def check_edges(in_nodes, out_nodes, edge_type):
for in_node in in_nodes:
for out_node in out_nodes:
edge = RelationGraphEdge(in_node, out_node, edge_type)
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
def test_max_series(self):
input_df = pd.DataFrame([[1, 2], [2, 3], [2, 0]])
input_00 = RelationGraphNode("I0", (0, 0),
get_node_type(input_df.iat[0, 0]))
input_01 = RelationGraphNode("I0", (0, 1),
get_node_type(input_df.iat[0, 1]))
input_10 = RelationGraphNode("I0", (1, 0),
get_node_type(input_df.iat[1, 0]))
input_11 = RelationGraphNode("I0", (1, 1),
get_node_type(input_df.iat[1, 1]))
input_20 = RelationGraphNode("I0", (2, 0),
get_node_type(input_df.iat[2, 0]))
input_21 = RelationGraphNode("I0", (2, 1),
get_node_type(input_df.iat[2, 1]))
output = pd.DataFrame.max(input_df)
output_00 = RelationGraphNode("O0", (0, 0),
get_node_type(output.iat[0]))
output_10 = RelationGraphNode("O0", (1, 0),
get_node_type(output.iat[1]))
options = RelationGraphOptions()
options.NODE_TYPES = True
rel_graph: RelationGraph = RelationGraph.build_relation_graph(
[input_df], output, options)
rel_graph_edges = rel_graph.edges
# positional edges
positional_edges = [
RelationGraphEdge(input_00, input_01,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_00, input_10,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_10, input_11,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_10, input_20,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_20, input_21,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_01, input_11,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_11, input_21,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(output_00, output_10,
RelationGraphEdgeType.ADJACENCY)
]
for edge in positional_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
# equality edges
equality_edges = [
RelationGraphEdge(input_10, output_00,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(input_20, output_00,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(input_01, output_00,
RelationGraphEdgeType.EQUALITY), # redundant
RelationGraphEdge(input_11, output_10,
RelationGraphEdgeType.EQUALITY)
]
for edge in equality_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
def test_idx_multi(self):
tuples = [("bar", "one"), ("bar", "two")]
index = pd.MultiIndex.from_tuples(tuples)
data = [[0], [1]]
input_df = pd.DataFrame(data, index=index)
# 0
# bar one 0
# two 1
output_df = input_df.unstack()
# 0
# one two
# bar 0 1
options = RelationGraphOptions()
options.COLUMN_NODES = True
options.INDEX_NODES = True
options.ADJACENCY_EDGES = True
options.EQUALITY_EDGES = True
options.NODE_TYPES = True
options.INDEX_EDGES = True
rel_graph: RelationGraph = RelationGraph.build_relation_graph(
[input_df], output_df, options)
rel_graph_edges = rel_graph.edges
bar_in_0 = RelationGraphNode("I0", (0, -2),
RelationGraphNodeType.INDEX)
bar_in_1 = RelationGraphNode("I0", (1, -2),
RelationGraphNodeType.INDEX)
bar_out = RelationGraphNode("O0", (0, -1), RelationGraphNodeType.INDEX)
one_in = RelationGraphNode("I0", (0, -1), RelationGraphNodeType.INDEX)
two_in = RelationGraphNode("I0", (1, -1), RelationGraphNodeType.INDEX)
one_out = RelationGraphNode("O0", (-1, 0),
RelationGraphNodeType.COLUMN)
two_out = RelationGraphNode("O0", (-1, 1),
RelationGraphNodeType.COLUMN)
in_0 = RelationGraphNode("I0", (0, 0), RelationGraphNodeType.INT)
in_1 = RelationGraphNode("I0", (1, 0), RelationGraphNodeType.INT)
out_0 = RelationGraphNode("O0", (0, 0), RelationGraphNodeType.INT)
out_1 = RelationGraphNode("O0", (0, 1), RelationGraphNodeType.INT)
adjacency_edges = [
RelationGraphEdge(bar_in_0, bar_in_1,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(bar_in_0, one_in,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(bar_in_1, two_in,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(one_in, two_in, RelationGraphEdgeType.ADJACENCY)
]
for edge in adjacency_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
indexing_edges = [
RelationGraphEdge(bar_in_0, in_0, RelationGraphEdgeType.INDEX),
RelationGraphEdge(one_in, in_0, RelationGraphEdgeType.INDEX),
RelationGraphEdge(bar_in_1, in_1, RelationGraphEdgeType.INDEX),
RelationGraphEdge(two_in, in_1, RelationGraphEdgeType.INDEX),
RelationGraphEdge(bar_out, out_0, RelationGraphEdgeType.INDEX),
RelationGraphEdge(bar_out, out_1, RelationGraphEdgeType.INDEX)
]
for edge in indexing_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
equality_edges = [
RelationGraphEdge(bar_in_0, bar_out,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(bar_in_1, bar_out,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(one_in, one_out, RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(two_in, two_out, RelationGraphEdgeType.EQUALITY)
]
for edge in equality_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
def test_column_multi(self):
column_labels = [['bar', 'bar', 'baz', 'baz'],
['one', 'two', 'one', 'two']]
tuples = list(zip(*column_labels))
col_index = pd.MultiIndex.from_tuples(tuples)
data = [[0, 1, 2, 3], [4, 5, 6, 7]]
input_df = pd.DataFrame(data, columns=col_index)
# bar baz
# one two one two
# 0 0 1 2 3
# 1 4 5 6 7
output_df = input_df.stack().reset_index()
# level_0 level_1 bar baz
# 0 0 one 0 2
# 1 0 two 1 3
# 2 1 one 4 6
# 3 1 two 5 7
options = RelationGraphOptions()
options.COLUMN_NODES = True
options.ADJACENCY_EDGES = True
options.EQUALITY_EDGES = True
