load a services.Dataset into a Neo4J instance

.env.template

@@ -5,3 +5,4 @@ POSTGRES_DB=glossary
 API_KEY="123456789"
 DATABASE_URL="postgresql+psycopg://admin:123456789@postgres:5432/glossary"
 SENTRY_DSN="https://…"
+NEO4J_AUTH=

Makefile

@@ -0,0 +1,6 @@
+start-neo4j:
+	docker compose -f docker-compose.neo4j.yml --env-file .env up -d
+
+stop-neo4j:
+	docker compose -f docker-compose.neo4j.yml down
+

dds_glossary/neo.py

@@ -0,0 +1,167 @@
+import logging
+from typing import Any, Dict, List, Tuple
+
+from . import services
+
+import neo4j
+from owlready2 import get_ontology
+
+logging.basicConfig(format="%(asctime)s - %(message)s", level=logging.INFO)
+
+
+def load_services_dataset(driver: neo4j.Driver, dataset: services.Dataset, dataset_path: str, import_chunk_size: int=1000):
+    """
+    Load a given dataset into a Neo4J instance
+    """
+    ontology = get_ontology(dataset.url).load()
+    ontology.save(file=str(dataset_path), format="rdfxml")
+    concept_schemes_raw, concepts_raw, _, semantic_relations_raw = services.GlossaryController.parse_dataset(dataset_path)
+
+    logging.info("Loaded the dataset in memory")
+
+    concept_schemes = [c.to_dict() for c in concept_schemes_raw]
+    concepts = [c.to_dict() for c in concepts_raw]
+    semantic_relations = [c.to_dict() for c in semantic_relations_raw]
+
+    # One service.Dataset has exactly one concept scheme
+    main_concept_scheme = concept_schemes[0]
+
+    logging.info("Importing concept schemes...")
+
+    # Load concept schemes
+    for chunk in chunk_list(concept_schemes, import_chunk_size):
+        query, args = build_nodes_import_query_and_args(["ConceptScheme"], chunk)
+        driver.execute_query(query, args)
+
+    logging.info("Imported concept schemes")
+
+    logging.info("Importing concepts...")
+
+    # Load concepts
+    for chunk in chunk_list(concepts, import_chunk_size):
+        query, args = build_nodes_import_query_and_args(["Concept"], chunk)
+        driver.execute_query(query, args)
+
+    logging.info("Imported concept schemes")
+
+    logging.info("Adding indices...")
+
+    # Index by the IRI
+    for key, label in [("ConceptScheme", "iri"), ("Concept", "iri")]:
+        driver.execute_query(build_index_query(key=key, label=label))
+
+    logging.info("Added indices...")
+
+    logging.info("Importing concept -> concept scheme relationships...")
+
+    # Load concept -> concept scheme relationships
+    for concept in concepts:
+        edge_type = "IsFrom"
+        edge =  ("Concept", concept["iri"], "ConceptScheme", main_concept_scheme["iri"])
+        query, args = build_edges_import_query_and_args([edge_type], [edge])
+        driver.execute_query(query, args)
+
+    logging.info("Imported concept -> concept scheme relationships")
+
+    logging.info("Importing concept -> concept 'broader' relationships...")
+
+    # Load concept "broader" their relationships
+    for semantic_relation in semantic_relations:
+        edge_type = semantic_relations[0]["type"]
+
+        edge =  ("Concept", semantic_relation["source_concept_iri"], "Concept", semantic_relation["target_concept_iri"])
+
+        query, args = build_edges_import_query_and_args([edge_type], [edge])
+        driver.execute_query(query, args)
+
+    logging.info("Imported concept -> concept 'broader' relationships")
+
+
+def build_nodes_import_query_and_args(labels: List[str], nodes: List[Dict[str, Any]]):
+    """
