Menu
▾
▴
[Bigdata-commit] SF.net SVN: bigdata:[4013] branches/QUADS_QUERY_BRANCH/bigdata/src
|
From: <tho...@us...> - 2010-12-18 14:31:30
|
Revision: 4013
http://bigdata.svn.sourceforge.net/bigdata/?rev=4013&view=rev
Author: thompsonbry
Date: 2010-12-18 14:31:23 +0000 (Sat, 18 Dec 2010)
Log Message:
-----------
Some work on optional join group support, primarily declaring the new annotations. I need to apply the annotations to the unit tests and modify the QueryEngine logic for setting up the altSink to handle the push/pop(bool:save) as indicated by the annotations.
Modified Paths:
--------------
branches/QUADS_QUERY_BRANCH/bigdata/src/java/com/bigdata/bop/PipelineOp.java
branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestAll.java
branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestQueryEngineOptionalJoins.java
branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/join/TestPipelineJoin.java
Modified: branches/QUADS_QUERY_BRANCH/bigdata/src/java/com/bigdata/bop/PipelineOp.java
===================================================================
--- branches/QUADS_QUERY_BRANCH/bigdata/src/java/com/bigdata/bop/PipelineOp.java 2010-12-17 15:05:06 UTC (rev 4012)
+++ branches/QUADS_QUERY_BRANCH/bigdata/src/java/com/bigdata/bop/PipelineOp.java 2010-12-18 14:31:23 UTC (rev 4013)
@@ -30,12 +30,9 @@
import java.util.Map;
import java.util.concurrent.Callable;
import java.util.concurrent.FutureTask;
-import java.util.concurrent.TimeUnit;
import com.bigdata.bop.engine.BOpStats;
import com.bigdata.bop.engine.QueryEngine;
-import com.bigdata.relation.accesspath.BlockingBuffer;
-import com.bigdata.relation.accesspath.IBlockingBuffer;
/**
* Abstract base class for pipeline operators where the data moving along the
@@ -85,6 +82,48 @@
boolean DEFAULT_SHARED_STATE = false;
+ /**
+ * Annotation used to mark the set of non-optional joins which may be
+ * input to either the static or runtime query optimizer. Joins within a
+ * join graph may be freely reordered by the query optimizer in order to
+ * minimize the amount of work required to compute the solutions.
+ * <p>
+ * Note: Optional joins MAY NOT appear within the a join graph. Optional
+ * joins SHOULD be evaluated as part of the "tail plan" following the
+ * join graph, but before operations such as SORT, DISTINCT, etc.
+ *
+ * @todo We should be able to automatically apply the static or runtime
+ * query optimizers to an operator tree using this annotation to
+ * identify the join graphs.
+ */
+ String JOIN_GRAPH = PipelineOp.class.getName() + ".joinGraph";
+
+ /**
+ * Annotation marks a high level join group, which may include optional
+ * joins. Join groups are marked in order to decide the re-entry point
+ * in the query plan when a join within an optional join group fails.
+ * Also, the top-level join group is not marked -- only nested join
+ * groups are marked. This is used by the decision rule to handle do
+ * {@link IBindingSet#push()} when entering a
+ * <p>
+ * This is different from a {@link #JOIN_GRAPH} primarily in that the
+ * latter may not include optional joins.
+ */
+ String JOIN_GROUP = PipelineOp.class.getName() + ".joinGroup";
+
+ /**
+ * Annotation is used to designate the target when a join within an
+ * optional join group fails. The value of this annotation must be the
+ * {@link #JOIN_GROUP} identifier corresponding to the next join group
+ * in the query plan. The target join group identifier is specified
+ * (rather than the bopId of the target join) since the joins in the
+ * target join group may be reordered by the query optimizer. The entry
+ * point for solutions redirected to the {@link #ALT_SINK_GROUP} is
+ * therefore the first operator in the target {@link #JOIN_GROUP}. This
+ * decouples the routing decisions from the join ordering decisions.
+ */
+ String ALT_SINK_GROUP = PipelineOp.class.getName() + ".altSinkGroup";
+
}
/**
Modified: branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestAll.java
===================================================================
--- branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestAll.java 2010-12-17 15:05:06 UTC (rev 4012)
+++ branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestAll.java 2010-12-18 14:31:23 UTC (rev 4013)
@@ -75,6 +75,9 @@
// stress test for SliceOp.
suite.addTestSuite(TestQueryEngine_Slice.class);
+ // test suite for optional join groups.
+ suite.addTestSuite(TestQueryEngineOptionalJoins.class);
+
// @todo test suite for query evaluation (DISTINCT, ORDER BY, GROUP BY).