options.NODE_TYPES = True
options.INDEX_EDGES = True
rel_graph: RelationGraph = RelationGraph.build_relation_graph(
[input_df], output_df, options)
rel_graph_edges = rel_graph.edges
col_nodes = [
[
RelationGraphNode("I0", (-2, 0), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-2, 1), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-2, 2), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-2, 3), RelationGraphNodeType.COLUMN)
],
[
RelationGraphNode("I0", (-1, 0), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-1, 1), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-1, 2), RelationGraphNodeType.COLUMN),
RelationGraphNode("I0", (-1, 3), RelationGraphNodeType.COLUMN)
],
]
adjacency_edges = [
RelationGraphEdge(col_nodes[0][0], col_nodes[1][0],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][0], col_nodes[0][1],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[1][0], col_nodes[1][1],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[1][1], col_nodes[1][2],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][1], col_nodes[1][1],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][1], col_nodes[0][2],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][2], col_nodes[1][2],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][2], col_nodes[0][3],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[1][2], col_nodes[1][3],
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(col_nodes[0][3], col_nodes[1][3],
RelationGraphEdgeType.ADJACENCY)
]
for edge in adjacency_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
# indexing edges
input_coli_elems = [[
RelationGraphNode("I0", (0, 0), RelationGraphNodeType.INT),
RelationGraphNode("I0", (1, 0), RelationGraphNodeType.INT)
],
[
RelationGraphNode("I0", (0, 1),
RelationGraphNodeType.INT),
RelationGraphNode("I0", (1, 1),
RelationGraphNodeType.INT)
],
[
RelationGraphNode("I0", (0, 2),
RelationGraphNodeType.INT),
RelationGraphNode("I0", (1, 2),
RelationGraphNodeType.INT)
],
[
RelationGraphNode("I0", (0, 3),
RelationGraphNodeType.INT),
RelationGraphNode("I0", (1, 3),
RelationGraphNodeType.INT)
]]
def check_edges(in_nodes, out_nodes, edge_type):
for in_node in in_nodes:
for out_node in out_nodes:
edge = RelationGraphEdge(in_node, out_node, edge_type)
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
for i in range(4):
in_nodes = [col_nodes[0][i], col_nodes[1][i]]
out_nodes = input_coli_elems[i]
check_edges(in_nodes, out_nodes, RelationGraphEdgeType.INDEX)
# equality_edges
bars = [col_nodes[0][0], col_nodes[0][1]]
bazs = [col_nodes[0][2], col_nodes[0][3]]
ones = [col_nodes[1][0], col_nodes[1][2]]
twos = [col_nodes[1][1], col_nodes[1][3]]
out_01 = RelationGraphNode("O0", (0, 1), RelationGraphNodeType.STR)
out_11 = RelationGraphNode("O0", (1, 1), RelationGraphNodeType.STR)
out_21 = RelationGraphNode("O0", (2, 1), RelationGraphNodeType.STR)
out_31 = RelationGraphNode("O0", (3, 1), RelationGraphNodeType.STR)
out_col_2 = RelationGraphNode("O0", (-1, 2),
RelationGraphNodeType.COLUMN)
out_col_3 = RelationGraphNode("O0", (-1, 3),
RelationGraphNodeType.COLUMN)
check_edges(bars, [out_col_2], RelationGraphEdgeType.EQUALITY)
check_edges(bazs, [out_col_3], RelationGraphEdgeType.EQUALITY)
check_edges(ones, [out_01, out_21], RelationGraphEdgeType.EQUALITY)
check_edges(twos, [out_11, out_31], RelationGraphEdgeType.EQUALITY)
def test_dict(self):
input_df = pd.DataFrame([[1, 2], [3, 4]])
input_00 = RelationGraphNode("I0", (0, 0),
get_node_type(input_df.iat[0, 0]))
input_01 = RelationGraphNode("I0", (0, 1),
get_node_type(input_df.iat[0, 1]))
input_10 = RelationGraphNode("I0", (1, 0),
get_node_type(input_df.iat[1, 0]))
input_11 = RelationGraphNode("I0", (1, 1),
get_node_type(input_df.iat[1, 1]))
output = {"A": [1, 3], "B": [2, 4]}
output_00 = RelationGraphNode("O0", (0, 0),
get_node_type(output['A'][0]))
output_01 = RelationGraphNode("O0", (0, 1),
get_node_type(output['B'][0]))
output_10 = RelationGraphNode("O0", (1, 0),
get_node_type(output['A'][1]))
output_11 = RelationGraphNode("O0", (1, 1),
get_node_type(output['B'][1]))
options = RelationGraphOptions()
options.NODE_TYPES = True
rel_graph: RelationGraph = RelationGraph.build_relation_graph(
[input_df], output, options)
rel_graph_edges = rel_graph.edges
positional_edges = [
RelationGraphEdge(input_00, input_01,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_00, input_10,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_10, input_11,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(input_01, input_11,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(output_00, output_01,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(output_00, output_10,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(output_10, output_11,
RelationGraphEdgeType.ADJACENCY),
RelationGraphEdge(output_01, output_11,
RelationGraphEdgeType.ADJACENCY)
]
for edge in positional_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))
equality_edges = [
RelationGraphEdge(input_00, output_00,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(input_10, output_10,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(input_01, output_01,
RelationGraphEdgeType.EQUALITY),
RelationGraphEdge(input_11, output_11,
RelationGraphEdgeType.EQUALITY)
]
for edge in equality_edges:
self.assertTrue(
edge in rel_graph_edges,
"Could not find edge %s in set of edges:\n%s" %
(edge, rel_graph_edges))