Merge branch 'master' into jarodwen/features/positionvar
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/base/LogicalOperatorDeepCopyVisitor.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/base/LogicalOperatorDeepCopyVisitor.java
index ba940d6..693a19e 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/base/LogicalOperatorDeepCopyVisitor.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/base/LogicalOperatorDeepCopyVisitor.java
@@ -39,6 +39,7 @@
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.EmptyTupleSourceOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExchangeOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExtensionOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExternalDataAccessByRIDOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.GroupByOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.IndexInsertDeleteOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.InnerJoinOperator;
@@ -387,6 +388,13 @@
public ILogicalOperator visitUnnestMapOperator(UnnestMapOperator op, ILogicalOperator arg) {
throw new UnsupportedOperationException();
}
+
+ @Override
+ public ILogicalOperator visitExternalDataAccessByRIDOperator(
+ ExternalDataAccessByRIDOperator op, ILogicalOperator arg)
+ throws AlgebricksException {
+ throw new UnsupportedOperationException();
+ }
@Override
public ILogicalOperator visitUnnestOperator(UnnestOperator op, ILogicalOperator arg) throws AlgebricksException {
@@ -438,4 +446,5 @@
public Map<LogicalVariable, LogicalVariable> getVariableMapping() {
return outVarMapping;
}
+
}
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/CommitRuntime.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/CommitRuntime.java
index a7c2fdb..6dd11bd 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/CommitRuntime.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/CommitRuntime.java
@@ -19,10 +19,11 @@
import edu.uci.ics.asterix.common.api.IAsterixAppRuntimeContext;
import edu.uci.ics.asterix.common.exceptions.ACIDException;
-import edu.uci.ics.asterix.common.transactions.DatasetId;
+import edu.uci.ics.asterix.common.transactions.ILogManager;
import edu.uci.ics.asterix.common.transactions.ITransactionContext;
import edu.uci.ics.asterix.common.transactions.ITransactionManager;
import edu.uci.ics.asterix.common.transactions.JobId;
+import edu.uci.ics.asterix.transaction.management.service.logging.LogRecord;
import edu.uci.ics.hyracks.algebricks.runtime.base.IPushRuntime;
import edu.uci.ics.hyracks.api.comm.IFrameWriter;
import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
@@ -39,11 +40,13 @@
private final IHyracksTaskContext hyracksTaskCtx;
private final ITransactionManager transactionManager;
+ private final ILogManager logMgr;
private final JobId jobId;
- private final DatasetId datasetId;
+ private final int datasetId;
private final int[] primaryKeyFields;
private final boolean isWriteTransaction;
private final long[] longHashes;
+ private final LogRecord logRecord;
private ITransactionContext transactionContext;
private RecordDescriptor inputRecordDesc;
@@ -56,12 +59,14 @@
IAsterixAppRuntimeContext runtimeCtx = (IAsterixAppRuntimeContext) ctx.getJobletContext()
.getApplicationContext().getApplicationObject();
this.transactionManager = runtimeCtx.getTransactionSubsystem().getTransactionManager();
+ this.logMgr = runtimeCtx.getTransactionSubsystem().getLogManager();
this.jobId = jobId;
- this.datasetId = new DatasetId(datasetId);
+ this.datasetId = datasetId;
this.primaryKeyFields = primaryKeyFields;
this.frameTupleReference = new FrameTupleReference();
this.isWriteTransaction = isWriteTransaction;
this.longHashes = new long[2];
+ this.logRecord = new LogRecord();
}
@Override
@@ -82,11 +87,9 @@
for (int t = 0; t < nTuple; t++) {
frameTupleReference.reset(frameTupleAccessor, t);
pkHash = computePrimaryKeyHashValue(frameTupleReference, primaryKeyFields);
- try {
- transactionManager.commitTransaction(transactionContext, datasetId, pkHash);
- } catch (ACIDException e) {
- throw new HyracksDataException(e);
- }
+ logRecord.formEntityCommitLogRecord(transactionContext, datasetId, pkHash, frameTupleReference,
+ primaryKeyFields);
+ logMgr.log(logRecord);
}
}
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/ExternalDataAccessByRIDPOperator.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/ExternalDataAccessByRIDPOperator.java
new file mode 100644
index 0000000..447555e
--- /dev/null
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/algebra/operators/physical/ExternalDataAccessByRIDPOperator.java
@@ -0,0 +1,110 @@
+package edu.uci.ics.asterix.algebra.operators.physical;
+
+import edu.uci.ics.asterix.metadata.declared.AqlDataSource;
+import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
+import edu.uci.ics.asterix.metadata.declared.AqlSourceId;
+import edu.uci.ics.asterix.metadata.entities.Dataset;
+import edu.uci.ics.asterix.metadata.entities.Index;
+import edu.uci.ics.asterix.om.functions.AsterixBuiltinFunctions;
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.algebricks.common.exceptions.AlgebricksException;
+import edu.uci.ics.hyracks.algebricks.common.utils.Pair;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.IHyracksJobBuilder;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.ILogicalExpression;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.ILogicalOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.IOptimizationContext;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.LogicalExpressionTag;
+import edu.uci.ics.hyracks.algebricks.core.algebra.base.PhysicalOperatorTag;
+import edu.uci.ics.hyracks.algebricks.core.algebra.expressions.AbstractFunctionCallExpression;
+import edu.uci.ics.hyracks.algebricks.core.algebra.functions.FunctionIdentifier;
+import edu.uci.ics.hyracks.algebricks.core.algebra.metadata.IDataSourcePropertiesProvider;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AbstractScanOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExternalDataAccessByRIDOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.IOperatorSchema;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.physical.AbstractScanPOperator;
+import edu.uci.ics.hyracks.algebricks.core.jobgen.impl.JobGenContext;
+import edu.uci.ics.hyracks.api.dataflow.IOperatorDescriptor;
+
+public class ExternalDataAccessByRIDPOperator extends AbstractScanPOperator{
+
+ private AqlSourceId datasetId;
+ private Dataset dataset;
+ private ARecordType recordType;
+ private Index secondaryIndex;
+ public ExternalDataAccessByRIDPOperator(AqlSourceId datasetId, Dataset dataset, ARecordType recordType,Index secondaryIndex)
+ {
+ this.datasetId = datasetId;
+ this.dataset = dataset;
+ this.recordType = recordType;
+ this.secondaryIndex = secondaryIndex;
+ }
+
+ public Dataset getDataset() {
+ return dataset;
+ }
+
+ public void setDataset(Dataset dataset) {
+ this.dataset = dataset;
+ }
+
+ public ARecordType getRecordType() {
+ return recordType;
+ }
+
+ public void setRecordType(ARecordType recordType) {
+ this.recordType = recordType;
+ }
+
+ public AqlSourceId getDatasetId() {
+ return datasetId;
+ }
+
+ public void setDatasetId(AqlSourceId datasetId) {
+ this.datasetId = datasetId;
+ }
+
+ @Override
