus, a zero-length walk raises an error.
with pytest.raises(ValueError):
subgraph = bfw.run(nodes=["root"], n=1, in_size=[], out_size=[])
