Merge pull request #12 from Milittle/main

.gitignore

@@ -139,3 +139,6 @@ dmypy.json
 
 # dgl
 .dgl
+
+homo*
+multi*

README.md

@@ -221,3 +221,36 @@ networkx.degree(nx_g)
 # get degree centrality
 networkx.degree_centrality(nx_g)
 ```
+
+
+## Multi-Part Loader for Nebula Graph
+
+1. For now, the Multi-Part Loader is slow like sequence scan, need to profile the performance.
+
+```python
+import yaml
+import networkx as nx
+import matplotlib.pyplot as plt
+
+from nebula_dgl import NebulaReducedLoader
+
+
+nebula_config = {
+    "graph_hosts": [
+                ('127.0.0.1', 9669)
+            ],
+    "nebula_user": "root",
+    "nebula_password": "nebula",
+}
+
+with open('example/homogeneous_graph.yaml', 'r') as f:
+    feature_mapper = yaml.safe_load(f)
+
+# you only need change the following line: from NebulaLoader to NebulaReducedLoader
+# Easy for you to use the multi-part loader 
+nebula_reduced_loader = NebulaReducedLoader(nebula_config, feature_mapper)
+homo_dgl_graph = nebula_reduced_loader.load()
+nx_g = homo_dgl_graph.to_networkx()
+nx.draw(nx_g, with_labels=True, pos=nx.spring_layout(nx_g))
+plt.savefig("multi_graph.png")
+```

nebula_dgl/__init__.py

@@ -4,9 +4,13 @@
 # Export for plugin use
 from nebula_dgl.nebula_exporter import NebulaExporter
 from nebula_dgl.nebula_loader import NebulaLoader
+from nebula_dgl.nebula_part_loader import NebulaPartLoader
+from nebula_dgl.nebula_reduced_loader import NebulaReducedLoader
 
 
 __all__ = (
     "NebulaExporter",
-    "NebulaLoader"
+    "NebulaLoader",
+    "NebulaPartLoader",
+    "NebulaReducedLoader"
 )

nebula_dgl/nebula_part_loader.py

@@ -0,0 +1,214 @@
+from nebula3.mclient import MetaCache
+from nebula3.sclient.GraphStorageClient import GraphStorageClient
+
+from typing import Dict, List, Callable, Tuple
+
+import logging
+
+
+logger = logging.getLogger(__package__)
+ENUMERATION_DEFAULT_VALUE = -1
+NEBULA_TYPE_MAP = {
+    "int": "as_int",
+    "double": "as_double",
+    "str": "as_string",
+    "float": "as_double",
+}
+
+
+class NebulaPartLoader():
+    """
+    NebulaPartLoader is a class that loads vertex and edge data from Nebula Graph Cluster.
+
+    """
+
+    def __init__(self, part_id: int, meta_hosts: List, edge_feature_dict: Dict, tag_feature_dict: Dict):
+        """
+        Initialize the NebulaLoader class.
+        """
+        self.part_id = part_id
+        self.edge_feature_dict = edge_feature_dict
+        self.tag_feature_dict = tag_feature_dict
+        self.meta_hosts = meta_hosts
+
+    def get_storage_client(self):
+        """
+        Get the GraphStorageClient.
+        """
+        meta_cache = MetaCache(self.meta_hosts, 50000)
+        storage_client = GraphStorageClient(meta_cache)
+        return storage_client
+
+    def get_feature_transform_function(self, features: Dict, prop_names: List) -> Callable:
+        """
+        Get the transform function for all the features.
+        """
+        prop_pos_index = {prop_name: i for i,
+                          prop_name in enumerate(prop_names)}
+
+        def transform_function(prop_values):
+            ret_feature = []
+            for feature_name in features:
+                feature = features[feature_name]
+                feature_props = feature.get('prop')
+                if feature_props is None:
+                    import pdb
+                    pdb.set_trace()
+
+                feature_prop_names = [prop.get('name')
+                                      for prop in feature_props]
+                feature_prop_types = [prop.get('type')
+                                      for prop in feature_props]
+                feature_prop_values = []
+                for index, prop_name in enumerate(feature_prop_names):
+                    raw_value = prop_values[prop_pos_index[prop_name]]
+                    # convert byte value according to type
+                    feature_prop_values.append(
+                        getattr(
+                            raw_value, NEBULA_TYPE_MAP[feature_prop_types[index]])()
+                    )
+                if feature.get('type') == 'value':
+                    ret_feature.append(feature_prop_values[0])
+                elif feature.get('type') == 'enumeration':
+                    enumeration_dict = feature.get('enumeration')
+                    ret_feature.append(enumeration_dict.get(
+                        feature_prop_values[0], ENUMERATION_DEFAULT_VALUE))
+                elif feature.get('type') == 'function':
+                    feature_filter_function = feature.get('function')
+                    ret_feature.append(
+                        feature_filter_function(*feature_prop_values))
+            if len(ret_feature) == 0:
+                return None
+            else:
+                return ret_feature
+
+        return transform_function
+
+    def load_vertexes(self) -> Tuple[Dict, Dict]:
+        """
+        Load the part vertexes data from Nebula Graph Cluster, return vertex ids and ndata.
+        """
+        # generate vertices per tag
+        vertex_id_dict = dict()
+        ndata = dict()
+
+        # assumed only one graph space though, we iterate it here.
+        for space_name in self.tag_feature_dict:
+            if space_name not in vertex_id_dict:
+                vertex_id_dict[space_name] = dict()
+            for tag_name in self.tag_feature_dict[space_name]:
+                if tag_name not in vertex_id_dict[space_name]:
+                    vertex_id_dict[space_name][tag_name] = dict()
+                _vertex_id_dict = vertex_id_dict[space_name][tag_name]
+                tag_features = self.tag_feature_dict[space_name][tag_name]
