def create_graph(session_id, graph_type, **kwargs):
"""Create an `CREATE_GRAPH` op, add op to default dag.
Args:
session_id (str): Refer to session that the graph will be create on.
graph_type (:enum:`GraphType`): GraphType defined in proto.types.proto.
**kwargs: additional properties respect to different `graph_type`.
Returns:
An op to create a graph in c++ side with necessary configurations.
"""
config = {
types_pb2.GRAPH_TYPE: utils.graph_type_to_attr(graph_type),
}
if graph_type == types_pb2.ARROW_PROPERTY:
attrs = kwargs.pop("attrs", None)
if attrs:
for k, v in attrs.items():
if isinstance(v, attr_value_pb2.AttrValue):
config[k] = v
elif graph_type == types_pb2.DYNAMIC_PROPERTY:
config[types_pb2.E_FILE] = utils.s_to_attr(kwargs["efile"])
config[types_pb2.V_FILE] = utils.s_to_attr(kwargs["vfile"])
config[types_pb2.DIRECTED] = utils.b_to_attr(kwargs["directed"])
else:
raise RuntimeError("Not supported graph type {}".format(graph_type))
op = Operation(session_id,
types_pb2.CREATE_GRAPH,
config=config,
output_types=types_pb2.GRAPH)
return op
def gremlin_to_subgraph(
interactive_query, gremlin_script, request_options=None, oid_type="int64"
):
"""Create a subgraph from gremlin output.
Args:
interactive_query (:class:`graphscope.interactive.query.InteractiveQueryDAGNode`):
The GIE instance holds the graph that gremlin query on.
gremlin_script (str):
gremlin script to be executed.
request_options (dict, optional): gremlin request options. format:
{
"engine": "gae"
}
oid_type (str, optional):
Type of vertex original id. Defaults to "int64".
Returns:
An op to create the subgraph from gremlin script
"""
config = {}
config[types_pb2.GIE_GREMLIN_QUERY_MESSAGE] = utils.s_to_attr(gremlin_script)
config[types_pb2.OID_TYPE] = utils.s_to_attr(oid_type)
if request_options:
config[types_pb2.GIE_GREMLIN_REQUEST_OPTIONS] = utils.s_to_attr(
json.dumps(request_options)
)
op = Operation(
interactive_query.session_id,
types_pb2.SUBGRAPH,
config=config,
inputs=[interactive_query.op],
output_types=types_pb2.GRAPH,
)
return op
def add_column(graph, results, selector):
"""Add a column to `graph`, produce a new graph.
Args:
graph (:class:`Graph`): Source ArrowProperty graph.
results (:class:`Context`): Results that generated by previous app querying.
selector (str): Used to select a subrange of data of results, add them
as one column of graph.
Returns:
A new graph with new columns added.
"""
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.GRAPH_TYPE: utils.graph_type_to_attr(graph.graph_type),
types_pb2.CTX_NAME: utils.s_to_attr(results.key),
types_pb2.SELECTOR: utils.s_to_attr(selector),
}
op = Operation(
graph._session_id,
types_pb2.ADD_COLUMN,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def copy_graph(graph, copy_type="identical"):
"""Create copy operation for nx graph.
Args:
graph (:class:`nx.Graph`): A nx graph.
copy_type (str): 'identical': copy graph to destination graph without any change.
'reverse': copy graph to destination graph with reversing the graph edges
Returns:
Operation
"""
check_argument(graph.graph_type in (graph_def_pb2.ARROW_PROPERTY,
graph_def_pb2.DYNAMIC_PROPERTY))
check_argument(copy_type in ("identical", "reverse"))
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.COPY_TYPE: utils.s_to_attr(copy_type),
}
op = Operation(
graph.session_id,
types_pb2.COPY_GRAPH,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def project_dynamic_property_graph(graph, v_prop, e_prop, v_prop_type, e_prop_type):
"""Create project graph operation for nx graph.
Args:
graph (:class:`nx.Graph`): A nx graph.
v_prop (str): The node attribute key to project.
e_prop (str): The edge attribute key to project.
v_prop_type (str): Type of the node attribute.
e_prop_type (str): Type of the edge attribute.