+ public PhysicalOperatorTag getOperatorTag() {
+ return PhysicalOperatorTag.EXTERNAL_ACCESS_BY_RID;
+ }
+
+ @Override
+ public void computeDeliveredProperties(ILogicalOperator op,
+ IOptimizationContext context) throws AlgebricksException {
+ AqlDataSource ds = new AqlDataSource(datasetId, dataset, recordType);
+ IDataSourcePropertiesProvider dspp = ds.getPropertiesProvider();
+ AbstractScanOperator as = (AbstractScanOperator) op;
+ deliveredProperties = dspp.computePropertiesVector(as.getVariables());
+ }
+
+ @Override
+ public void contributeRuntimeOperator(IHyracksJobBuilder builder,
+ JobGenContext context, ILogicalOperator op,
+ IOperatorSchema propagatedSchema, IOperatorSchema[] inputSchemas,
+ IOperatorSchema outerPlanSchema) throws AlgebricksException {
+ ExternalDataAccessByRIDOperator edabro = (ExternalDataAccessByRIDOperator) op;
+ ILogicalExpression expr = edabro.getExpressionRef().getValue();
+ if (expr.getExpressionTag() != LogicalExpressionTag.FUNCTION_CALL) {
+ throw new IllegalStateException();
+ }
+ AbstractFunctionCallExpression funcExpr = (AbstractFunctionCallExpression) expr;
+ FunctionIdentifier funcIdent = funcExpr.getFunctionIdentifier();
+ if (!funcIdent.equals(AsterixBuiltinFunctions.EXTERNAL_ACCESS_BY_RID)) {
+ return;
+ }
+ AqlMetadataProvider metadataProvider = (AqlMetadataProvider) context.getMetadataProvider();
+ Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> externalAccessByRID = metadataProvider.buildExternalDataAccesByRIDRuntime(
+ builder.getJobSpec(), dataset,secondaryIndex);
+ builder.contributeHyracksOperator(edabro, externalAccessByRID.first);
+ builder.contributeAlgebricksPartitionConstraint(externalAccessByRID.first, externalAccessByRID.second);
+ ILogicalOperator srcExchange = edabro.getInputs().get(0).getValue();
+ builder.contributeGraphEdge(srcExchange, 0, edabro, 0);
+ }
+
+ @Override
+ public boolean isMicroOperator() {
+ return false;
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/AccessMethodUtils.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/AccessMethodUtils.java
index ab0fd79..95d29e3 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/AccessMethodUtils.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/AccessMethodUtils.java
@@ -22,8 +22,12 @@
import org.apache.commons.lang3.mutable.Mutable;
import org.apache.commons.lang3.mutable.MutableObject;
+import edu.uci.ics.asterix.algebra.operators.physical.ExternalDataAccessByRIDPOperator;
import edu.uci.ics.asterix.aql.util.FunctionUtils;
+import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.DatasetConfig.IndexType;
+import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
+import edu.uci.ics.asterix.metadata.declared.AqlSourceId;
import edu.uci.ics.asterix.metadata.entities.Dataset;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.utils.DatasetUtils;
@@ -34,6 +38,7 @@
import edu.uci.ics.asterix.om.constants.AsterixConstantValue;
import edu.uci.ics.asterix.om.functions.AsterixBuiltinFunctions;
import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.asterix.om.types.BuiltinType;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.asterix.om.util.NonTaggedFormatUtil;
import edu.uci.ics.hyracks.algebricks.common.exceptions.AlgebricksException;
@@ -52,6 +57,7 @@
import edu.uci.ics.hyracks.algebricks.core.algebra.functions.IFunctionInfo;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator.ExecutionMode;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExternalDataAccessByRIDOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.OrderOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.OrderOperator.IOrder;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.UnnestMapOperator;
@@ -60,290 +66,384 @@
* Static helper functions for rewriting plans using indexes.
*/
public class AccessMethodUtils {
- public static void appendPrimaryIndexTypes(Dataset dataset, IAType itemType, List<Object> target)
- throws IOException {
- ARecordType recordType = (ARecordType) itemType;
- List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
- for (String partitioningKey : partitioningKeys) {
- target.add(recordType.getFieldType(partitioningKey));
- }
- target.add(itemType);
- }
+ public static void appendPrimaryIndexTypes(Dataset dataset, IAType itemType, List<Object> target)
+ throws IOException {
+ ARecordType recordType = (ARecordType) itemType;
+ List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
+ for (String partitioningKey : partitioningKeys) {
+ target.add(recordType.getFieldType(partitioningKey));
+ }
+ target.add(itemType);
+ }
- public static ConstantExpression createStringConstant(String str) {
- return new ConstantExpression(new AsterixConstantValue(new AString(str)));
- }
+ public static void appendExternalRecTypes(Dataset dataset, IAType itemType, List<Object> target){
+ target.add(itemType);
+ }
- public static ConstantExpression createInt32Constant(int i) {
- return new ConstantExpression(new AsterixConstantValue(new AInt32(i)));
- }
+ public static void appendExternalRecPrimaryKey(Dataset dataset, List<Object> target){
+ target.add(BuiltinType.ASTRING);
+ target.add(BuiltinType.AINT64);
+ if(DatasetUtils.getExternalRIDSize(dataset) == 3)
+ {
+ target.add(BuiltinType.AINT32);
+ }
+ }
- public static ConstantExpression createBooleanConstant(boolean b) {
- if (b) {
- return new ConstantExpression(new AsterixConstantValue(ABoolean.TRUE));
- } else {
- return new ConstantExpression(new AsterixConstantValue(ABoolean.FALSE));
- }
- }
+ public static ConstantExpression createStringConstant(String str) {
+ return new ConstantExpression(new AsterixConstantValue(new AString(str)));
+ }
- public static String getStringConstant(Mutable<ILogicalExpression> expr) {
- IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
- return ((AString) obj).getStringValue();
- }
+ public static ConstantExpression createInt32Constant(int i) {
+ return new ConstantExpression(new AsterixConstantValue(new AInt32(i)));
+ }
- public static int getInt32Constant(Mutable<ILogicalExpression> expr) {
- IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
- return ((AInt32) obj).getIntegerValue();
- }
+ public static ConstantExpression createBooleanConstant(boolean b) {
+ if (b) {
+ return new ConstantExpression(new AsterixConstantValue(ABoolean.TRUE));
+ } else {
+ return new ConstantExpression(new AsterixConstantValue(ABoolean.FALSE));
+ }
+ }
- public static boolean getBooleanConstant(Mutable<ILogicalExpression> expr) {
- IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
- return ((ABoolean) obj).getBoolean();
- }
+ public static String getStringConstant(Mutable<ILogicalExpression> expr) {
+ IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
+ return ((AString) obj).getStringValue();
+ }
- public static boolean analyzeFuncExprArgsForOneConstAndVar(AbstractFunctionCallExpression funcExpr,
- AccessMethodAnalysisContext analysisCtx) {
- IAlgebricksConstantValue constFilterVal = null;
- LogicalVariable fieldVar = null;
- ILogicalExpression arg1 = funcExpr.getArguments().get(0).getValue();
- ILogicalExpression arg2 = funcExpr.getArguments().get(1).getValue();
- // One of the args must be a constant, and the other arg must be a variable.