+                props = set()
+                for feature_name in tag_features:
+                    feature = tag_features[feature_name]
+                    if feature_name not in ndata:
+                        ndata[feature_name] = {tag_name: []}
+                    else:
+                        assert tag_name not in ndata[feature_name], \
+                            f'tag {tag_name} already exists in ndata[{feature_name}]'
+                        ndata[feature_name][tag_name] = []
+                    for prop in feature.get('prop', []):
+                        props.add(prop['name'])
+                prop_names = list(props)
+
+                graph_storage_client = self.get_storage_client()
+                resp = graph_storage_client.scan_vertex_with_part(
+                    space_name=space_name,
+                    part=self.part_id,
+                    tag_name=tag_name,
+                    prop_names=prop_names)
+                vertex_index = 0
+                transform_function = self.get_feature_transform_function(
+                    tag_features, prop_names)
+                while resp.has_next():
+                    result = resp.next()
+                    for vertex_data in result:
+                        _vertex_id_dict[vertex_data.get_id()] = vertex_index
+                        vertex_index += 1
+                        # feature data for vertex(node)
+                        if not tag_features:
+                            continue
+                        prop_values = vertex_data.get_prop_values()
+                        feature_values = transform_function(prop_values)
+                        for index, feature_name in enumerate(tag_features):
+                            feature = tag_features[feature_name]
+                            if feature_name not in ndata:
+                                ndata[feature_name] = {tag_name: []}
+                            ndata[feature_name][tag_name].append(
+                                feature_values[index])
+                if prop_names:
+                    assert vertex_index == len(
+                        ndata[feature_name][tag_name]), \
+                        f'vertex index {vertex_index} != len(ndata[{prop_names[0]}][{tag_name}])'
+
+        return vertex_id_dict, ndata
+
+    def load_edges(self, vertex_id_dict) -> Tuple[Dict, Dict]:
+        """
+        Load the part edge data from Nebula Graph Cluster, return edge and edata
+        """
+        data_dict = dict()
+        edata = dict()
+        for space_name in self.edge_feature_dict:
+
+            for edge_name in self.edge_feature_dict[space_name]:
+                edge = self.edge_feature_dict[space_name][edge_name]
+                edge_features = edge.get('features', {})
+                start_vertex_tag, end_vertex_tag = edge.get(
+                    'start_vertex_tag'), edge.get('end_vertex_tag')
+                assert (start_vertex_tag, edge_name, end_vertex_tag) not in data_dict, \
+                    f'edge {start_vertex_tag}-{edge_name}-{end_vertex_tag} already exists'
+                props = set()
+                for feature_name in edge_features:
+                    feature = edge_features[feature_name]
+                    if feature_name not in edata:
+                        edata[feature_name] = {edge_name: []}
+                    else:
+                        assert edge_name not in edata[feature_name], \
+                            f'tag {edge_name} already exists in edata[{feature_name}]'
+                        edata[feature_name][edge_name] = []
+                    for prop in feature.get('prop', []):
+                        props.add(prop['name'])
+                prop_names = list(props)
+
+                graph_storage_client = self.get_storage_client()
+                resp = graph_storage_client.scan_edge_with_part(
+                    space_name=space_name,
+                    part=self.part_id,
+                    edge_name=edge_name,
+                    prop_names=prop_names)
+                transform_function = self.get_feature_transform_function(
+                    edge_features, prop_names)
+                start_vertices, end_vertices = [], []
+                start_vertex_id_dict = vertex_id_dict[space_name][start_vertex_tag]
+                end_vertex_id_dict = vertex_id_dict[space_name][end_vertex_tag]
+                while resp.has_next():
+                    result = resp.next()
+                    for edge_data in result:
+                        # edge data for edge
+                        start_vertices.append(
+                            start_vertex_id_dict[edge_data.get_src_id()]
+                        )
+                        end_vertices.append(
+                            end_vertex_id_dict[edge_data.get_dst_id()]
+                        )
+
+                        # feature data for edge
+                        if not edge_features:
+                            continue
+                        prop_values = edge_data.get_prop_values()
+                        feature_values = transform_function(prop_values)
+                        for index, feature_name in enumerate(edge_features):
+                            feature = edge_features[feature_name]
+                            if feature_name not in edata:
+                                edata[feature_name] = {edge_name: []}
+                            edata[feature_name][edge_name].append(
+                                feature_values[index])
+
+                data_dict[(start_vertex_tag, edge_name, end_vertex_tag)] = (
+                    start_vertices, end_vertices)
+
+        return data_dict, edata