def test_node_pickle(self):
"""
Test cases to check that nodes can be accurately
pickled and un-pickled.
"""
import pickle
# Node definitions
node = TransformationProxy(Node(None, None)) # Head node
n1 = node.Define("a", b="c") # First child node
n2 = n1.Count() # noqa: avoid PEP8 F841
n3 = node.Filter("b")
n4 = n3.Count() # noqa: avoid PEP8 F841
# Pickled representation of nodes
pickled_node = pickle.dumps(node.proxied_node)
# n3 is of class TransformationProxy, so the proxied node must be
# accessed before pickling.
pickled_n3_node = pickle.dumps(n3.proxied_node)
# Un-pickled node objects
unpickled_node = pickle.loads(pickled_node)
unpickled_n3_node = pickle.loads(pickled_n3_node)
self.assertIsInstance(unpickled_node, type(node.proxied_node))
self.assertIsInstance(unpickled_n3_node, type(n3.proxied_node))
self.assertGraphs(node, unpickled_node)
self.assertGraphs(n3.proxied_node, unpickled_n3_node)
def test_mapper_with_pruning(self):
"""
A test case to check that the mapper works even in the case of
pruning.
"""
# A mock RDF object
t = CallableGeneratorTest.Temp()
# Head node
node = TransformationProxy(Node(None, None))
# Set of operations to build the graph
n1 = node.Define()
n2 = node.Filter().Filter()
n4 = n2.Count()
n5 = n1.Count()
n6 = node.Filter() # noqa: avoid PEP8 F841
# Reason for pruning (change of reference)
n5 = n1.Filter() # noqa: avoid PEP8 F841
# Generate and execute the mapper
generator = CallableGenerator(node.proxied_node)
mapper_func = generator.get_callable()
values = mapper_func(t)
nodes = generator.get_action_nodes()
reqd_order = [1, 2, 2, 2, 3, 2]
self.assertEqual(t.ord_list, reqd_order)
self.assertListEqual(nodes, [n4.proxied_node])
self.assertListEqual(values, [t])
def test_mapper_from_graph(self):
"""A simple test case to check the working of mapper."""
# A mock RDF object
t = CallableGeneratorTest.Temp()
# Head node
node = TransformationProxy(Node(None, None))
# Set of operations to build the graph
n1 = node.Define()
n2 = node.Filter().Filter()
n4 = n2.Count()
n5 = n1.Count()
n6 = node.Filter() # noqa: avoid PEP8 F841
# Generate and execute the mapper
generator = CallableGenerator(node.proxied_node)
mapper_func = generator.get_callable()
values = mapper_func(t)
nodes = generator.get_action_nodes()
reqd_order = [1, 3, 2, 2, 3, 2]
self.assertEqual(t.ord_list, reqd_order)
self.assertListEqual(nodes, [n5.proxied_node, n4.proxied_node])
self.assertListEqual(values, [t, t])
def test_set_state(self):
"""
Test cases to check the working of
__setstate__ method on Node class.
"""
node = TransformationProxy(Node(None, None)) # Head node
n1 = TransformationProxy(Node(None, None)) # First child node
# State dictionaries
node_dict = {"children": [n1]}
n1_dict = {
'operation_name': "Define",
'operation_args': ["a"],
'operation_kwargs': {
"b": "c"
},
'children': []
}
# Set node objects with state dicts
node.proxied_node.__setstate__(node_dict)
n1.proxied_node.__setstate__(n1_dict)
self.assertListEqual([node.operation, node.children],
[None, node_dict["children"]])
self.assertListEqual([
n1.operation.name, n1.operation.args, n1.operation.kwargs,
n1.children
], [
n1_dict["operation_name"], n1_dict["operation_args"],
n1_dict["operation_kwargs"], n1_dict["children"]
])
def test_type_return_transformation(self):
"""
TransformationProxy object is of type `PyRDF.TransformationProxy` and
wraps a node object.
"""
node = Node(None, None)
proxy = TransformationProxy(node)
self.assertIsInstance(proxy, TransformationProxy)
self.assertIsInstance(proxy.proxied_node, Node)
def test_undefined_attr_transformation(self):
"""
When a non-defined Node class attribute is called on a
TransformationProxy object, it raises an AttributeError.