- if (arg1.getExpressionTag() == LogicalExpressionTag.CONSTANT
- && arg2.getExpressionTag() == LogicalExpressionTag.VARIABLE) {
- ConstantExpression constExpr = (ConstantExpression) arg1;
- constFilterVal = constExpr.getValue();
- VariableReferenceExpression varExpr = (VariableReferenceExpression) arg2;
- fieldVar = varExpr.getVariableReference();
- } else if (arg1.getExpressionTag() == LogicalExpressionTag.VARIABLE
- && arg2.getExpressionTag() == LogicalExpressionTag.CONSTANT) {
- ConstantExpression constExpr = (ConstantExpression) arg2;
- constFilterVal = constExpr.getValue();
- VariableReferenceExpression varExpr = (VariableReferenceExpression) arg1;
- fieldVar = varExpr.getVariableReference();
- } else {
- return false;
- }
- analysisCtx.matchedFuncExprs.add(new OptimizableFuncExpr(funcExpr, fieldVar, constFilterVal));
- return true;
- }
+ public static int getInt32Constant(Mutable<ILogicalExpression> expr) {
+ IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
+ return ((AInt32) obj).getIntegerValue();
+ }
- public static boolean analyzeFuncExprArgsForTwoVars(AbstractFunctionCallExpression funcExpr,
- AccessMethodAnalysisContext analysisCtx) {
- LogicalVariable fieldVar1 = null;
- LogicalVariable fieldVar2 = null;
- ILogicalExpression arg1 = funcExpr.getArguments().get(0).getValue();
- ILogicalExpression arg2 = funcExpr.getArguments().get(1).getValue();
- if (arg1.getExpressionTag() == LogicalExpressionTag.VARIABLE
- && arg2.getExpressionTag() == LogicalExpressionTag.VARIABLE) {
- fieldVar1 = ((VariableReferenceExpression) arg1).getVariableReference();
- fieldVar2 = ((VariableReferenceExpression) arg2).getVariableReference();
- } else {
- return false;
- }
- analysisCtx.matchedFuncExprs.add(new OptimizableFuncExpr(funcExpr,
- new LogicalVariable[] { fieldVar1, fieldVar2 }, null));
- return true;
- }
+ public static boolean getBooleanConstant(Mutable<ILogicalExpression> expr) {
+ IAObject obj = ((AsterixConstantValue) ((ConstantExpression) expr.getValue()).getValue()).getObject();
+ return ((ABoolean) obj).getBoolean();
+ }
- public static int getNumSecondaryKeys(Index index, ARecordType recordType) throws AlgebricksException {
- switch (index.getIndexType()) {
- case BTREE:
- case SINGLE_PARTITION_WORD_INVIX:
- case SINGLE_PARTITION_NGRAM_INVIX:
- case LENGTH_PARTITIONED_WORD_INVIX:
- case LENGTH_PARTITIONED_NGRAM_INVIX: {
- return index.getKeyFieldNames().size();
- }
- case RTREE: {
- Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(index.getKeyFieldNames().get(0),
- recordType);
- IAType keyType = keyPairType.first;
- int numDimensions = NonTaggedFormatUtil.getNumDimensions(keyType.getTypeTag());
- return numDimensions * 2;
- }
- default: {
- throw new AlgebricksException("Unknown index kind: " + index.getIndexType());
- }
- }
- }
+ public static boolean analyzeFuncExprArgsForOneConstAndVar(AbstractFunctionCallExpression funcExpr,
+ AccessMethodAnalysisContext analysisCtx) {
+ IAlgebricksConstantValue constFilterVal = null;
+ LogicalVariable fieldVar = null;
+ ILogicalExpression arg1 = funcExpr.getArguments().get(0).getValue();
+ ILogicalExpression arg2 = funcExpr.getArguments().get(1).getValue();
+ // One of the args must be a constant, and the other arg must be a variable.
+ if (arg1.getExpressionTag() == LogicalExpressionTag.CONSTANT
+ && arg2.getExpressionTag() == LogicalExpressionTag.VARIABLE) {
+ ConstantExpression constExpr = (ConstantExpression) arg1;
+ constFilterVal = constExpr.getValue();
+ VariableReferenceExpression varExpr = (VariableReferenceExpression) arg2;
+ fieldVar = varExpr.getVariableReference();
+ } else if (arg1.getExpressionTag() == LogicalExpressionTag.VARIABLE
+ && arg2.getExpressionTag() == LogicalExpressionTag.CONSTANT) {
+ ConstantExpression constExpr = (ConstantExpression) arg2;
+ constFilterVal = constExpr.getValue();
+ VariableReferenceExpression varExpr = (VariableReferenceExpression) arg1;
+ fieldVar = varExpr.getVariableReference();
+ } else {
+ return false;
+ }
+ analysisCtx.matchedFuncExprs.add(new OptimizableFuncExpr(funcExpr, fieldVar, constFilterVal));
+ return true;
+ }
- /**
- * Appends the types of the fields produced by the given secondary index to dest.
- */
- public static void appendSecondaryIndexTypes(Dataset dataset, ARecordType recordType, Index index,
- boolean primaryKeysOnly, List<Object> dest) throws AlgebricksException {
- if (!primaryKeysOnly) {
- switch (index.getIndexType()) {
- case BTREE:
- case SINGLE_PARTITION_WORD_INVIX:
- case SINGLE_PARTITION_NGRAM_INVIX: {
- for (String sk : index.getKeyFieldNames()) {
- Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(sk, recordType);
- dest.add(keyPairType.first);
- }
- break;
- }
- case RTREE: {
- Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(index.getKeyFieldNames()
- .get(0), recordType);
- IAType keyType = keyPairType.first;
- IAType nestedKeyType = NonTaggedFormatUtil.getNestedSpatialType(keyType.getTypeTag());
- int numKeys = getNumSecondaryKeys(index, recordType);
- for (int i = 0; i < numKeys; i++) {
- dest.add(nestedKeyType);
- }
- break;
- }
- }
- }
- // Primary keys.
- List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
- for (String partitioningKey : partitioningKeys) {
- try {
- dest.add(recordType.getFieldType(partitioningKey));
- } catch (IOException e) {
- throw new AlgebricksException(e);
- }
- }
- }
+ public static boolean analyzeFuncExprArgsForTwoVars(AbstractFunctionCallExpression funcExpr,
+ AccessMethodAnalysisContext analysisCtx) {
+ LogicalVariable fieldVar1 = null;
+ LogicalVariable fieldVar2 = null;
+ ILogicalExpression arg1 = funcExpr.getArguments().get(0).getValue();
+ ILogicalExpression arg2 = funcExpr.getArguments().get(1).getValue();
+ if (arg1.getExpressionTag() == LogicalExpressionTag.VARIABLE
+ && arg2.getExpressionTag() == LogicalExpressionTag.VARIABLE) {
+ fieldVar1 = ((VariableReferenceExpression) arg1).getVariableReference();
+ fieldVar2 = ((VariableReferenceExpression) arg2).getVariableReference();
+ } else {
+ return false;
+ }
+ analysisCtx.matchedFuncExprs.add(new OptimizableFuncExpr(funcExpr,
+ new LogicalVariable[] { fieldVar1, fieldVar2 }, null));
+ return true;
+ }
- public static void appendSecondaryIndexOutputVars(Dataset dataset, ARecordType recordType, Index index,
- boolean primaryKeysOnly, IOptimizationContext context, List<LogicalVariable> dest)
- throws AlgebricksException {
- int numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
- int numSecondaryKeys = getNumSecondaryKeys(index, recordType);
- int numVars = (primaryKeysOnly) ? numPrimaryKeys : numPrimaryKeys + numSecondaryKeys;
- for (int i = 0; i < numVars; i++) {
- dest.add(context.newVar());
- }
- }
+ public static int getNumSecondaryKeys(Index index, ARecordType recordType) throws AlgebricksException {
+ switch (index.getIndexType()) {
+ case BTREE:
+ case SINGLE_PARTITION_WORD_INVIX:
+ case SINGLE_PARTITION_NGRAM_INVIX:
+ case LENGTH_PARTITIONED_WORD_INVIX:
+ case LENGTH_PARTITIONED_NGRAM_INVIX: {
+ return index.getKeyFieldNames().size();
+ }
+ case RTREE: {
+ Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(index.getKeyFieldNames().get(0),
+ recordType);
+ IAType keyType = keyPairType.first;
+ int numDimensions = NonTaggedFormatUtil.getNumDimensions(keyType.getTypeTag());
+ return numDimensions * 2;
+ }
+ default: {
+ throw new AlgebricksException("Unknown index kind: " + index.getIndexType());
+ }
+ }
+ }
- public static List<LogicalVariable> getPrimaryKeyVarsFromUnnestMap(Dataset dataset, ILogicalOperator unnestMapOp) {
- int numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
- List<LogicalVariable> primaryKeyVars = new ArrayList<LogicalVariable>();
- List<LogicalVariable> sourceVars = ((UnnestMapOperator) unnestMapOp).getVariables();
- // Assumes the primary keys are located at the end.
