Set rootNodeCount after all nodes are loaded

Before we set the rootNodeCount as soon as the first node was added to the LazyIdMapBuilder. However, lead to a bug where the number of pages did not match the node count.

Co-authored-by: Paul Horn <paul.horn@neotechnology.com>

Author: FlorentinD

core/src/main/java/org/neo4j/gds/core/loading/AdjacencyBuffer.java

@@ -136,7 +136,8 @@ public static AdjacencyBuffer of(
 
         return new AdjacencyBuffer(
             importMetaData,
-            adjacencyCompressorFactory, localBuilders,
+            adjacencyCompressorFactory,
+            localBuilders,
             compressedAdjacencyLists,
             paging,
             atLeastOnePropertyToLoad

core/src/main/java/org/neo4j/gds/core/loading/LazyIdMapBuilder.java

@@ -51,6 +51,10 @@ public LazyIdMapBuilder(int concurrency, boolean hasLabelInformation, boolean ha
             .build();
     }
 
+    public void prepareForFlush() {
+        isEmpty.set(false);
+    }
+
     public long addNode(long nodeId, NodeLabelToken nodeLabels) {
         var intermediateId = this.intermediateIdMapBuilder.toMappedNodeId(nodeId);
 
@@ -61,7 +65,6 @@ public long addNode(long nodeId, NodeLabelToken nodeLabels) {
 
         intermediateId = this.intermediateIdMapBuilder.addNode(nodeId);
 
-        isEmpty.lazySet(false);
         this.nodesBuilder.addNode(intermediateId, nodeLabels);
 
         return intermediateId;
@@ -80,7 +83,6 @@ public long addNodeWithProperties(
         }
 
         intermediateId = this.intermediateIdMapBuilder.addNode(nodeId);
-        isEmpty.lazySet(false);
         if (properties.isEmpty()) {
             this.nodesBuilder.addNode(intermediateId, nodeLabels);
         } else {

cypher-aggregation/src/main/java/org/neo4j/gds/projection/CypherAggregation.java

@@ -567,6 +567,8 @@ private static double[] loadMultipleRelationshipProperties(
             // validate again before doing the heavier graph building
             validateGraphName(graphName);
 
+            Objects.requireNonNull(idMapBuilder).prepareForFlush();
+
             var graphStoreBuilder = new GraphStoreBuilder()
                 .concurrency(config.readConcurrency())
                 .capabilities(ImmutableStaticCapabilities.of(true))

cypher-aggregation/src/test/java/org/neo4j/gds/projection/CypherAggregationTest.java

@@ -31,6 +31,7 @@
 import org.neo4j.gds.BaseProcTest;
 import org.neo4j.gds.BaseTest;
 import org.neo4j.gds.NodeLabel;
+import org.neo4j.gds.ResourceUtil;
 import org.neo4j.gds.api.DefaultValue;
 import org.neo4j.gds.api.Graph;
 import org.neo4j.gds.api.GraphStore;
@@ -208,6 +209,28 @@ void testLargerGraphSize() {
         }
     }
 
+    @Nested
+    class CoraGraphTest extends BaseTest {
+        @Test
+        void load_cora_rels() {
+            List<String> lines = ResourceUtil.lines("cora_rels.csv");
+            List<long[]> rows = lines.stream().map(i -> {
+                var values = i.split(",");
+                return new long[]{Long.parseLong(values[0]), Long.parseLong(values[1])};
+            }).collect(Collectors.toList());
+
+            runQuery(
+                "UNWIND $data AS data WITH data RETURN gds.alpha.graph.project('g', data[0], data[1], {}, {}, {readConcurrency: 1})",
+                Map.of("data", rows)
+            );
+
+            assertThat(GraphStoreCatalog.exists("", db.databaseName(), "g")).isTrue();
+            var graph = GraphStoreCatalog.get("", db.databaseName(), "g").graphStore().getUnion();
+            assertThat(graph.nodeCount()).isEqualTo(2708);
+            assertThat(graph.relationshipCount()).isEqualTo(rows.size());
+        }
+    }
+
     @Nested
     class InverseGraphTest extends BaseTest {