Returns:
Operation to project a dynamic property graph. Results in a simple graph.
"""
check_argument(graph.graph_type == types_pb2.DYNAMIC_PROPERTY)
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.GRAPH_TYPE: utils.graph_type_to_attr(types_pb2.DYNAMIC_PROJECTED),
types_pb2.V_PROP_KEY: utils.s_to_attr(v_prop),
types_pb2.E_PROP_KEY: utils.s_to_attr(e_prop),
types_pb2.V_DATA_TYPE: utils.s_to_attr(utils.data_type_to_cpp(v_prop_type)),
types_pb2.E_DATA_TYPE: utils.s_to_attr(utils.data_type_to_cpp(e_prop_type)),
}
op = Operation(
graph.session_id,
types_pb2.PROJECT_GRAPH,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def context_to_numpy(context, selector=None, vertex_range=None, axis=0):
"""Retrieve results as a numpy ndarray.
Args:
results (:class:`Context`): Results return by `run_app` operation, store the query results.
selector (str): Select the type of data to retrieve.
vertex_range (str): Specify a range to retrieve.
Returns:
An op to retrieve query results and convert to numpy ndarray.
"""
config = {}
if selector is not None:
config[types_pb2.SELECTOR] = utils.s_to_attr(selector)
if vertex_range is not None:
config[types_pb2.VERTEX_RANGE] = utils.s_to_attr(vertex_range)
if axis is not None:
config[types_pb2.AXIS] = utils.i_to_attr(axis)
op = Operation(
context.session_id,
types_pb2.CONTEXT_TO_NUMPY,
config=config,
inputs=[context.op],
output_types=types_pb2.TENSOR,
)
return op
def modify_edges(graph, modify_type, edges, attr={}, weight=None):
"""Create modify edges operation for nx graph.
Args:
graph (:class:`nx.Graph`): A nx graph.
modify_type (`type_pb2.(NX_ADD_EDGES | NX_DEL_EDGES | NX_UPDATE_EDGES)`): The modify type
edges (list): List of edges to be inserted into or delete from graph based on `modify_type`
Returns:
An op to modify edges on the graph.
"""
check_argument(graph.graph_type == graph_def_pb2.DYNAMIC_PROPERTY)
config = {}
config[types_pb2.GRAPH_NAME] = utils.s_to_attr(graph.key)
config[types_pb2.MODIFY_TYPE] = utils.modify_type_to_attr(modify_type)
config[types_pb2.PROPERTIES] = utils.s_to_attr(json.dumps(attr))
if weight:
config[types_pb2.EDGE_KEY] = utils.s_to_attr(weight)
op = Operation(
graph.session_id,
types_pb2.MODIFY_EDGES,
config=config,
large_attr=utils.bytes_to_large_attr(edges),
output_types=types_pb2.GRAPH,
)
return op
def arrow_to_dynamic(graph):
"""Transform a :class:`Graph` object to :class:`nx.Graph`.
Args:
graph (:class:`Graph`): Source graph, which type should be ARROW_PROPERTY.
Returns:
An op of transform arrow graph to dynamic graph with necessary configurations.
"""
check_argument(graph.graph_type == graph_def_pb2.ARROW_PROPERTY)
config = {
types_pb2.GRAPH_NAME:
utils.s_to_attr(graph.key),
types_pb2.GRAPH_TYPE:
utils.graph_type_to_attr(graph_def_pb2.ARROW_PROPERTY),
types_pb2.DST_GRAPH_TYPE:
utils.graph_type_to_attr(graph_def_pb2.DYNAMIC_PROPERTY),
types_pb2.OID_TYPE:
utils.s_to_attr(utils.data_type_to_cpp(graph.schema.oid_type)),
types_pb2.VID_TYPE:
utils.s_to_attr(utils.data_type_to_cpp(graph.schema.vid_type)),
types_pb2.DEFAULT_LABEL_ID:
utils.i_to_attr(graph._default_label_id),
}
op = Operation(
graph.session_id,
types_pb2.TRANSFORM_GRAPH,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def graph_to_dataframe(graph, selector=None, vertex_range=None):
"""Retrieve graph raw data as a pandas DataFrame.
