Cleaned up mdsal-common-api and mdsal-dom-spi
[mdsal.git] / common / mdsal-common-api / src / main / java / org / opendaylight / mdsal / common / api / AsyncWriteTransaction.java
diff --git a/common/mdsal-common-api/src/main/java/org/opendaylight/mdsal/common/api/AsyncWriteTransaction.java b/common/mdsal-common-api/src/main/java/org/opendaylight/mdsal/common/api/AsyncWriteTransaction.java
new file mode 100644 (file)
index 0000000..21ad979
--- /dev/null
@@ -0,0 +1,476 @@
+/*
+ * Copyright (c) 2014 Cisco Systems, Inc. and others.  All rights reserved.
+ *
+ * This program and the accompanying materials are made available under the
+ * terms of the Eclipse Public License v1.0 which accompanies this distribution,
+ * and is available at http://www.eclipse.org/legal/epl-v10.html
+ */
+package org.opendaylight.mdsal.common.api;
+
+import com.google.common.util.concurrent.CheckedFuture;
+import com.google.common.util.concurrent.ListenableFuture;
+import org.opendaylight.yangtools.concepts.Path;
+
+/**
+ * Write transaction provides mutation capabilities for a data tree.
+ *
+ * <p>
+ * Initial state of write transaction is a stable snapshot of the current data tree.
+ * The state is captured when the transaction is created and its state and underlying
+ * data tree are not affected by other concurrently running transactions.
+ * <p>
+ * Write transactions are isolated from other concurrent write transactions. All
+ * writes are local to the transaction and represent only a proposal of state
+ * change for the data tree and it is not visible to any other concurrently running
+ * transaction.
+ * <p>
+ * Applications make changes to the local data tree in the transaction by via the
+ * <b>put</b>, <b>merge</b>, and <b>delete</b> operations.
+ *
+ * <h2>Put operation</h2>
+ * Stores a piece of data at a specified path. This acts as an add / replace
+ * operation, which is to say that whole subtree will be replaced by the
+ * specified data.
+ * <p>
+ * Performing the following put operations:
+ *
+ * <pre>
+ * 1) container { list [ a ] }
+ * 2) container { list [ b ] }
+ * </pre>
+ *
+ * will result in the following data being present:
+ *
+ * <pre>
+ * container { list [ b ] }
+ * </pre>
+ * <h2>Merge operation</h2>
+ * Merges a piece of data with the existing data at a specified path. Any pre-existing data
+ * which is not explicitly overwritten will be preserved. This means that if you store a container,
+ * its child lists will be merged.
+ * <p>
+ * Performing the following merge operations:
+ *
+ * <pre>
+ * 1) container { list [ a ] }
+ * 2) container { list [ b ] }
+ * </pre>
+ *
+ * will result in the following data being present:
+ *
+ * <pre>
+ * container { list [ a, b ] }
+ * </pre>
+ *
+ * This also means that storing the container will preserve any
+ * augmentations which have been attached to it.
+ *
+ * <h2>Delete operation</h2>
+ * Removes a piece of data from a specified path.
+ * <p>
+ * After applying changes to the local data tree, applications publish the changes proposed in the
+ * transaction by calling {@link #submit} on the transaction. This seals the transaction
+ * (preventing any further writes using this transaction) and submits it to be
+ * processed and applied to global conceptual data tree.
+ * <p>
+ * The transaction commit may fail due to a concurrent transaction modifying and committing data in
+ * an incompatible way. See {@link #submit} for more concrete commit failure examples.
+ * <p>
+ * <b>Implementation Note:</b> This interface is not intended to be implemented
+ * by users of MD-SAL, but only to be consumed by them.
+ *
+ * @param <P>
+ *            Type of path (subtree identifier), which represents location in
+ *            tree
+ * @param <D>
+ *            Type of data (payload), which represents data payload
+ */
+public interface AsyncWriteTransaction<P extends Path<P>, D> extends AsyncTransaction<P, D> {
+    /**
+     * Cancels the transaction.
+     *
+     * Transactions can only be cancelled if it was not yet submited.
