GH-5553 Lazily compute LMDB index names on demand

Author: hmottestad

core/queryalgebra/evaluation/src/main/java/org/eclipse/rdf4j/query/algebra/evaluation/impl/evaluationsteps/StatementPatternQueryEvaluationStep.java

@@ -276,8 +276,8 @@ private JoinStatementWithBindingSetIterator getIteration(BindingSet bindings) {
 			}
 
 			if (iteration instanceof IndexReportingIterator) {
-				String indexName = ((IndexReportingIterator) iteration).getIndexName();
-				statementPattern.setIndexName(indexName);
+				statementPattern.setIndexNameSupplier(
+						((IndexReportingIterator) iteration)::getIndexName);
 			} else {
 				statementPattern.setIndexName(null);
 			}
@@ -332,8 +332,8 @@ private ConvertStatementToBindingSetIterator getIteration() {
 				iteration = tripleSource.getStatements((Resource) subject, (IRI) predicate, object, contexts);
 			}
 			if (iteration instanceof IndexReportingIterator) {
-				String indexName = ((IndexReportingIterator) iteration).getIndexName();
-				statementPattern.setIndexName(indexName);
+				statementPattern.setIndexNameSupplier(
+						((IndexReportingIterator) iteration)::getIndexName);
 			} else {
 				statementPattern.setIndexName(null);
 			}

core/queryalgebra/model/src/main/java/org/eclipse/rdf4j/query/algebra/StatementPattern.java

@@ -18,6 +18,7 @@
 import java.util.List;
 import java.util.Objects;
 import java.util.Set;
+import java.util.function.Supplier;
 import java.util.stream.Collectors;
 
 import org.eclipse.rdf4j.common.annotation.Experimental;
@@ -65,6 +66,7 @@ public enum Scope {
 	private StatementOrder statementOrder;
 
 	private String indexName;
+	private transient Supplier<String> indexNameSupplier;
 
 	private Set<String> assuredBindingNames;
 	private List<Var> varList;
@@ -537,11 +539,22 @@ public Var getOrder() {
 
 	@Experimental
 	public String getIndexName() {
+		if (indexNameSupplier != null) {
+			indexName = indexNameSupplier.get();
+			indexNameSupplier = null;
+		}
 		return indexName;
 	}
 
 	@Experimental
 	public void setIndexName(String indexName) {
 		this.indexName = indexName;
+		this.indexNameSupplier = null;
+	}
+
+	@Experimental
+	public void setIndexNameSupplier(Supplier<String> indexNameSupplier) {
+		this.indexNameSupplier = indexNameSupplier;
+		this.indexName = null;
 	}
 }

core/queryalgebra/model/src/main/java/org/eclipse/rdf4j/query/algebra/helpers/QueryModelTreeToGenericPlanNode.java

@@ -17,6 +17,7 @@
 import org.eclipse.rdf4j.query.algebra.BinaryTupleOperator;
 import org.eclipse.rdf4j.query.algebra.QueryModelNode;
 import org.eclipse.rdf4j.query.algebra.QueryRoot;
+import org.eclipse.rdf4j.query.algebra.StatementPattern;
 import org.eclipse.rdf4j.query.algebra.VariableScopeChange;
 import org.eclipse.rdf4j.query.explanation.GenericPlanNode;
 
@@ -44,6 +45,9 @@ public GenericPlanNode getGenericPlanNode() {
 
 	@Override
 	protected void meetNode(QueryModelNode node) {
+		if (node instanceof StatementPattern) {
+			((StatementPattern) node).getIndexName();
+		}
 		GenericPlanNode genericPlanNode = new GenericPlanNode(node.getSignature());
 		genericPlanNode.setCostEstimate(node.getCostEstimate());
 		genericPlanNode.setResultSizeEstimate(node.getResultSizeEstimate());

core/sail/lmdb/src/main/java/org/eclipse/rdf4j/sail/lmdb/LmdbRecordIterator.java

@@ -54,7 +54,7 @@ class LmdbRecordIterator implements RecordIterator {
 
 	private final TripleIndex index;
 