+    Bulk import nodes into Neo4J
+
+    ## Example
+
+    > build_nodes_import_query_and_args(["Hello", "World"], [{"a": 1, "b": 2}, {"a": 1, "c": 10}])
+    (
+        "MERGE (e_0:Hello:World) {a: $a_0, b: $b_0}\nMERGE (e_1:Hello:World) {a: $a_1, c: $c_1}",
+        {'a_0': 1, 'b_0': 2, 'a_1': 1, 'c_1': 10}
+    )
+    """
+    query_args = {}
+    for idx, node in enumerate(nodes):
+        for k, v in node.items():
+            query_args[f"{k}_{idx}"] = v
+
+    schema_keys = set()
+    for node in nodes:
+        for k in node.keys():
+            schema_keys.add(k)
+
+    node_labels_str = ':'.join(labels)
+
+    query_rows = []
+    for idx, node in enumerate(nodes):
+        schema_kv = [f"{k}: ${k}_{idx}" for k in node.keys()]
+        query_row = f"MERGE (e_{idx}:{node_labels_str} {{{', '.join(schema_kv)}}})"
+        query_rows.append(query_row)
+
+    query = "\n".join(query_rows)
+    return query, query_args
+
+
+def build_edges_import_query_and_args(labels: List[str], edges: List[Tuple[str, str, str, str]]):
+    """
+    Bulk import nodes into Neo4J
+
+    ## Example
+
+    > build_edges_import_query_and_args(["IsFrom"], [("Concept", "def", "ConceptScheme", "abcd")])
+    (
+        "MATCH (src_0:Concept {iri: $iri_src_0}), (tgt_0: ConceptScheme {iri: $iri_tgt_0})\nWITH src_0, tgt_0\nMERGE (src_0)-[r_0:IsFrom]->(tgt_0)",
+        {'iri_src_0': 'def', 'iri_tgt_0': 'abcd'}
+    )
+    """
+    query_args = {}
+    for idx, edge in enumerate(edges):
+        _, source_iri, _, target_iri = edge 
+        query_args[f"iri_src_{idx}"] = source_iri
+        query_args[f"iri_tgt_{idx}"] = target_iri
+
+    edge_labels_str = ":".join(labels)
+
+    matches = []
+    withs = []
+    merges = []
+    for idx, edge in enumerate(edges):
+        source_label, source_iri, target_label, target_iri = edge 
+        matches.extend([
+            f"(src_{idx}:{source_label} {{iri: $iri_src_{idx}}})",
+            f"(tgt_{idx}:{target_label} {{iri: $iri_tgt_{idx}}})"
+        ])
+        withs.extend([f"src_{idx}", f"tgt_{idx}"])
+        merges.extend([f"(src_{idx})-[r_{idx}:{edge_labels_str}]->(tgt_{idx})"])
+
+    query = f"""
+    MATCH {', '.join(matches)}
+    """
+    for merge in merges:
+        query += f"MERGE {merge}"
+
+    return query, query_args
+
+
+def build_index_query(label: str, key: str):
+    """
+    Build indices for a list of keys on labels
+    """
+    return f"CREATE INDEX {label}_{key}_index IF NOT EXISTS FOR (c:{label}) ON (c.{key})"
+
+
+def chunk_list(lst: list, n: int):
+    """
+    Yield successive n-sized chunks from list `lst`.
+    """
+    for i in range(0, len(lst), n):
+        yield lst[i:i + n]

dds_glossary/services.py

@@ -118,8 +118,8 @@ def get_scheme_members(
         """
         return [member for member in members if member.member_type == member_type]
 
+    @staticmethod
     def parse_dataset(
-        self,
         dataset_path: Path,
     ) -> tuple[
         list[ConceptScheme],

docker-compose.neo4j.yml

@@ -0,0 +1,14 @@
+services:
+  neo4j:
+    image: "neo4j:5.20.0-community-bullseye"
+    volumes:
+      - neo4j_data:/data
+      - neo4j_logs:/logs
+    ports:
+      - 7474:7474
+      - 7687:7687
+    environment:
+      - NEO4J_AUTH=${NEO4J_AUTH}
+volumes:
+  neo4j_data:
+  neo4j_logs:

pyproject.toml

@@ -39,6 +39,7 @@ dependencies = [
     "SQLAlchemy>=2.0.0",
     "sqlalchemy-utils",
     "uvicorn[standard]",
+    "neo4j"
 ]
 
 [project.urls]