+     *
+     * Invoking cancel() on failed or already canceled will have no effect, and transaction is
+     * considered cancelled.
+     *
+     * Invoking cancel() on finished transaction (future returned by {@link #submit()} already
+     * successfully completed) will always fail (return false).
+     *
+     * @return <tt>false</tt> if the task could not be cancelled, typically because it has already
+     *         completed normally; <tt>true</tt> otherwise
+     *
+     */
+    boolean cancel();
+
+    /**
+     * Removes a piece of data from specified path. This operation does not fail if the specified
+     * path does not exist.
+     *
+     * @param store Logical data store which should be modified
+     * @param path Data object path
+     * @throws IllegalStateException if the transaction was submitted or canceled.
+     */
+    void delete(LogicalDatastoreType store, P path);
+
+    /**
+     * Submits this transaction to be asynchronously applied to update the logical data tree. The
+     * returned CheckedFuture conveys the result of applying the data changes.
+     * <p>
+     * <b>Note:</b> It is strongly recommended to process the CheckedFuture result in an
+     * asynchronous manner rather than using the blocking get() method. See example usage below.
+     * <p>
+     * This call logically seals the transaction, which prevents the client from further changing
+     * data tree using this transaction. Any subsequent calls to
+     * <code>put(LogicalDatastoreType, Path, Object)</code>,
+     * <code>merge(LogicalDatastoreType, Path, Object)</code>,
+     * <code>delete(LogicalDatastoreType, Path)</code> will fail with {@link IllegalStateException}.
+     *
+     * The transaction is marked as submitted and enqueued into the data store back-end for
+     * processing.
+     *
+     * <p>
+     * Whether or not the commit is successful is determined by versioning of the data tree and
+     * validation of registered commit participants if the transaction changes the data tree.
+     * <p>
+     * The effects of a successful commit of data depends on data change listeners (
+     * {@link AsyncDataChangeListener}) and commit participants that are registered with the data
+     * broker.
+     * <p>
+     * <h3>Example usage:</h3>
+     *
+     * <pre>
+     *  private void doWrite( final int tries ) {
+     *      WriteTransaction writeTx = dataBroker.newWriteOnlyTransaction();
+     * 
+     *      MyDataObject data = ...;
+     *      InstanceIdentifier&lt;MyDataObject&gt; path = ...;
+     *      writeTx.put( LogicalDatastoreType.OPERATIONAL, path, data );
+     * 
+     *      Futures.addCallback( writeTx.submit(), new FutureCallback&lt;Void&gt;() {
+     *          public void onSuccess( Void result ) {
+     *              // succeeded
+     *          }
+     * 
+     *          public void onFailure( Throwable t ) {
+     *              if( t instanceof OptimisticLockFailedException ) {
+     *                  if( ( tries - 1 ) &gt; 0 ) {
+     *                      // do retry
+     *                      doWrite( tries - 1 );
+     *                  } else {
+     *                      // out of retries
+     *                  }
+     *              } else {
+     *                  // failed due to another type of TransactionCommitFailedException.
+     *              }
+     *          } );
+     * }
+     * ...
+     * doWrite( 2 );
+     * </pre>
+     *
+     * <h2>Failure scenarios</h2>
+     * <p>
+     * Transaction may fail because of multiple reasons, such as
+     * <ul>
+     * <li>Another transaction finished earlier and modified the same node in a non-compatible way
+     * (see below). In this case the returned future will fail with an
+     * {@link OptimisticLockFailedException}. It is the responsibility of the caller to create a new
+     * transaction and submit the same modification again in order to update data tree.
+     * <i><b>Warning</b>: In most cases, retrying after an OptimisticLockFailedException will result
+     * in a high probability of success. However, there are scenarios, albeit unusual, where any
+     * number of retries will not succeed. Therefore it is strongly recommended to limit the number
+     * of retries (2 or 3) to avoid an endless loop.</i></li>
+     * <li>Data change introduced by this transaction did not pass validation by commit handlers or
+     * data was incorrectly structured. Returned future will fail with a
+     * {@link DataValidationFailedException}. User should not retry to create new transaction with
+     * same data, since it probably will fail again.</li>
+     * </ul>
+     *
+     * <h3>Change compatibility</h3>
+     *
+     * There are several sets of changes which could be considered incompatible between two
+     * transactions which are derived from same initial state. Rules for conflict detection applies
+     * recursively for each subtree level.