Args:
graph (:class:`Graph`): Source graph.
selector (str): Select the type of data to retrieve.
vertex_range (str): Specify a range to retrieve.
Returns:
An op to convert a graph's data to pandas DataFrame.
"""
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
}
if selector is not None:
config[types_pb2.SELECTOR] = utils.s_to_attr(selector)
if vertex_range is not None:
config[types_pb2.VERTEX_RANGE] = utils.s_to_attr(vertex_range)
op = Operation(
graph._session_id,
types_pb2.GRAPH_TO_DATAFRAME,
config=config,
output_types=types_pb2.DATAFRAME,
)
return op
def to_vineyard_tensor(context, selector=None, vertex_range=None, axis=None):
"""Retrieve results as vineyard tensor.
Parameters:
results (:class:`Context`): Results return by `run_app` operation, store the query results.
selector (str): Select the type of data to retrieve.
vertex_range (str): Specify a range to retrieve.
Returns:
An op to convert query results into a vineyard tensor.
"""
config = {}
if selector is not None:
config[types_pb2.SELECTOR] = utils.s_to_attr(selector)
if vertex_range is not None:
config[types_pb2.VERTEX_RANGE] = utils.s_to_attr(vertex_range)
if axis is not None:
config[types_pb2.AXIS] = utils.i_to_attr(axis)
op = Operation(
context.session_id,
types_pb2.TO_VINEYARD_TENSOR,
config=config,
inputs=[context.op],
output_types=types_pb2.VINEYARD_TENSOR,
)
return op
def gremlin_query(interactive_query, query, request_options=None):
"""Execute a gremlin query.
Args:
interactive_query (:class:`graphscope.interactive.query.InteractiveQueryDAGNode`):
The GIE instance holds the graph that gremlin query on.
query (str):
Scripts that written in gremlin quering language.
request_options (dict, optional): gremlin request options. format:
{
"engine": "gae"
}
Returns:
An op to execute a gremlin query on the GIE instance.
"""
config = {}
config[types_pb2.GIE_GREMLIN_QUERY_MESSAGE] = utils.s_to_attr(query)
if request_options:
config[types_pb2.GIE_GREMLIN_REQUEST_OPTIONS] = utils.s_to_attr(
json.dumps(request_options))
op = Operation(
interactive_query.session_id,
types_pb2.GREMLIN_QUERY,
config=config,
inputs=[interactive_query.op],
output_types=types_pb2.GREMLIN_RESULTS,
)
return op
def report_graph(
graph, report_type, node=None, edge=None, fid=None, lid=None, key=None
):
"""Create report operation for nx graph.
This operation is used to simulate networkx graph reporting methods with variaty
report type and corresponding config parameters.
Args:
graph (`nx.Graph`): A nx graph.
report_type: report type, can be
type_pb2.(NODE_NUM,
EDGE_NUM,
HAS_NODE,
HAS_EDGE,
NODE_DATA,
EDGE_DATA,
NEIGHBORS_BY_NODE,
SUCCS_BY_NODE,
PREDS_BY_NODE,
NEIGHBORS_BY_LOC,
SUCCS_BY_LOC,
PREDS_BY_LOC,
DEG_BY_NODE,
IN_DEG_BY_NODE,
OUT_DEG_BY_NODE,
DEG_BY_LOC,
IN_DEG_BY_LOC,
OUT_DEG_BY_LOC,
NODES_BY_LOC)
node (str): node id, used as node id with 'NODE' report types. (optional)
edge (str): an edge with 'EDGE' report types. (optional)
fid (int): fragment id, with 'LOC' report types. (optional)
lid (int): local id of node in grape_engine, with 'LOC; report types. (optional)
key (str): edge key for MultiGraph or MultiDiGraph, with 'EDGE' report types. (optional)
Returns:
An op to do reporting job.