// suite.addTestSuite(TestQueryEngine2.class);
Modified: branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestQueryEngineOptionalJoins.java
===================================================================
--- branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestQueryEngineOptionalJoins.java 2010-12-17 15:05:06 UTC (rev 4012)
+++ branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/engine/TestQueryEngineOptionalJoins.java 2010-12-18 14:31:23 UTC (rev 4013)
@@ -28,17 +28,9 @@
package com.bigdata.bop.engine;
import java.util.Iterator;
-import java.util.LinkedList;
-import java.util.List;
import java.util.Map;
import java.util.Properties;
import java.util.UUID;
-import java.util.concurrent.CancellationException;
-import java.util.concurrent.ExecutionException;
-import java.util.concurrent.Executor;
-import java.util.concurrent.FutureTask;
-import java.util.concurrent.TimeUnit;
-import java.util.concurrent.TimeoutException;
import java.util.concurrent.atomic.AtomicInteger;
import junit.framework.TestCase2;
@@ -62,14 +54,10 @@
import com.bigdata.bop.bindingSet.HashBindingSet;
import com.bigdata.bop.bset.ConditionalRoutingOp;
import com.bigdata.bop.bset.StartOp;
-import com.bigdata.bop.constraint.EQ;
import com.bigdata.bop.constraint.EQConstant;
import com.bigdata.bop.constraint.NEConstant;
-import com.bigdata.bop.fed.TestFederatedQueryEngine;
import com.bigdata.bop.join.PipelineJoin;
import com.bigdata.bop.solutions.SliceOp;
-import com.bigdata.bop.solutions.SliceOp.SliceStats;
-import com.bigdata.io.DirectBufferPoolAllocator.IAllocationContext;
import com.bigdata.journal.BufferMode;
import com.bigdata.journal.ITx;
import com.bigdata.journal.Journal;
@@ -78,16 +66,16 @@
import com.bigdata.striterator.ChunkedArrayIterator;
import com.bigdata.striterator.Dechunkerator;
import com.bigdata.striterator.ICloseableIterator;
-import com.bigdata.util.InnerCause;
-import com.bigdata.util.concurrent.LatchedExecutor;
-import com.ibm.icu.impl.ByteBuffer;
/**
- * Test suite for the {@link QueryEngine} against a local database instance.
- * <p>
- * Note: The {@link BOp}s are unit tested separately. This test suite is focused
- * on interactions when {@link BOp}s are chained together in a query, such as a
- * sequence of pipeline joins, a slice applied to a query, etc.
+ * Test suite for handling of optional join groups during query evaluation
+ * against a local database instance. Optional join groups are handled using
+ * {@link IBindingSet#push()} when entering the join group and
+ * {@link IBindingSet#pop(boolean)} when exiting the join group. If the join
+ * group was successful for a given binding set, then <code>save:=true</code> is
+ * specified for {@link IBindingSet#pop(boolean)} and the applied bindings will
+ * be visible to the downstream consumer. Otherwise the bindings applied during
+ * the join group are simply discarded.
*
* <pre>
* -Dlog4j.configuration=bigdata/src/resources/logging/log4j.properties
@@ -95,10 +83,6 @@
*
* @author <a href="mailto:tho...@us...">Bryan Thompson</a>
* @version $Id: TestQueryEngine.java 3950 2010-11-17 02:14:08Z thompsonbry $
- *
- * @see TestFederatedQueryEngine
- *
- * @todo write a unit and stress tests for deadlines.
*/
public class TestQueryEngineOptionalJoins extends TestCase2 {
@@ -270,14 +254,14 @@
*/
public void test_query_join2_optionals() throws Exception {
- final int startId = 1;
- final int joinId1 = 2;
- final int predId1 = 3;
- final int joinId2 = 4;
- final int predId2 = 5;
- final int joinId3 = 6;
- final int predId3 = 7;
- final int sliceId = 8;
+ final int startId = 1; //
+ final int joinId1 = 2; // : base join group.
+ final int predId1 = 3; // (a b)
+ final int joinId2 = 4; // : joinGroup1
+ final int predId2 = 5; // (b c)
+ final int joinId3 = 6; // : joinGroup1
+ final int predId3 = 7; // (c d)
+ final int sliceId = 8; //
final IVariable<?> a = Var.var("a");
final IVariable<?> b = Var.var("b");
Modified: branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/join/TestPipelineJoin.java
===================================================================
--- branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/join/TestPipelineJoin.java 2010-12-17 15:05:06 UTC (rev 4012)
+++ branches/QUADS_QUERY_BRANCH/bigdata/src/test/com/bigdata/bop/join/TestPipelineJoin.java 2010-12-18 14:31:23 UTC (rev 4013)
@@ -535,16 +535,16 @@
}
- /**
- * Unit tests for optional joins. For an optional join, an alternative sink
- * may be specified in the {@link BOpContext}. When specified, it is used if
- * the join fails (if not specified, the binding sets which do not join are
- * forwarded to the primary sink). Binding sets which join go to the primary
- * sink regardless.
- *
- * @throws ExecutionException
- * @throws InterruptedException
- */
+ /**
+ * Unit tests for optional joins. For an optional join, an alternative sink
+ * may be specified for the join. When specified, it is used if the join
+ * fails (if not specified, the binding sets which do not join are forwarded
+ * to the primary sink). Binding sets which join go to the primary sink
+ * regardless.
+ *
+ * @throws ExecutionException
+ * @throws InterruptedException
+ */
public void test_optionalJoin() throws InterruptedException, ExecutionException {
final Var<?> x = Var.var("x");
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
