2 * Copyright (c) 2014 Cisco Systems, Inc. and others. All rights reserved.
4 * This program and the accompanying materials are made available under the
5 * terms of the Eclipse Public License v1.0 which accompanies this distribution,
6 * and is available at http://www.eclipse.org/legal/epl-v10.html
8 package org.opendaylight.mdsal.common.api;
10 import org.opendaylight.yangtools.concepts.Path;
13 * A chain of transactions. Transactions in a chain need to be committed in
14 * sequence and each transaction should see the effects of previous committed transactions
15 * as they occurred. A chain makes no guarantees of atomicity across the chained transactions -
16 * the transactions are committed as soon as possible in the order that they were submitted.
18 * This behaviour is different from the default AsyncDataBroker, where a
19 * transaction is always created from the current global state, not taking into
20 * account any transactions previously committed by the calling thread. Due to
21 * the asynchronous nature of transaction submission this can lead to surprising
22 * results. If a thread executes the following sequence sufficiently quickly:
25 * AsyncWriteTransaction t1 = broker.newWriteOnlyTransaction();
29 * AsyncReadTransaction t2 = broker.newReadOnlyTransaction();
30 * Optional<?> maybeData = t2.read(id).get();
33 * it may happen, that it sees maybeData.isPresent() == false, simply because
34 * t1 has not completed the processes of being applied and t2 is actually
35 * allocated from the previous state. This is obviously bad for users who create
36 * incremental state in the datastore and actually read what they write in
37 * subsequent transactions.
39 * Using a TransactionChain instead of a broker solves this particular problem,
40 * and leads to expected behavior: t2 will always see the data written in t1
43 public interface TransactionChain<P extends Path<P>, D> extends AutoCloseable,
44 AsyncDataTransactionFactory<P, D> {
47 * Create a new read only transaction which will continue the chain.
50 * The previous write transaction has to be either SUBMITTED
51 * ({@link AsyncWriteTransaction#submit submit} was invoked) or CANCELLED
52 * ({@link #close close} was invoked).
54 * The returned read-only transaction presents an isolated view of the data if the previous
55 * write transaction was successful - in other words, this read-only transaction will see the
56 * state changes made by the previous write transaction in the chain. However, state which
57 * was introduced by other transactions outside this transaction chain after creation of
58 * the previous transaction is not visible.
60 * @return New transaction in the chain.
61 * @throws IllegalStateException
62 * if the previous transaction was not SUBMITTED or CANCELLED.
63 * @throws TransactionChainClosedException
64 * if the chain has been closed.
67 public AsyncReadOnlyTransaction<P, D> newReadOnlyTransaction();
70 * Create a new write-only transaction which will continue the chain.
73 * The previous write transaction has to be either SUBMITTED
74 * ({@link AsyncWriteTransaction#submit submit} was invoked) or CANCELLED
75 * ({@link #close close} was invoked).
77 * The returned write-only transaction presents an isolated view of the data if the previous
78 * write transaction was successful - in other words, this write-only transaction will see the
79 * state changes made by the previous write transaction in the chain. However, state which
80 * was introduced by other transactions outside this transaction chain after creation of
81 * the previous transaction is not visible.
83 * Committing this write-only transaction using {@link AsyncWriteTransaction#submit submit}
84 * will submit the state changes in this transaction to be visible to any subsequent
85 * transaction in this chain and also to any transaction outside this chain.
87 * @return New transaction in the chain.
88 * @throws IllegalStateException
89 * if the previous transaction was not SUBMITTED or CANCELLED.
90 * @throws TransactionChainClosedException
91 * if the chain has been closed.
94 public AsyncWriteTransaction<P, D> newWriteOnlyTransaction();