+     *
+     * <h4>Change compatibility of leafs, leaf-list items</h4>
+     *
+     * Following table shows state changes and failures between two concurrent transactions, which
+     * are based on same initial state, Tx 1 completes successfully before Tx 2 is submitted.
+     *
+     * <table summary="Change compatibility of leaf values">
+     * <tr>
+     * <th>Initial state</th>
+     * <th>Tx 1</th>
+     * <th>Tx 2</th>
+     * <th>Result</th>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(A,1)</td>
+     * <td>put(A,2)</td>
+     * <td>Tx 2 will fail, state is A=1</td>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(A,1)</td>
+     * <td>merge(A,2)</td>
+     * <td>A=2</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>Empty</td>
+     * <td>merge(A,1)</td>
+     * <td>put(A,2)</td>
+     * <td>Tx 2 will fail, state is A=1</td>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>merge(A,1)</td>
+     * <td>merge(A,2)</td>
+     * <td>A=2</td>
+     * </tr>
+     *
+     *
+     * <tr>
+     * <td>A=0</td>
+     * <td>put(A,1)</td>
+     * <td>put(A,2)</td>
+     * <td>Tx 2 will fail, A=1</td>
+     * </tr>
+     * <tr>
+     * <td>A=0</td>
+     * <td>put(A,1)</td>
+     * <td>merge(A,2)</td>
+     * <td>A=2</td>
+     * </tr>
+     * <tr>
+     * <td>A=0</td>
+     * <td>merge(A,1)</td>
+     * <td>put(A,2)</td>
+     * <td>Tx 2 will fail, A=1</td>
+     * </tr>
+     * <tr>
+     * <td>A=0</td>
+     * <td>merge(A,1)</td>
+     * <td>merge(A,2)</td>
+     * <td>A=2</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>A=0</td>
+     * <td>delete(A)</td>
+     * <td>put(A,2)</td>
+     * <td>Tx 2 will fail, A does not exists</td>
+     * </tr>
+     * <tr>
+     * <td>A=0</td>
+     * <td>delete(A)</td>
+     * <td>merge(A,2)</td>
+     * <td>A=2</td>
+     * </tr>
+     * </table>
+     *
+     * <h4>Change compatibility of subtrees</h4>
+     *
+     * Following table shows state changes and failures between two concurrent transactions, which
+     * are based on same initial state, Tx 1 completes successfully before Tx 2 is submitted.
+     *
+     * <table summary="Change compatibility of containers">
+     * <tr>
+     * <th>Initial state</th>
+     * <th>Tx 1</th>
+     * <th>Tx 2</th>
+     * <th>Result</th>
+     * </tr>
+     *
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(TOP,[])</td>
+     * <td>put(TOP,[])</td>
+     * <td>Tx 2 will fail, state is TOP=[]</td>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(TOP,[])</td>
+     * <td>merge(TOP,[])</td>
+     * <td>TOP=[]</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(TOP,[FOO=1])</td>
+     * <td>put(TOP,[BAR=1])</td>
+     * <td>Tx 2 will fail, state is TOP=[FOO=1]</td>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>put(TOP,[FOO=1])</td>
+     * <td>merge(TOP,[BAR=1])</td>
+     * <td>TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>Empty</td>
+     * <td>merge(TOP,[FOO=1])</td>
+     * <td>put(TOP,[BAR=1])</td>
+     * <td>Tx 2 will fail, state is TOP=[FOO=1]</td>
+     * </tr>
+     * <tr>
+     * <td>Empty</td>
+     * <td>merge(TOP,[FOO=1])</td>
+     * <td>merge(TOP,[BAR=1])</td>
+     * <td>TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>put(TOP,[FOO=1])</td>
+     * <td>put(TOP,[BAR=1])</td>
+     * <td>Tx 2 will fail, state is TOP=[FOO=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>put(TOP,[FOO=1])</td>
+     * <td>merge(TOP,[BAR=1])</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>merge(TOP,[FOO=1])</td>
+     * <td>put(TOP,[BAR=1])</td>
+     * <td>Tx 2 will fail, state is TOP=[FOO=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>merge(TOP,[FOO=1])</td>
+     * <td>merge(TOP,[BAR=1])</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>delete(TOP)</td>
+     * <td>put(TOP,[BAR=1])</td>
+     * <td>Tx 2 will fail, state is empty store</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>delete(TOP)</td>