- int start = sourceVars.size() - numPrimaryKeys;
- int stop = sourceVars.size();
- for (int i = start; i < stop; i++) {
- primaryKeyVars.add(sourceVars.get(i));
- }
- return primaryKeyVars;
- }
+ /**
+ * Appends the types of the fields produced by the given secondary index to dest.
+ */
+ public static void appendSecondaryIndexTypes(Dataset dataset, ARecordType recordType, Index index,
+ boolean primaryKeysOnly, List<Object> dest) throws AlgebricksException {
+ if (!primaryKeysOnly) {
+ switch (index.getIndexType()) {
+ case BTREE:
+ case SINGLE_PARTITION_WORD_INVIX:
+ case SINGLE_PARTITION_NGRAM_INVIX: {
+ for (String sk : index.getKeyFieldNames()) {
+ Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(sk, recordType);
+ dest.add(keyPairType.first);
+ }
+ break;
+ }
+ case RTREE: {
+ Pair<IAType, Boolean> keyPairType = Index.getNonNullableKeyFieldType(index.getKeyFieldNames()
+ .get(0), recordType);
+ IAType keyType = keyPairType.first;
+ IAType nestedKeyType = NonTaggedFormatUtil.getNestedSpatialType(keyType.getTypeTag());
+ int numKeys = getNumSecondaryKeys(index, recordType);
+ for (int i = 0; i < numKeys; i++) {
+ dest.add(nestedKeyType);
+ }
+ break;
+ }
+ }
+ }
+ // Primary keys.
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ //add primary keys
+ appendExternalRecPrimaryKey(dataset, dest);
+ }
+ else
+ {
+ List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
+ for (String partitioningKey : partitioningKeys) {
+ try {
+ dest.add(recordType.getFieldType(partitioningKey));
+ } catch (IOException e) {
+ throw new AlgebricksException(e);
+ }
+ }
+ }
+ }
- /**
- * Returns the search key expression which feeds a secondary-index search. If we are optimizing a selection query then this method returns
- * the a ConstantExpression from the first constant value in the optimizable function expression.
- * If we are optimizing a join, then this method returns the VariableReferenceExpression that should feed the secondary index probe.
- */
- public static ILogicalExpression createSearchKeyExpr(IOptimizableFuncExpr optFuncExpr,
- OptimizableOperatorSubTree indexSubTree, OptimizableOperatorSubTree probeSubTree) {
- if (probeSubTree == null) {
- // We are optimizing a selection query. Search key is a constant.
- return new ConstantExpression(optFuncExpr.getConstantVal(0));
- } else {
- // We are optimizing a join query. Determine which variable feeds the secondary index.
- if (optFuncExpr.getOperatorSubTree(0) == null || optFuncExpr.getOperatorSubTree(0) == probeSubTree) {
- return new VariableReferenceExpression(optFuncExpr.getLogicalVar(0));
- } else {
- return new VariableReferenceExpression(optFuncExpr.getLogicalVar(1));
- }
- }
- }
+ public static void appendSecondaryIndexOutputVars(Dataset dataset, ARecordType recordType, Index index,
+ boolean primaryKeysOnly, IOptimizationContext context, List<LogicalVariable> dest)
+ throws AlgebricksException {
+ int numPrimaryKeys=0;
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ numPrimaryKeys = DatasetUtils.getExternalRIDSize(dataset);
+ }
+ else
+ {
+ numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
+ }
+ int numSecondaryKeys = getNumSecondaryKeys(index, recordType);
+ int numVars = (primaryKeysOnly) ? numPrimaryKeys : numPrimaryKeys + numSecondaryKeys;
+ for (int i = 0; i < numVars; i++) {
+ dest.add(context.newVar());
+ }
+ }
- /**
- * Returns the first expr optimizable by this index.
- */
- public static IOptimizableFuncExpr chooseFirstOptFuncExpr(Index chosenIndex, AccessMethodAnalysisContext analysisCtx) {
- List<Integer> indexExprs = analysisCtx.getIndexExprs(chosenIndex);
- int firstExprIndex = indexExprs.get(0);
- return analysisCtx.matchedFuncExprs.get(firstExprIndex);
- }
+ public static List<LogicalVariable> getPrimaryKeyVarsFromUnnestMap(Dataset dataset, ILogicalOperator unnestMapOp) {
+ int numPrimaryKeys;
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ numPrimaryKeys = DatasetUtils.getExternalRIDSize(dataset);
+ }
+ else
+ {
+ numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
+ }
+ List<LogicalVariable> primaryKeyVars = new ArrayList<LogicalVariable>();
+ List<LogicalVariable> sourceVars = ((UnnestMapOperator) unnestMapOp).getVariables();
+ // Assumes the primary keys are located at the end.
+ int start = sourceVars.size() - numPrimaryKeys;
+ int stop = sourceVars.size();
+ for (int i = start; i < stop; i++) {
+ primaryKeyVars.add(sourceVars.get(i));
+ }
+ return primaryKeyVars;
+ }
- public static UnnestMapOperator createSecondaryIndexUnnestMap(Dataset dataset, ARecordType recordType, Index index,
- ILogicalOperator inputOp, AccessMethodJobGenParams jobGenParams, IOptimizationContext context,
- boolean outputPrimaryKeysOnly, boolean retainInput) throws AlgebricksException {
- // The job gen parameters are transferred to the actual job gen via the UnnestMapOperator's function arguments.
- ArrayList<Mutable<ILogicalExpression>> secondaryIndexFuncArgs = new ArrayList<Mutable<ILogicalExpression>>();
- jobGenParams.writeToFuncArgs(secondaryIndexFuncArgs);
- // Variables and types coming out of the secondary-index search.
- List<LogicalVariable> secondaryIndexUnnestVars = new ArrayList<LogicalVariable>();
- List<Object> secondaryIndexOutputTypes = new ArrayList<Object>();
- // Append output variables/types generated by the secondary-index search (not forwarded from input).
- appendSecondaryIndexOutputVars(dataset, recordType, index, outputPrimaryKeysOnly, context,
- secondaryIndexUnnestVars);
- appendSecondaryIndexTypes(dataset, recordType, index, outputPrimaryKeysOnly, secondaryIndexOutputTypes);
- // An index search is expressed as an unnest over an index-search function.
- IFunctionInfo secondaryIndexSearch = FunctionUtils.getFunctionInfo(AsterixBuiltinFunctions.INDEX_SEARCH);
- UnnestingFunctionCallExpression secondaryIndexSearchFunc = new UnnestingFunctionCallExpression(
- secondaryIndexSearch, secondaryIndexFuncArgs);
- secondaryIndexSearchFunc.setReturnsUniqueValues(true);
- // This is the operator that jobgen will be looking for. It contains an unnest function that has all necessary arguments to determine
- // which index to use, which variables contain the index-search keys, what is the original dataset, etc.