"""
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.REPORT_TYPE: utils.report_type_to_attr(report_type),
}
if node is not None:
config[types_pb2.NODE] = utils.s_to_attr(node)
if edge is not None:
config[types_pb2.EDGE] = utils.s_to_attr(edge)
if fid is not None:
config[types_pb2.FID] = utils.i_to_attr(fid)
if lid is not None:
config[types_pb2.LID] = utils.i_to_attr(lid)
config[types_pb2.EDGE_KEY] = utils.s_to_attr(str(key) if key is not None else "")
op = Operation(
graph.session_id,
types_pb2.REPORT_GRAPH,
config=config,
output_types=types_pb2.RESULTS,
)
return op
def output(context, fd, selector, vertex_range, **kwargs):
"""Output result to `fd`, this will be handled by registered vineyard C++ adaptor.
Args:
results (:class:`Context`): Results return by `run_app` operation, store the query results.
fd (str): Such as `file:///tmp/result_path`
selector (str): Select the type of data to retrieve.
vertex_range (str): Specify a range to retrieve.
Returns:
An op to output results to `fd`.
"""
config = {}
config[types_pb2.FD] = utils.s_to_attr(fd)
config[types_pb2.SELECTOR] = utils.s_to_attr(selector)
config[types_pb2.STORAGE_OPTIONS] = utils.s_to_attr(json.dumps(kwargs))
if vertex_range is not None:
config[types_pb2.VERTEX_RANGE] = utils.s_to_attr(vertex_range)
op = Operation(
context.session_id,
types_pb2.OUTPUT,
config=config,
inputs=[context.op],
output_types=types_pb2.NULL_OUTPUT,
)
return op
def context_to_dataframe(results, selector=None, vertex_range=None):
"""Retrieve results as a pandas DataFrame.
Args:
results (:class:`Context`): Results return by `run_app` operation, store the query results.
selector (str): Select the type of data to retrieve.
vertex_range (str): Specify a range to retrieve.
Returns:
An op to retrieve query results and convert to pandas DataFrame.
"""
config = {
types_pb2.CTX_NAME: utils.s_to_attr(results.key),
}
if selector is not None:
config[types_pb2.SELECTOR] = utils.s_to_attr(selector)
if vertex_range is not None:
config[types_pb2.VERTEX_RANGE] = utils.s_to_attr(vertex_range)
op = Operation(
results._session_id,
types_pb2.CONTEXT_TO_DATAFRAME,
config=config,
output_types=types_pb2.DATAFRAME,
)
return op
def project_arrow_property_graph_to_simple(
graph,
v_prop,
e_prop,
):
"""Project arrow property graph to a simple graph.
Args:
graph (:class:`graphscope.Graph`): Source graph, which type should be ARROW_PROPERTY
v_prop (str): The node attribute key to project.
e_prop (str): The edge attribute key to project.
Returns:
Operation to project a property graph, results in a simple ARROW_PROJECTED graph.
"""
check_argument(graph.graph_type == graph_def_pb2.ARROW_PROPERTY)
config = {
types_pb2.V_PROP_KEY: utils.s_to_attr(v_prop),
types_pb2.E_PROP_KEY: utils.s_to_attr(e_prop),
}
op = Operation(
graph.session_id,
types_pb2.PROJECT_TO_SIMPLE,
config=config,
inputs=[graph.op],
output_types=types_pb2.GRAPH,
)
return op
def run_app(graph, app, *args, **kwargs):
"""Run `app` on the `graph`.
Args:
graph (:class:`Graph`): A loaded graph.
app (:class:`App`): A loaded app that will be queried.
key (str): Key of query results, can be used to retrieve results.
*args: Additional query params that will be used in evaluation.
**kwargs: Key-value formated query params that mostly used in Cython apps.
Returns:
An op to run app on the specified graph, with optional query parameters.
"""
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.APP_NAME: utils.s_to_attr(app.key),
}
output_prefix = kwargs.pop("output_prefix", ".")