+     * <td>merge(TOP,[BAR=1])</td>
+     * <td>state is TOP=[BAR=1]</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>put(TOP/FOO,1)</td>
+     * <td>put(TOP/BAR,1])</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>put(TOP/FOO,1)</td>
+     * <td>merge(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>merge(TOP/FOO,1)</td>
+     * <td>put(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>merge(TOP/FOO,1)</td>
+     * <td>merge(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=1,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>delete(TOP)</td>
+     * <td>put(TOP/BAR,1)</td>
+     * <td>Tx 2 will fail, state is empty store</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[]</td>
+     * <td>delete(TOP)</td>
+     * <td>merge(TOP/BAR,1]</td>
+     * <td>Tx 2 will fail, state is empty store</td>
+     * </tr>
+     *
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>put(TOP/FOO,2)</td>
+     * <td>put(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=2,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>put(TOP/FOO,2)</td>
+     * <td>merge(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=2,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>merge(TOP/FOO,2)</td>
+     * <td>put(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=2,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>merge(TOP/FOO,2)</td>
+     * <td>merge(TOP/BAR,1)</td>
+     * <td>state is TOP=[FOO=2,BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>delete(TOP/FOO)</td>
+     * <td>put(TOP/BAR,1)</td>
+     * <td>state is TOP=[BAR=1]</td>
+     * </tr>
+     * <tr>
+     * <td>TOP=[FOO=1]</td>
+     * <td>delete(TOP/FOO)</td>
+     * <td>merge(TOP/BAR,1]</td>
+     * <td>state is TOP=[BAR=1]</td>
+     * </tr>
+     * </table>
+     *
+     *
+     * <h3>Examples of failure scenarios</h3>
+     *
+     * <h4>Conflict of two transactions</h4>
+     *
+     * This example illustrates two concurrent transactions, which derived from same initial state
+     * of data tree and proposes conflicting modifications.
+     *
+     * <pre>
+     * txA = broker.newWriteTransaction(); // allocates new transaction, data tree is empty
+     * txB = broker.newWriteTransaction(); // allocates new transaction, data tree is empty
+     * 
+     * txA.put(CONFIGURATION, PATH, A);    // writes to PATH value A
+     * txB.put(CONFIGURATION, PATH, B)     // writes to PATH value B
+     * 
+     * ListenableFuture futureA = txA.submit(); // transaction A is sealed and submitted
+     * ListenebleFuture futureB = txB.submit(); // transaction B is sealed and submitted
+     * </pre>
+     *
+     * Commit of transaction A will be processed asynchronously and data tree will be updated to
+     * contain value <code>A</code> for <code>PATH</code>. Returned {@link ListenableFuture} will
+     * successfully complete once state is applied to data tree.
+     *
+     * Commit of Transaction B will fail, because previous transaction also modified path in a
+     * concurrent way. The state introduced by transaction B will not be applied. Returned
+     * {@link ListenableFuture} object will fail with {@link OptimisticLockFailedException}
+     * exception, which indicates to client that concurrent transaction prevented the submitted
+     * transaction from being applied. <br>
+     *
+     * @return a CheckFuture containing the result of the commit. The Future blocks until the commit
+     *         operation is complete. A successful commit returns nothing. On failure, the Future
+     *         will fail with a {@link TransactionCommitFailedException} or an exception derived
+     *         from TransactionCommitFailedException.
+     *
+     * @throws IllegalStateException if the transaction is already submitted or was canceled.
+     */
+    CheckedFuture<Void,TransactionCommitFailedException> submit();
+
+}