- UnnestMapOperator secondaryIndexUnnestOp = new UnnestMapOperator(secondaryIndexUnnestVars,
- new MutableObject<ILogicalExpression>(secondaryIndexSearchFunc), secondaryIndexOutputTypes, retainInput);
- secondaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
- context.computeAndSetTypeEnvironmentForOperator(secondaryIndexUnnestOp);
- secondaryIndexUnnestOp.setExecutionMode(ExecutionMode.PARTITIONED);
- return secondaryIndexUnnestOp;
- }
+ /**
+ * Returns the search key expression which feeds a secondary-index search. If we are optimizing a selection query then this method returns
+ * the a ConstantExpression from the first constant value in the optimizable function expression.
+ * If we are optimizing a join, then this method returns the VariableReferenceExpression that should feed the secondary index probe.
+ */
+ public static ILogicalExpression createSearchKeyExpr(IOptimizableFuncExpr optFuncExpr,
+ OptimizableOperatorSubTree indexSubTree, OptimizableOperatorSubTree probeSubTree) {
+ if (probeSubTree == null) {
+ // We are optimizing a selection query. Search key is a constant.
+ return new ConstantExpression(optFuncExpr.getConstantVal(0));
+ } else {
+ // We are optimizing a join query. Determine which variable feeds the secondary index.
+ if (optFuncExpr.getOperatorSubTree(0) == null || optFuncExpr.getOperatorSubTree(0) == probeSubTree) {
+ return new VariableReferenceExpression(optFuncExpr.getLogicalVar(0));
+ } else {
+ return new VariableReferenceExpression(optFuncExpr.getLogicalVar(1));
+ }
+ }
+ }
- public static UnnestMapOperator createPrimaryIndexUnnestMap(DataSourceScanOperator dataSourceScan, Dataset dataset,
- ARecordType recordType, ILogicalOperator inputOp, IOptimizationContext context, boolean sortPrimaryKeys,
- boolean retainInput, boolean requiresBroadcast) throws AlgebricksException {
- List<LogicalVariable> primaryKeyVars = AccessMethodUtils.getPrimaryKeyVarsFromUnnestMap(dataset, inputOp);
- // Optionally add a sort on the primary-index keys before searching the primary index.
- OrderOperator order = null;
- if (sortPrimaryKeys) {
- order = new OrderOperator();
- for (LogicalVariable pkVar : primaryKeyVars) {
- Mutable<ILogicalExpression> vRef = new MutableObject<ILogicalExpression>(
- new VariableReferenceExpression(pkVar));
- order.getOrderExpressions().add(
- new Pair<IOrder, Mutable<ILogicalExpression>>(OrderOperator.ASC_ORDER, vRef));
- }
- // The secondary-index search feeds into the sort.
- order.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
- order.setExecutionMode(ExecutionMode.LOCAL);
- context.computeAndSetTypeEnvironmentForOperator(order);
- }
- // The job gen parameters are transferred to the actual job gen via the UnnestMapOperator's function arguments.
- List<Mutable<ILogicalExpression>> primaryIndexFuncArgs = new ArrayList<Mutable<ILogicalExpression>>();
- BTreeJobGenParams jobGenParams = new BTreeJobGenParams(dataset.getDatasetName(), IndexType.BTREE,
- dataset.getDataverseName(), dataset.getDatasetName(), retainInput, requiresBroadcast);
- // Set low/high inclusive to true for a point lookup.
- jobGenParams.setLowKeyInclusive(true);
- jobGenParams.setHighKeyInclusive(true);
- jobGenParams.setLowKeyVarList(primaryKeyVars, 0, primaryKeyVars.size());
- jobGenParams.setHighKeyVarList(primaryKeyVars, 0, primaryKeyVars.size());
- jobGenParams.writeToFuncArgs(primaryIndexFuncArgs);
- // Variables and types coming out of the primary-index search.
- List<LogicalVariable> primaryIndexUnnestVars = new ArrayList<LogicalVariable>();
- List<Object> primaryIndexOutputTypes = new ArrayList<Object>();
- // Append output variables/types generated by the primary-index search (not forwarded from input).
- primaryIndexUnnestVars.addAll(dataSourceScan.getVariables());
- try {
- appendPrimaryIndexTypes(dataset, recordType, primaryIndexOutputTypes);
- } catch (IOException e) {
- throw new AlgebricksException(e);
- }
- // An index search is expressed as an unnest over an index-search function.
- IFunctionInfo primaryIndexSearch = FunctionUtils.getFunctionInfo(AsterixBuiltinFunctions.INDEX_SEARCH);
- AbstractFunctionCallExpression primaryIndexSearchFunc = new ScalarFunctionCallExpression(primaryIndexSearch,
- primaryIndexFuncArgs);
- // This is the operator that jobgen will be looking for. It contains an unnest function that has all necessary arguments to determine
- // which index to use, which variables contain the index-search keys, what is the original dataset, etc.
- UnnestMapOperator primaryIndexUnnestOp = new UnnestMapOperator(primaryIndexUnnestVars,
- new MutableObject<ILogicalExpression>(primaryIndexSearchFunc), primaryIndexOutputTypes, retainInput);
- // Fed by the order operator or the secondaryIndexUnnestOp.
- if (sortPrimaryKeys) {
- primaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(order));
- } else {
- primaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
- }
- context.computeAndSetTypeEnvironmentForOperator(primaryIndexUnnestOp);
- primaryIndexUnnestOp.setExecutionMode(ExecutionMode.PARTITIONED);
- return primaryIndexUnnestOp;
- }
+ /**
+ * Returns the first expr optimizable by this index.
+ */
+ public static IOptimizableFuncExpr chooseFirstOptFuncExpr(Index chosenIndex, AccessMethodAnalysisContext analysisCtx) {
+ List<Integer> indexExprs = analysisCtx.getIndexExprs(chosenIndex);
+ int firstExprIndex = indexExprs.get(0);
+ return analysisCtx.matchedFuncExprs.get(firstExprIndex);
+ }
+
+ public static void writeVarList(List<LogicalVariable> varList, List<Mutable<ILogicalExpression>> funcArgs) {
+ Mutable<ILogicalExpression> numKeysRef = new MutableObject<ILogicalExpression>(new ConstantExpression(
+ new AsterixConstantValue(new AInt32(varList.size()))));
+ funcArgs.add(numKeysRef);
+ for (LogicalVariable keyVar : varList) {
+ Mutable<ILogicalExpression> keyVarRef = new MutableObject<ILogicalExpression>(
+ new VariableReferenceExpression(keyVar));
+ funcArgs.add(keyVarRef);
+ }
+ }
+
+ public static ExternalDataAccessByRIDOperator createExternalDataAccessByRIDUnnestMap(DataSourceScanOperator dataSourceScan, Dataset dataset,
+ ARecordType recordType, ILogicalOperator inputOp, IOptimizationContext context, Index secondaryIndex) throws AlgebricksException {
+ List<LogicalVariable> primaryKeyVars = AccessMethodUtils.getPrimaryKeyVarsFromUnnestMap(dataset, inputOp);
+
+ // add a sort on the RID fields before fetching external data.
+ OrderOperator order = new OrderOperator();
+ for (LogicalVariable pkVar : primaryKeyVars) {
+ Mutable<ILogicalExpression> vRef = new MutableObject<ILogicalExpression>(
+ new VariableReferenceExpression(pkVar));
+ order.getOrderExpressions().add(
+ new Pair<IOrder, Mutable<ILogicalExpression>>(OrderOperator.ASC_ORDER, vRef));
+ }
+ // The secondary-index search feeds into the sort.