config[types_pb2.OUTPUT_PREFIX] = utils.s_to_attr(output_prefix)
# optional query arguments.
params = utils.pack_query_params(*args, **kwargs)
query_args = query_args_pb2.QueryArgs()
query_args.args.extend(params)
op = Operation(
graph._session_id,
types_pb2.RUN_APP,
config=config,
output_types=types_pb2.RESULTS,
query_args=query_args,
)
return op
def _get_config(
edges: Sequence[EdgeLabel],
vertices: Sequence[VertexLabel],
directed: bool,
oid_type: str,
generate_eid: bool,
) -> Dict:
config = {}
attr = attr_value_pb2.AttrValue()
for label in chain(edges, vertices):
label.finish(oid_type)
for edge in edges:
attr.list.func.extend([process_edge(edge)])
attr.list.func.extend([process_vertex(vertex) for vertex in vertices])
directed_attr = utils.b_to_attr(directed)
generate_eid_attr = utils.b_to_attr(generate_eid)
config[types_pb2.ARROW_PROPERTY_DEFINITION] = attr
config[types_pb2.DIRECTED] = directed_attr
config[types_pb2.OID_TYPE] = utils.s_to_attr(oid_type)
config[types_pb2.GENERATE_EID] = generate_eid_attr
# vid_type is fixed
config[types_pb2.VID_TYPE] = utils.s_to_attr("uint64_t")
config[types_pb2.IS_FROM_VINEYARD_ID] = utils.b_to_attr(False)
return config
def get_attr(self):
attr_list = attr_value_pb2.NameAttrList()
attr_list.name = "{}_{}".format(self.source_label,
self.destination_label)
attr_list.attr[types_pb2.SRC_LABEL].CopyFrom(
utils.s_to_attr(self.source_label))
attr_list.attr[types_pb2.DST_LABEL].CopyFrom(
utils.s_to_attr(self.destination_label))
attr_list.attr[types_pb2.LOAD_STRATEGY].CopyFrom(
utils.s_to_attr(self.load_strategy))
attr_list.attr[types_pb2.SRC_VID].CopyFrom(
utils.s_to_attr(str(self.source_vid)))
attr_list.attr[types_pb2.DST_VID].CopyFrom(
utils.s_to_attr(str(self.destination_vid)))
attr_list.attr[types_pb2.LOADER].CopyFrom(self.loader.get_attr())
props = []
for prop in self.properties[2:]:
prop_attr = attr_value_pb2.NameAttrList()
prop_attr.name = prop[0]
prop_attr.attr[0].CopyFrom(utils.type_to_attr(prop[1]))
props.append(prop_attr)
attr_list.attr[types_pb2.PROPERTIES].list.func.extend(props)
return attr_list
def attr(self) -> Sequence[attr_value_pb2.Chunk]:
chunk_list = []
for sub_label in self.sub_labels.values():
chunk = sub_label.get_attr()
chunk.attr[types_pb2.CHUNK_NAME].CopyFrom(utils.s_to_attr("edge"))
chunk.attr[types_pb2.CHUNK_TYPE].CopyFrom(
utils.s_to_attr("loader"))
chunk.attr[types_pb2.LABEL].CopyFrom(utils.s_to_attr(self.label))
chunk_list.append(chunk)
return chunk_list
def _construct_op_of_empty_graph(self):
config = {}
config[types_pb2.ARROW_PROPERTY_DEFINITION] = attr_value_pb2.AttrValue()
config[types_pb2.DIRECTED] = utils.b_to_attr(self._directed)
config[types_pb2.GENERATE_EID] = utils.b_to_attr(self._generate_eid)
config[types_pb2.OID_TYPE] = utils.s_to_attr(self._oid_type)
config[types_pb2.VID_TYPE] = utils.s_to_attr("uint64_t")
config[types_pb2.IS_FROM_VINEYARD_ID] = utils.b_to_attr(False)
return dag_utils.create_graph(
self.session_id, graph_def_pb2.ARROW_PROPERTY, inputs=None, attrs=config
)
def project_arrow_property_graph(
graph,
v_label_id,
v_prop_id,
e_label_id,
e_prop_id,
v_data_type,
e_data_type,
oid_type=None,
vid_type=None,
):
"""Project arrow property graph to a simple graph.
Args:
graph (:class:`Graph`): Source graph, which type should be ARROW_PROPERTY
dst_graph_key (str): The key of projected graph.
v_label_id (int): Label id of vertex used to project.
v_prop_id (int): Property id of vertex used to project.
e_label_id (int): Label id of edge used to project.
e_prop_id (int): Property id of edge used to project.
Returns:
An op to project `graph`, results in a simple ARROW_PROJECTED graph.
