Github user afs commented on a diff in the pull request:
https://github.com/apache/jena/pull/369#discussion_r172271582
--- Diff:
jena-arq/src/main/java/org/apache/jena/sparql/core/TxnDataset2Graph.java ---
@@ -0,0 +1,240 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.jena.sparql.core;
+
+import java.util.*;
+import java.util.function.Consumer;
+
+import org.apache.jena.graph.Graph;
+import org.apache.jena.graph.TransactionHandler;
+import org.apache.jena.query.ReadWrite;
+import org.apache.jena.query.TxnType;
+import org.apache.jena.reasoner.InfGraph;
+import org.apache.jena.shared.LockMRSW;
+import org.apache.jena.sparql.JenaTransactionException;
+
+/**
+ * A {@link Transactional} that passes the transaction operations down to
transactions on
+ * independent graphs.
+ * <p>
+ * There are limitations:
+ * <ul>
+ * <li>we can't atomically do all the commits together in the crash
situation.
+ * <li>This {@code Transactional} maintains a MRSW policy because that is
all that is
+ * required of graphs in general.
+ * </ul>
+ * It does cover the important case of one graph ({@link DatasetGraphOne})
where the one
+ * graph is an InfGraph and should work when the graphs in the dataset is
not changing or
+ * when a new memory graph is added mid-transaction.
+ * <p>
+ * This is not "nested transactions" - theer is no overall "commit" or
"abort". If
+ * failure/restart occurs, some graphs may have commited and others not.
It is the best
+ * that can be done given for an arbitrary collection of graphs, backed by
different
+ * storage and having different capabilities.
+ * <p>
+ * Best practice is to change the graph membership outside of any
transaction,
+ * ideally at setup time of the object using this class. (Caution: SPARQL
Update
+ * can create graphs.
+ * @See {@link DatasetGraphMapLink}
+ * @See {@link DatasetGraphOne}
+ */
+public class TxnDataset2Graph extends TransactionalLock {
+ /**
+ * Control whether to pass down transactions from the dataset to the
graph in the
+ * dataset. This should be set to "true"; setting it "false" causes
the onld,
+ * no-transaction passing behaviour.
+ * <p>
+ * This is temporary flag during the transition because the change at
Jena 3.7.0 needs
+ * to be proven in real deployments as well as testing. "false"
restores the Jena
+ * 3.6.0 and before behaviour (transactions not passed down). See
JENA-1492.
+ *
+ * @deprecated This flag will be removed.
+ */
+ @Deprecated
+ public static boolean TXN_DSG_GRAPH = true;
+
+ private Graph primary;
+ // Object key may be a graph or a DSG is the graph is a GraphView.
+ // This avoids starting a tranasction on the same storage unit twice.
+ private Map<Object, TransactionHandler> handlers = new HashMap<>();
+
+ private Object lock = new Object();
+
+ public TxnDataset2Graph(Graph primaryGraph, Graph ... otherGraphs) {
+ super(new LockMRSW());
+ primary = primaryGraph;
+ handlers = buildHandlerSet(primary, Arrays.asList(otherGraphs));
+ }
+
+ private static Map<Object, TransactionHandler> buildHandlerSet(Graph
primary, Collection<Graph> graphs) {
+ Map<Object, TransactionHandler> handlers = new HashMap<>();
+ addHandler(handlers, primary);
+ graphs.forEach(g->addHandler(handlers,g));
+ return handlers;
+ }
+
+ private static void addHandler(Map<Object, TransactionHandler>
handlers, Graph graph) {
+ TransactionHandler th = graph.getTransactionHandler();
+ if ( ! th.transactionsSupported() )
+ return;
+ Object key = calcKey(graph);
+ if ( th.transactionsSupported() )
+ handlers.put(key, th) ;
+ }
+
+ // Determine the key - an object that is the unit of transactions.