+ order.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
+ order.setExecutionMode(ExecutionMode.LOCAL);
+ context.computeAndSetTypeEnvironmentForOperator(order);
+
+ List<Mutable<ILogicalExpression>> externalRIDAccessFuncArgs = new ArrayList<Mutable<ILogicalExpression>>();
+ AccessMethodUtils.writeVarList(primaryKeyVars,externalRIDAccessFuncArgs);
+
+ // Variables and types coming out of the external access.
+ List<LogicalVariable> externalAccessByRIDVars = new ArrayList<LogicalVariable>();
+ List<Object> externalAccessOutputTypes = new ArrayList<Object>();
+ // Append output variables/types generated by the data scan (not forwarded from input).
+ externalAccessByRIDVars.addAll(dataSourceScan.getVariables());
+ appendExternalRecTypes(dataset, recordType, externalAccessOutputTypes);
+
+ IFunctionInfo externalAccessByRID = FunctionUtils.getFunctionInfo(AsterixBuiltinFunctions.EXTERNAL_ACCESS_BY_RID);
+ AbstractFunctionCallExpression externalAccessFunc = new ScalarFunctionCallExpression(externalAccessByRID,
+ externalRIDAccessFuncArgs);
+
+ ExternalDataAccessByRIDOperator externalAccessByRIDOp = new ExternalDataAccessByRIDOperator(externalAccessByRIDVars,
+ new MutableObject<ILogicalExpression>(externalAccessFunc), externalAccessOutputTypes);
+ // Fed by the order operator or the secondaryIndexUnnestOp.
+ externalAccessByRIDOp.getInputs().add(new MutableObject<ILogicalOperator>(order));
+
+ context.computeAndSetTypeEnvironmentForOperator(externalAccessByRIDOp);
+ externalAccessByRIDOp.setExecutionMode(ExecutionMode.PARTITIONED);
+
+ //set the physical operator
+ AqlSourceId dataSourceId = new AqlSourceId(dataset.getDataverseName(),
+ dataset.getDatasetName());
+ externalAccessByRIDOp.setPhysicalOperator(new ExternalDataAccessByRIDPOperator(dataSourceId, dataset, recordType, secondaryIndex));
+ return externalAccessByRIDOp;
+ }
+
+ public static UnnestMapOperator createSecondaryIndexUnnestMap(Dataset dataset, ARecordType recordType, Index index,
+ ILogicalOperator inputOp, AccessMethodJobGenParams jobGenParams, IOptimizationContext context,
+ boolean outputPrimaryKeysOnly, boolean retainInput) throws AlgebricksException {
+ // The job gen parameters are transferred to the actual job gen via the UnnestMapOperator's function arguments.
+ ArrayList<Mutable<ILogicalExpression>> secondaryIndexFuncArgs = new ArrayList<Mutable<ILogicalExpression>>();
+ jobGenParams.writeToFuncArgs(secondaryIndexFuncArgs);
+ // Variables and types coming out of the secondary-index search.
+ List<LogicalVariable> secondaryIndexUnnestVars = new ArrayList<LogicalVariable>();
+ List<Object> secondaryIndexOutputTypes = new ArrayList<Object>();
+ // Append output variables/types generated by the secondary-index search (not forwarded from input).
+ appendSecondaryIndexOutputVars(dataset, recordType, index, outputPrimaryKeysOnly, context,
+ secondaryIndexUnnestVars);
+ appendSecondaryIndexTypes(dataset, recordType, index, outputPrimaryKeysOnly, secondaryIndexOutputTypes);
+ // An index search is expressed as an unnest over an index-search function.
+ IFunctionInfo secondaryIndexSearch = FunctionUtils.getFunctionInfo(AsterixBuiltinFunctions.INDEX_SEARCH);
+ UnnestingFunctionCallExpression secondaryIndexSearchFunc = new UnnestingFunctionCallExpression(
+ secondaryIndexSearch, secondaryIndexFuncArgs);
+ secondaryIndexSearchFunc.setReturnsUniqueValues(true);
+ // This is the operator that jobgen will be looking for. It contains an unnest function that has all necessary arguments to determine
+ // which index to use, which variables contain the index-search keys, what is the original dataset, etc.
+ UnnestMapOperator secondaryIndexUnnestOp = new UnnestMapOperator(secondaryIndexUnnestVars,
+ new MutableObject<ILogicalExpression>(secondaryIndexSearchFunc), secondaryIndexOutputTypes, retainInput);
+ secondaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
+ context.computeAndSetTypeEnvironmentForOperator(secondaryIndexUnnestOp);
+ secondaryIndexUnnestOp.setExecutionMode(ExecutionMode.PARTITIONED);
+ return secondaryIndexUnnestOp;
+ }
+
+ public static UnnestMapOperator createPrimaryIndexUnnestMap(DataSourceScanOperator dataSourceScan, Dataset dataset,
+ ARecordType recordType, ILogicalOperator inputOp, IOptimizationContext context, boolean sortPrimaryKeys,
+ boolean retainInput, boolean requiresBroadcast) throws AlgebricksException {
+ List<LogicalVariable> primaryKeyVars = AccessMethodUtils.getPrimaryKeyVarsFromUnnestMap(dataset, inputOp);
+ // Optionally add a sort on the primary-index keys before searching the primary index.
+ OrderOperator order = null;
+ if (sortPrimaryKeys) {
+ order = new OrderOperator();
+ for (LogicalVariable pkVar : primaryKeyVars) {
+ Mutable<ILogicalExpression> vRef = new MutableObject<ILogicalExpression>(
+ new VariableReferenceExpression(pkVar));
+ order.getOrderExpressions().add(
+ new Pair<IOrder, Mutable<ILogicalExpression>>(OrderOperator.ASC_ORDER, vRef));
+ }
+ // The secondary-index search feeds into the sort.
+ order.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
+ order.setExecutionMode(ExecutionMode.LOCAL);
+ context.computeAndSetTypeEnvironmentForOperator(order);
+ }
+ // The job gen parameters are transferred to the actual job gen via the UnnestMapOperator's function arguments.
+ List<Mutable<ILogicalExpression>> primaryIndexFuncArgs = new ArrayList<Mutable<ILogicalExpression>>();
+ BTreeJobGenParams jobGenParams = new BTreeJobGenParams(dataset.getDatasetName(), IndexType.BTREE,
+ dataset.getDataverseName(), dataset.getDatasetName(), retainInput, requiresBroadcast);
+ // Set low/high inclusive to true for a point lookup.
+ jobGenParams.setLowKeyInclusive(true);
+ jobGenParams.setHighKeyInclusive(true);
+ jobGenParams.setLowKeyVarList(primaryKeyVars, 0, primaryKeyVars.size());
+ jobGenParams.setHighKeyVarList(primaryKeyVars, 0, primaryKeyVars.size());
+ jobGenParams.writeToFuncArgs(primaryIndexFuncArgs);
+ // Variables and types coming out of the primary-index search.
+ List<LogicalVariable> primaryIndexUnnestVars = new ArrayList<LogicalVariable>();
+ List<Object> primaryIndexOutputTypes = new ArrayList<Object>();
+ // Append output variables/types generated by the primary-index search (not forwarded from input).
+ primaryIndexUnnestVars.addAll(dataSourceScan.getVariables());
+ try {
+ appendPrimaryIndexTypes(dataset, recordType, primaryIndexOutputTypes);
+ } catch (IOException e) {
+ throw new AlgebricksException(e);
+ }
+ // An index search is expressed as an unnest over an index-search function.