"""
check_argument(graph.graph_type == types_pb2.ARROW_PROPERTY)
config = {
types_pb2.GRAPH_NAME:
utils.s_to_attr(graph.key),
types_pb2.GRAPH_TYPE:
utils.graph_type_to_attr(types_pb2.ARROW_PROJECTED),
types_pb2.V_LABEL_ID:
utils.i_to_attr(v_label_id),
types_pb2.V_PROP_ID:
utils.i_to_attr(v_prop_id),
types_pb2.E_LABEL_ID:
utils.i_to_attr(e_label_id),
types_pb2.E_PROP_ID:
utils.i_to_attr(e_prop_id),
types_pb2.OID_TYPE:
utils.s_to_attr(oid_type),
types_pb2.VID_TYPE:
utils.s_to_attr(vid_type),
types_pb2.V_DATA_TYPE:
utils.s_to_attr(utils.data_type_to_cpp(v_data_type)),
types_pb2.E_DATA_TYPE:
utils.s_to_attr(utils.data_type_to_cpp(e_data_type)),
}
op = Operation(
graph._session_id,
types_pb2.PROJECT_GRAPH,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def _from_vineyard_id(self, vineyard_id):
config = {}
config[types_pb2.IS_FROM_VINEYARD_ID] = b_to_attr(True)
config[types_pb2.VINEYARD_ID] = i_to_attr(vineyard_id)
# FIXME(hetao) hardcode oid/vid type for codegen, when loading from vineyard
#
# the metadata should be retrived from vineyard
config[types_pb2.OID_TYPE] = s_to_attr("int64_t")
config[types_pb2.VID_TYPE] = s_to_attr("uint64_t")
op = create_graph(self._session_id, types_pb2.ARROW_PROPERTY, attrs=config)
graph_def = op.eval()
return graph_def
def _from_vineyard_name(self, vineyard_name):
config = {}
config[types_pb2.IS_FROM_VINEYARD_ID] = utils.b_to_attr(True)
config[types_pb2.VINEYARD_NAME] = utils.s_to_attr(str(vineyard_name))
# FIXME(hetao) hardcode oid/vid type for codegen, when loading from vineyard
#
# the metadata should be retrived from vineyard
config[types_pb2.OID_TYPE] = utils.s_to_attr("int64_t")
config[types_pb2.VID_TYPE] = utils.s_to_attr("uint64_t")
return dag_utils.create_graph(self.session_id,
types_pb2.ARROW_PROPERTY,
attrs=config)
def get_context_data(results, node):
config = {
types_pb2.CONTEXT_KEY: utils.s_to_attr(results.key),
types_pb2.NODE: utils.s_to_attr(node),
}
op = Operation(
results._session_id,
types_pb2.GET_CONTEXT_DATA,
config=config,
output_types=types_pb2.RESULTS,
)
return op
def _construct_op_from_vineyard_id(self, vineyard_id):
assert self._session is not None
config = {}
config[types_pb2.IS_FROM_VINEYARD_ID] = utils.b_to_attr(True)
config[types_pb2.VINEYARD_ID] = utils.i_to_attr(int(vineyard_id))
# FIXME(hetao) hardcode oid/vid type for codegen, when loading from vineyard
#
# the metadata should be retrived from vineyard
config[types_pb2.OID_TYPE] = utils.s_to_attr("int64_t")
config[types_pb2.VID_TYPE] = utils.s_to_attr("uint64_t")
return dag_utils.create_graph(
self.session_id, graph_def_pb2.ARROW_PROPERTY, attrs=config
)
def get_attr(self):
if not self.finished:
self.finish()
attr = attr_value_pb2.AttrValue()
attr.func.name = "loader"
attr.func.attr[types_pb2.PROTOCOL].CopyFrom(utils.s_to_attr(self.protocol))
# Let graphscope handle local files cause it's implemented in c++ and
# doesn't add an additional stream layer.
# Maybe handled by vineyard in the near future
if self.protocol == "file":
source = "{}#{}".format(self.source, self.options)
attr.func.attr[types_pb2.VALUES].CopyFrom(
utils.bytes_to_attr(source.encode("utf-8"))
)
elif self.protocol in ("numpy", "pandas"):
attr.func.attr[types_pb2.ROW_NUM].CopyFrom(utils.i_to_attr(self.row_num))
attr.func.attr[types_pb2.COLUMN_NUM].CopyFrom(
utils.i_to_attr(self.column_num)
)
# Use key start from 10000 + col_index to store raw bytes.
for i in range(len(self.property_bytes)):
attr.func.attr[10000 + i].CopyFrom(
utils.bytes_to_attr(self.property_bytes[i])
)
else: # Let vineyard handle other data source.