-	private final String indexName;
+	private volatile String indexName;
 
 	private final long subj;
 	private final long pred;
@@ -109,7 +109,6 @@ class LmdbRecordIterator implements RecordIterator {
 		this.keyData = pool.getVal();
 		this.valueData = pool.getVal();
 		this.index = index;
-		this.indexName = computeIndexName(index, subj, pred, obj, context);
 		if (rangeSearch) {
 			minKeyBuf = pool.getKeyBuffer();
 			index.getMinKey(minKeyBuf, subj, pred, obj, context);
@@ -537,7 +536,17 @@ public long[] next() {
 
 	@Override
 	public String getIndexName() {
-		return indexName;
+		String current = indexName;
+		if (current == null) {
+			synchronized (this) {
+				current = indexName;
+				if (current == null) {
+					current = computeIndexName(index, subj, pred, obj, context);
+					indexName = current;
+				}
+			}
+		}
+		return current;
 	}
 
 	private boolean matches() {

core/sail/lmdb/src/main/java/org/eclipse/rdf4j/sail/lmdb/LmdbStatementIterator.java

@@ -36,8 +36,6 @@ class LmdbStatementIterator extends AbstractCloseableIteration<Statement> implem
 	private final ValueStore valueStore;
 	private Statement nextElement;
 
-	private final String indexName;
-
 	/*--------------*
 	 * Constructors *
 	 *--------------*/
@@ -48,7 +46,6 @@ class LmdbStatementIterator extends AbstractCloseableIteration<Statement> implem
 	public LmdbStatementIterator(RecordIterator recordIt, ValueStore valueStore) {
 		this.recordIt = recordIt;
 		this.valueStore = valueStore;
-		this.indexName = recordIt.getIndexName();
 	}
 
 	/*---------*
@@ -142,6 +139,6 @@ public void remove() {
 
 	@Override
 	public String getIndexName() {
-		return indexName;
+		return recordIt.getIndexName();
 	}
 }

core/sail/lmdb/src/test/java/org/eclipse/rdf4j/sail/lmdb/LmdbExplainIndexRecommendationTest.java

@@ -99,6 +99,22 @@ void recommendationPrefersIndexesWithHigherDemand() {
 		assertThat(plan).contains("(scan; consider opcs)");
 	}
 
+	@Test
+	void trackerRemainsEmptyUntilIndexNameRequested() {
+		runSelectQuery("psoc", PERSON_QUERY, connection -> connection.add(
+				connection.getValueFactory().createIRI("http://example.com/alice"), RDF.TYPE, FOAF.PERSON));
+
+		ConcurrentMap<String, LongAdder> tracked = LmdbRecordIterator.getRecommendedIndexTracker();
+		assertThat(tracked).isEmpty();
+
+		String plan = runExplainQuery("psoc", PERSON_QUERY, connection -> {
+			connection.add(connection.getValueFactory().createIRI("http://example.com/bob"), RDF.TYPE, FOAF.PERSON);
+		});
+
+		assertThat(plan).contains("(scan; consider posc)");
+		assertThat(tracked).isNotEmpty();
+	}
+
 	private static Stream<Arguments> allTripleIndexPermutations() {
 		List<String> permutations = new ArrayList<>();
 		permute("spoc".toCharArray(), new boolean[4], new StringBuilder(), permutations);
@@ -141,6 +157,23 @@ private String runExplainQuery(String index, String query, RepositoryConnectionC
 		}
 	}
 
+	private void runSelectQuery(String index, String query, RepositoryConnectionConsumer consumer) {
+		File storeDir = new File(dataDir, index + "-select-" + query.hashCode() + "-" + System.nanoTime());
+		storeDir.mkdirs();
+
+		SailRepository repository = new SailRepository(new LmdbStore(storeDir, new LmdbStoreConfig(index)));
+		repository.init();
+
+		try (SailRepositoryConnection connection = repository.getConnection()) {
+			consumer.accept(connection);
+			connection.prepareTupleQuery(query).evaluate().close();
+		} catch (RepositoryException e) {
+			throw new RuntimeException(e);
+		} finally {
+			repository.shutDown();
+		}
+	}
+
 	@FunctionalInterface
 	private interface RepositoryConnectionConsumer {
 		void accept(SailRepositoryConnection connection) throws RepositoryException;