+ IFunctionInfo primaryIndexSearch = FunctionUtils.getFunctionInfo(AsterixBuiltinFunctions.INDEX_SEARCH);
+ AbstractFunctionCallExpression primaryIndexSearchFunc = new ScalarFunctionCallExpression(primaryIndexSearch,
+ primaryIndexFuncArgs);
+ // This is the operator that jobgen will be looking for. It contains an unnest function that has all necessary arguments to determine
+ // which index to use, which variables contain the index-search keys, what is the original dataset, etc.
+ UnnestMapOperator primaryIndexUnnestOp = new UnnestMapOperator(primaryIndexUnnestVars,
+ new MutableObject<ILogicalExpression>(primaryIndexSearchFunc), primaryIndexOutputTypes, retainInput);
+ // Fed by the order operator or the secondaryIndexUnnestOp.
+ if (sortPrimaryKeys) {
+ primaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(order));
+ } else {
+ primaryIndexUnnestOp.getInputs().add(new MutableObject<ILogicalOperator>(inputOp));
+ }
+ context.computeAndSetTypeEnvironmentForOperator(primaryIndexUnnestOp);
+ primaryIndexUnnestOp.setExecutionMode(ExecutionMode.PARTITIONED);
+ return primaryIndexUnnestOp;
+ }
}
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/BTreeAccessMethod.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/BTreeAccessMethod.java
index ddcf768..ce2a1f7 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/BTreeAccessMethod.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/BTreeAccessMethod.java
@@ -27,6 +27,7 @@
import org.apache.commons.lang3.mutable.Mutable;
import org.apache.commons.lang3.mutable.MutableObject;
+import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.DatasetConfig.IndexType;
import edu.uci.ics.asterix.metadata.entities.Dataset;
import edu.uci.ics.asterix.metadata.entities.Index;
@@ -51,6 +52,7 @@
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator.ExecutionMode;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AssignOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExternalDataAccessByRIDOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.SelectOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.UnnestMapOperator;
@@ -416,10 +418,16 @@
// Generate the rest of the upstream plan which feeds the search results into the primary index.
UnnestMapOperator primaryIndexUnnestOp;
boolean isPrimaryIndex = chosenIndex.getIndexName().equals(dataset.getDatasetName());
- if (!isPrimaryIndex) {
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ ExternalDataAccessByRIDOperator externalDataAccessOp = AccessMethodUtils.createExternalDataAccessByRIDUnnestMap(dataSourceScan, dataset,
+ recordType, secondaryIndexUnnestOp, context, chosenIndex);
+ indexSubTree.dataSourceScanRef.setValue(externalDataAccessOp);
+ return externalDataAccessOp;
+ }
+ else if (!isPrimaryIndex) {
primaryIndexUnnestOp = AccessMethodUtils.createPrimaryIndexUnnestMap(dataSourceScan, dataset, recordType,
secondaryIndexUnnestOp, context, true, retainInput, false);
-
// Replace the datasource scan with the new plan rooted at
// primaryIndexUnnestMap.
indexSubTree.dataSourceScanRef.setValue(primaryIndexUnnestOp); //kisskys
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/OptimizableOperatorSubTree.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/OptimizableOperatorSubTree.java
index 28aee7a..ea2adad 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/OptimizableOperatorSubTree.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/OptimizableOperatorSubTree.java
@@ -114,7 +114,8 @@
if (dataset == null) {
throw new AlgebricksException("No metadata for dataset " + datasetName);
}
- if (dataset.getDatasetType() != DatasetType.INTERNAL && dataset.getDatasetType() != DatasetType.FEED) {
+ if (dataset.getDatasetType() != DatasetType.INTERNAL && dataset.getDatasetType() != DatasetType.FEED
+ && dataset.getDatasetType() != DatasetType.EXTERNAL ) {
return false;
}
// Get the record type for that dataset.
diff --git a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/RTreeAccessMethod.java b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/RTreeAccessMethod.java
index b8125aa..8af4ac1 100644
--- a/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/RTreeAccessMethod.java
+++ b/asterix-algebra/src/main/java/edu/uci/ics/asterix/optimizer/rules/am/RTreeAccessMethod.java
@@ -21,6 +21,7 @@
import org.apache.commons.lang3.mutable.MutableObject;
import edu.uci.ics.asterix.aql.util.FunctionUtils;
+import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.DatasetConfig.IndexType;
import edu.uci.ics.asterix.metadata.entities.Dataset;
import edu.uci.ics.asterix.metadata.entities.Index;
@@ -44,6 +45,7 @@
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator.ExecutionMode;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.AssignOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator;
+import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.ExternalDataAccessByRIDOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.SelectOperator;
import edu.uci.ics.hyracks.algebricks.core.algebra.operators.logical.UnnestMapOperator;
@@ -196,10 +198,19 @@
UnnestMapOperator secondaryIndexUnnestOp = AccessMethodUtils.createSecondaryIndexUnnestMap(dataset, recordType,
chosenIndex, assignSearchKeys, jobGenParams, context, false, retainInput);
// Generate the rest of the upstream plan which feeds the search results into the primary index.
- UnnestMapOperator primaryIndexUnnestOp = AccessMethodUtils.createPrimaryIndexUnnestMap(dataSourceScan, dataset,
- recordType, secondaryIndexUnnestOp, context, true, retainInput, false);
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ ExternalDataAccessByRIDOperator externalDataAccessOp = AccessMethodUtils.createExternalDataAccessByRIDUnnestMap(dataSourceScan, dataset,
+ recordType, secondaryIndexUnnestOp, context, chosenIndex);
+ return externalDataAccessOp;
+ }
+ else
+ {
+ UnnestMapOperator primaryIndexUnnestOp = AccessMethodUtils.createPrimaryIndexUnnestMap(dataSourceScan, dataset,
+ recordType, secondaryIndexUnnestOp, context, true, retainInput, false);
- return primaryIndexUnnestOp;
+ return primaryIndexUnnestOp;
+ }
}
@Override
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/aql/translator/AqlTranslator.java b/asterix-app/src/main/java/edu/uci/ics/asterix/aql/translator/AqlTranslator.java
index f22d2fb..9c3a853 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/aql/translator/AqlTranslator.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/aql/translator/AqlTranslator.java
@@ -414,7 +414,8 @@
case EXTERNAL: {
String adapter = ((ExternalDetailsDecl) dd.getDatasetDetailsDecl()).getAdapter();
Map<String, String> properties = ((ExternalDetailsDecl) dd.getDatasetDetailsDecl()).getProperties();
- datasetDetails = new ExternalDatasetDetails(adapter, properties);
+ String ngName = ((ExternalDetailsDecl) dd.getDatasetDetailsDecl()).getNodegroupName().getValue();
+ datasetDetails = new ExternalDatasetDetails(adapter, properties,ngName);
break;
}
case FEED: {
@@ -590,6 +591,18 @@
//#. create the index artifact in NC.