attr.func.attr[types_pb2.VALUES].CopyFrom(
utils.bytes_to_attr(self.source.encode("utf-8"))
)
return attr
def add_edges(graph, **kwargs):
"""Create an `ADD_EDGES` op, add op to default dag.
Args:
graph (:class:`Graph`): a :class:`Graph` instance.
**kwargs: additional properties respect to different `graph_type`.
Returns:
An op to add edges to a graph in c++ side with necessary configurations.
"""
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
types_pb2.GRAPH_TYPE: utils.graph_type_to_attr(graph.graph_type),
}
if graph.graph_type == types_pb2.ARROW_PROPERTY:
attrs = kwargs.pop("attrs", None)
if attrs:
for k, v in attrs.items():
if isinstance(v, attr_value_pb2.AttrValue):
config[k] = v
else:
raise ValueError(f"Not support add edges on graph type {graph.graph_type}")
op = Operation(
graph.session_id,
types_pb2.ADD_EDGES,
config=config,
output_types=types_pb2.GRAPH,
)
return op
def run_app(app, *args, **kwargs):
"""Run `bound app` on the `graph`.
Args:
app (:class:`AppDAGNode`): A :class:`AppDAGNode` instance which represent a bound app.
key (str): Key of query results, can be used to retrieve results.
*args: Additional query params that will be used in evaluation.
**kwargs: Key-value formated query params that mostly used in Cython apps.
Returns:
An op to run app on the specified graph, with optional query parameters.
"""
inputs = [app.op]
config = {}
output_prefix = kwargs.pop("output_prefix", ".")
config[types_pb2.OUTPUT_PREFIX] = utils.s_to_attr(output_prefix)
# optional query arguments.
params = utils.pack_query_params(*args, **kwargs)
query_args = query_args_pb2.QueryArgs()
query_args.args.extend(params)
op = Operation(
app.session_id,
types_pb2.RUN_APP,
inputs=inputs,
config=config,
output_types=types_pb2.RESULTS,
query_args=query_args,
)
return op
def create_interactive_query(graph, engine_params):
"""Create a interactive engine that query on the :code:`graph`
Args:
graph (:class:`graphscope.framework.graph.GraphDAGNode`):
Source property graph.
engine_params (dict, optional):
Configuration to startup the interactive engine. See detail in:
`interactive_engine/deploy/docker/dockerfile/executor.vineyard.properties`
Returns:
An op to create a interactive engine based on a graph.
"""
config = {}
if engine_params is not None:
config[types_pb2.GIE_GREMLIN_ENGINE_PARAMS] = utils.s_to_attr(
json.dumps(engine_params))
op = Operation(
graph.session_id,
types_pb2.CREATE_INTERACTIVE_QUERY,
config=config,
inputs=[graph.op],
output_types=types_pb2.INTERACTIVE_QUERY,
)
return op
def create_subgraph(graph, nodes=None, edges=None):
"""Create subgraph operation for nx graph.
Args:
graph (:class:`nx.Graph`): A nx graph.
nodes (list): the nodes to induce a subgraph.
edges (list): the edges to induce a edge-induced subgraph.
Returns:
Operation
"""
check_argument(graph.graph_type == graph_def_pb2.DYNAMIC_PROPERTY)
config = {
types_pb2.GRAPH_NAME: utils.s_to_attr(graph.key),
}
if nodes is not None:
config[types_pb2.NODES] = utils.bytes_to_attr(nodes)
if edges is not None:
config[types_pb2.EDGES] = utils.bytes_to_attr(edges)
op = Operation(
graph.session_id,
types_pb2.INDUCE_SUBGRAPH,
config=config,
output_types=types_pb2.GRAPH,
)
return op