"""
node = Node(None, None)
proxy = TransformationProxy(node)
with self.assertRaises(AttributeError):
proxy.attribute
def test_proxied_node_has_user_references(self):
"""
Check that the user reference holds until the proxy lives. When the
Python garbage collector attempts to remove the proxy object, its
`__del__` method switches the node attribute `has_user_references` from
`True` to `False`.
"""
node = Node(None, None)
proxy = TransformationProxy(node)
self.assertTrue(node.has_user_references)
proxy = None # noqa: avoid PEP8 F841
self.assertFalse(node.has_user_references)
def test_dfs_graph_with_pruning_actions(self):
"""
Test case to check that action nodes with no user references get
pruned.
"""
# Head node
node = TransformationProxy(Node(None, None))
# Graph nodes
n1 = node.Define()
n2 = node.Filter()
n3 = n2.Filter()
n4 = n3.Count() # noqa: avoid PEP8 F841
n5 = n1.Count()
n6 = node.Filter() # noqa: avoid PEP8 F841
# Action pruning, n5 was an action node earlier
n5 = n1.Filter() # noqa: avoid PEP8 F841
obtained_order = DfsTest.traverse(node=node.get_head())
reqd_order = [1, 2, 2, 2, 3, 2]
self.assertEqual(obtained_order, reqd_order)
def test_dfs_graph_with_parent_pruning(self):
"""
Test case to check that parent nodes with no user references don't
get pruned.
"""
# Head node
node = TransformationProxy(Node(None, None))
# Graph nodes
n1 = node.Define()
n2 = node.Filter()
n3 = n2.Filter()
n4 = n3.Count() # noqa: avoid PEP8 F841
n5 = n1.Filter() # noqa: avoid PEP8 F841
n6 = node.Filter() # noqa: avoid PEP8 F841
# Remove references from n2 (which shouldn't affect the graph)
n2 = None
obtained_order = DfsTest.traverse(node=node.get_head())
reqd_order = [1, 2, 2, 2, 3, 2]
# Removing references from n2 will not prune any node
# because n2 still has children
self.assertEqual(obtained_order, reqd_order)
def test_dfs_graph_with_computed_values_pruning(self):
"""
Test case to check that computed values in action nodes get
pruned.
"""
# Head node
node = TransformationProxy(Node(None, None))
# Graph nodes
n1 = node.Define()
n2 = node.Filter()
n3 = n2.Filter()
n4 = n3.Count() # noqa: avoid PEP8 F841
n5 = n1.Filter()
n6 = n5.Count()
n7 = node.Filter()
# This is to make sure action nodes with
# already computed values are pruned.
n6.proxied_node.value = 1
# This is to make sure that transformation
# leaf nodes with value (possibly set intentionally)
# don't get pruned.
n7.value = 1 # noqa: avoid PEP8 F841
obtained_order = DfsTest.traverse(node=node.get_head())
# The node 'n6' will be pruned. Hence,
# there's only one '3' in this list.
reqd_order = [1, 2, 2, 2, 3, 2]
self.assertEqual(obtained_order, reqd_order)
def test_get_state(self):
"""
Test cases to check the working of __getstate__ method on
Node class.
"""
node = TransformationProxy(Node(None, None)) # Head node
n1 = node.Define("a", b="c") # First child node
# Required dictionaries
node_dict = {"children": [n1.proxied_node]}
n1_dict = {
'operation_name': "Define",
'operation_args': ["a"],
'operation_kwargs': {
"b": "c"
},
'children': []
}
# nodes are of class TransformationProxy, so the proxied nodes must be
# accessed in order to extract their dictionaries.
self.assertDictEqual(node.proxied_node.__getstate__(), node_dict)
self.assertDictEqual(n1.proxied_node.__getstate__(), n1_dict)
def __new__(cls, *args):
"""
Creates the head node of the graph with the arguments provided by the
user, then returns a proxy to that node.
Args:
*args (list): A list of arguments that were provided by the user
to construct the RDataFrame object.
"""
head_node = HeadNode(*args)
proxy_head = TransformationProxy(head_node)
# Logger debug statements
logger.debug("Created RDataFrame head node and proxy")
return proxy_head
def test_node_attr_transformation(self):
"""
When a node attribute is called on a TransformationProxy object, it
correctly returns the attribute of the proxied node.