+ // For two graphs form the same DatasetGraph, i.e. GraphView, there
should be one transaction.
+ private static Object calcKey(Graph graph) {
+ if ( graph instanceof GraphView )
+ // Use the database as the key so that transactions are
started once-per-storage.
+ // This the case of a graph from some storage being plavced in
a general dataset.
+ return ((GraphView)graph).getDataset();
+ if ( graph instanceof InfGraph )
+ // InfGraph does actual pass done in its TransactionHandler.
+ // This allows the base graph to be included in the dataset as
well as the InfGraph.
+ return calcKey(((InfGraph)graph).getRawGraph());
+
+// if ( graph instanceof GraphWrapper )
+// return calcKey(((GraphWrapper)graph).get());
+// if ( graph instanceof WrappedGraph )
+// return calcKey(((WrappedGraph)graph).getWrapped());
+ return graph;
+ }
+
+ private static void removeHandler(Map<Object, TransactionHandler>
handlers, Graph graph) {
+ Object key = calcKey(graph);
+ handlers.remove(graph);
+ }
+
+ // Attempt to manage the graph transactions during a transaction.
+ // Imperfect for removal, we don't know whether to call commit() or
abort().
+ // Works for adding.
+ // Generally better not to change the graphs during a transaction,
just set them once
+ // on creation.
+
+ public void addGraph(Graph graph) {
+ checkNotReadMode();
+ if ( graph == null )
+ return;
+ if ( ! handlers.containsKey(graph) ) {
+ // Add if new.
+ addHandler(handlers, graph) ;
+ if ( super.isInTransaction() ) {
+ // If we are in a transaction, start the subtransaction.
+ TransactionHandler th = handlers.get(graph);
+ if ( th != null )
+ th.begin();
+ }
+ }
+ }
+
+ public void removeGraph(Graph graph) {
+ checkNotReadMode();
+ if ( graph == null )
+ return;
+ if ( ! super.isInTransaction() ) {
+ // Not in transaction, do now.
+ removeHandler(handlers, graph);
+ return;
+ }
+ // Queue to be removed at the end.
+ Set<Graph> toBeRemoved = removedGraphs.get();
+ if ( toBeRemoved == null ) {
+ // Lazy set of the HashSet.
+ toBeRemoved = new HashSet<>();
+ removedGraphs.set(toBeRemoved);
+ }
+ removedGraphs.get().add(graph);
+ }
+
+ public void setPrimaryGraph(Graph graph) {
+ checkNotReadMode();
+ if ( graph == null )
+ return;
+ removeGraph(graph);
+ addGraph(graph);
+ }
+
+ private void handlers(Consumer<TransactionHandler> action) {
+ synchronized (lock) {
+ handlers.forEach((g,th)->action.accept(th));
+ }
+ }
+
+ private void checkNotReadMode() {
+ if ( !super.isInTransaction() )
+ return;
+ if ( super.isTransactionMode(ReadWrite.READ) )
+ throw new JenaTransactionException("In READ mode in
transaction");
+ }
+
+ private ThreadLocal<Set<Graph>> removedGraphs =
ThreadLocal.withInitial(()->null);
+ private void start() {}
+ private void finish() {
+ if ( ! super.isTransactionMode(ReadWrite.WRITE) )
+ return;
+ // This is called inside the lock of super.
+ Set<Graph> toBeRemoved = removedGraphs.get();
+ removedGraphs.remove();
+ if ( toBeRemoved == null )
+ return ;
+ toBeRemoved.forEach(g->removeHandler(handlers, g));
+ }
+
+ // TransactionalLock.begin(ReadWrite) calls begin(TxnType)
+ @Override
+ public void begin(TxnType type) {
+ super.begin(type);
+ // Whatever the type. Graph Transactions do not allow for
"read-only".
+ start();
+ handlers(h->h.begin());
--- End diff --
I don't think it always does :-) `h.` means the IDE method finding works.
---