"""
node = Node(None, None)
proxy = TransformationProxy(node)
node_attributes = [
"get_head", "operation", "children", "_new_op_name", "value",
"pyroot_node", "has_user_references"
]
for attr in node_attributes:
self.assertEqual(getattr(proxy, attr),
getattr(proxy.proxied_node, attr))
def test_other_dunder_methods(self):
"""
Test cases to check the working of other dunder methods on
Node class.
"""
node = TransformationProxy(Node(None, None))
# Regular dunder method must not throw an error
node.__format__('')
with self.assertRaises(AttributeError):
node.__random__() # Unknown dunder method
def test_supported_transformation(self):
"""
TransformationProxy object reads the right input attributes,
returning the methods of the proxied node.
"""
node = Node(None, None)
proxy = TransformationProxy(node)
transformations = {
"Define": ["x", "tdfentry_"],
"Filter": ["tdfentry_ > 0"],
"Range": ["tdfentry_"]
}
for transformation, args in transformations.items():
newProxy = getattr(proxy, transformation)(*args)
self.assertEqual(proxy.proxied_node._new_op_name, transformation)
self.assertIsInstance(newProxy, TransformationProxy)
self.assertEqual(newProxy.proxied_node.operation.name,
transformation)
self.assertEqual(newProxy.proxied_node.operation.args, args)
def test_dfs_graph_without_pruning(self):
"""
Test case to check that node pruning does not occur if every node either
has children or some user references.
"""
# Head node
node = TransformationProxy(Node(None, None))
# Graph nodes
n1 = node.Define()
n2 = node.Filter()
n3 = n2.Filter()
n4 = n3.Count() # noqa: avoid PEP8 F841
n5 = n1.Count() # noqa: avoid PEP8 F841
n6 = node.Filter() # noqa: avoid PEP8 F841
obtained_order = DfsTest.traverse(node=node.get_head())
reqd_order = [1, 3, 2, 2, 3, 2]
self.assertEqual(obtained_order, reqd_order)
def test_dfs_graph_with_recursive_pruning(self):
"""
Test case to check that nodes in a PyRDF graph with no user references
and no children get pruned recursively.
"""
# Head node
node = TransformationProxy(Node(None, None))
# Graph nodes
n1 = node.Define()
n2 = node.Filter()
n3 = n2.Filter()
n4 = n3.Count() # noqa: avoid PEP8 F841
n5 = n1.Filter() # noqa: avoid PEP8 F841
n6 = node.Filter() # noqa: avoid PEP8 F841
# Remove references from n4 and it's parent nodes
n4 = n3 = n2 = None # noqa: avoid PEP8 F841
obtained_order = DfsTest.traverse(node=node.get_head())
reqd_order = [1, 2, 2]
self.assertEqual(obtained_order, reqd_order)
def test_kwargs_read(self):
"""Named arguments are read accurately."""
node = TransformationProxy(Node(None, None))
newNode = node.Define(1, "b", a="1", b=2)
self.assertEqual(newNode.operation.kwargs, {"a": "1", "b": 2})
def test_transformation_proxy_return(self):
"""Node objects are returned for transformation nodes."""
node = TransformationProxy(Node(None, None))
newNode = node.Define(1)
self.assertIsInstance(newNode, TransformationProxy)
self.assertIsInstance(newNode.proxied_node, Node)
def test_action_proxy_return(self):
"""Proxy objects are returned for action nodes."""
node = TransformationProxy(Node(None, None))
newNode = node.Count()
self.assertIsInstance(newNode, ActionProxy)
self.assertIsInstance(newNode.proxied_node, Node)
def test_attr_read(self):
"""Function names are read accurately."""
node = TransformationProxy(Node(None, None))
func = node.Define # noqa: avoid PEP8 F841
self.assertEqual(node._new_op_name, "Define")
def test_args_read(self):
"""Arguments (unnamed) are read accurately."""
node = TransformationProxy(Node(None, None))
newNode = node.Define(1, "b", a="1", b=2)
self.assertEqual(newNode.operation.args, [1, "b"])