runJob(hcc, spec, true);
+ //if external data and optimization is turned on, load file names
+ if(ds.getDatasetType() == DatasetType.EXTERNAL && AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ //load the file names into external files index
+ mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
+ bActiveTxn = true;
+ metadataProvider.setMetadataTxnContext(mdTxnCtx);
+ IndexOperations.addExternalDatasetFilesToMetadata(metadataProvider, ds);
+ MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
+ bActiveTxn = false;
+ }
+
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
@@ -654,6 +667,8 @@
throw new IllegalStateException("System is inconsistent state: pending index(" + dataverseName
+ "." + datasetName + "." + indexName + ") couldn't be removed from the metadata", e);
}
+
+ //if external dataset, remove external files from metadata
}
throw e;
} finally {
@@ -730,7 +745,7 @@
for (int j = 0; j < datasets.size(); j++) {
String datasetName = datasets.get(j).getDatasetName();
DatasetType dsType = datasets.get(j).getDatasetType();
- if (dsType == DatasetType.INTERNAL || dsType == DatasetType.FEED) {
+ if (dsType == DatasetType.INTERNAL || dsType == DatasetType.FEED || dsType == DatasetType.EXTERNAL) {
List<Index> indexes = MetadataManager.INSTANCE.getDatasetIndexes(mdTxnCtx, dataverseName,
datasetName);
@@ -843,7 +858,7 @@
}
}
- if (ds.getDatasetType() == DatasetType.INTERNAL || ds.getDatasetType() == DatasetType.FEED) {
+ if (ds.getDatasetType() == DatasetType.INTERNAL || ds.getDatasetType() == DatasetType.FEED || ds.getDatasetType() == DatasetType.EXTERNAL) {
//#. prepare jobs to drop the datatset and the indexes in NC
List<Index> indexes = MetadataManager.INSTANCE.getDatasetIndexes(mdTxnCtx, dataverseName, datasetName);
@@ -944,7 +959,7 @@
+ dataverseName);
}
- if (ds.getDatasetType() == DatasetType.INTERNAL || ds.getDatasetType() == DatasetType.FEED) {
+ if (ds.getDatasetType() == DatasetType.INTERNAL || ds.getDatasetType() == DatasetType.FEED || ds.getDatasetType() == DatasetType.EXTERNAL) {
indexName = stmtIndexDrop.getIndexName().getValue();
Index index = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataverseName, datasetName, indexName);
if (index == null) {
@@ -982,8 +997,8 @@
//#. finally, delete the existing index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, indexName);
} else {
- throw new AlgebricksException(datasetName
- + " is an external dataset. Indexes are not maintained for external datasets.");
+ //throw new AlgebricksException(datasetName
+ // + " is an external dataset. Indexes are not maintained for external datasets.");
}
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/file/IndexOperations.java b/asterix-app/src/main/java/edu/uci/ics/asterix/file/IndexOperations.java
index af56894..de4d075 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/file/IndexOperations.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/file/IndexOperations.java
@@ -14,14 +14,18 @@
*/
package edu.uci.ics.asterix.file;
+import java.util.ArrayList;
+
import edu.uci.ics.asterix.common.config.AsterixStorageProperties;
import edu.uci.ics.asterix.common.config.OptimizationConfUtil;
import edu.uci.ics.asterix.common.context.AsterixVirtualBufferCacheProvider;
import edu.uci.ics.asterix.common.exceptions.AsterixException;
import edu.uci.ics.asterix.common.ioopcallbacks.LSMBTreeIOOperationCallbackFactory;
import edu.uci.ics.asterix.metadata.MetadataException;
+import edu.uci.ics.asterix.metadata.MetadataManager;
import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
import edu.uci.ics.asterix.metadata.entities.Dataset;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.om.util.AsterixAppContextInfo;
import edu.uci.ics.asterix.transaction.management.opcallbacks.SecondaryIndexOperationTrackerProvider;
import edu.uci.ics.asterix.transaction.management.service.transaction.AsterixRuntimeComponentsProvider;
@@ -55,6 +59,17 @@
metadataProvider, physicalOptimizationConfig);
return secondaryIndexCreator.buildLoadingJobSpec();
}
+
+ public static void addExternalDatasetFilesToMetadata(AqlMetadataProvider metadataProvider,
+ Dataset dataset) throws AlgebricksException, MetadataException{
+ //get the file list
+ ArrayList<ExternalFile> files = metadataProvider.getExternalDatasetFiles(dataset);
+ //add files to metadata
+ for(int i=0; i < files.size(); i++)
+ {
+ MetadataManager.INSTANCE.addExternalFile(metadataProvider.getMetadataTxnContext(), files.get(i));
+ }
+ }
public static JobSpecification buildDropSecondaryIndexJobSpec(CompiledIndexDropStatement indexDropStmt,
AqlMetadataProvider metadataProvider, Dataset dataset) throws AlgebricksException, MetadataException {
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryBTreeCreator.java b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryBTreeCreator.java
index e3832d4..baf16de 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryBTreeCreator.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryBTreeCreator.java
@@ -17,20 +17,31 @@
import edu.uci.ics.asterix.common.api.ILocalResourceMetadata;
import edu.uci.ics.asterix.common.config.AsterixStorageProperties;
import edu.uci.ics.asterix.common.config.IAsterixPropertiesProvider;
+import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.context.AsterixVirtualBufferCacheProvider;
import edu.uci.ics.asterix.common.exceptions.AsterixException;
import edu.uci.ics.asterix.common.ioopcallbacks.LSMBTreeIOOperationCallbackFactory;
import edu.uci.ics.asterix.transaction.management.opcallbacks.SecondaryIndexOperationTrackerProvider;
+
+import edu.uci.ics.asterix.external.data.operator.ExternalDataIndexingOperatorDescriptor;
+import edu.uci.ics.asterix.external.util.ExternalIndexHashPartitionComputerFactory;
+import edu.uci.ics.asterix.metadata.utils.DatasetUtils;
+import edu.uci.ics.asterix.runtime.formats.NonTaggedDataFormat;
import edu.uci.ics.asterix.transaction.management.resource.LSMBTreeLocalResourceMetadata;
import edu.uci.ics.asterix.transaction.management.resource.PersistentLocalResourceFactoryProvider;
import edu.uci.ics.asterix.transaction.management.service.transaction.AsterixRuntimeComponentsProvider;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraintHelper;
import edu.uci.ics.hyracks.algebricks.common.exceptions.AlgebricksException;
+import edu.uci.ics.hyracks.algebricks.common.utils.Pair;
import edu.uci.ics.hyracks.algebricks.core.jobgen.impl.ConnectorPolicyAssignmentPolicy;
import edu.uci.ics.hyracks.algebricks.core.rewriter.base.PhysicalOptimizationConfig;
import edu.uci.ics.hyracks.algebricks.runtime.operators.meta.AlgebricksMetaOperatorDescriptor;
+import edu.uci.ics.hyracks.api.dataflow.IConnectorDescriptor;
+import edu.uci.ics.hyracks.api.dataflow.value.IBinaryHashFunctionFactory;
import edu.uci.ics.hyracks.api.job.JobSpecification;
import edu.uci.ics.hyracks.dataflow.std.base.AbstractOperatorDescriptor;
+import edu.uci.ics.hyracks.dataflow.std.connectors.MToNPartitioningConnectorDescriptor;
import edu.uci.ics.hyracks.dataflow.std.connectors.OneToOneConnectorDescriptor;
import edu.uci.ics.hyracks.dataflow.std.sort.ExternalSortOperatorDescriptor;
import edu.uci.ics.hyracks.storage.am.btree.dataflow.BTreeSearchOperatorDescriptor;
@@ -49,18 +60,16 @@
super(physOptConf, propertiesProvider);
}
- @Override
- public JobSpecification buildCreationJobSpec() throws AsterixException, AlgebricksException {
- JobSpecification spec = JobSpecificationUtils.createJobSpecification();
-
- AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
- //prepare a LocalResourceMetadata which will be stored in NC's local resource repository
- ILocalResourceMetadata localResourceMetadata = new LSMBTreeLocalResourceMetadata(
- secondaryRecDesc.getTypeTraits(), secondaryComparatorFactories, secondaryBloomFilterKeyFields, true,
- dataset.getDatasetId());
- ILocalResourceFactoryProvider localResourceFactoryProvider = new PersistentLocalResourceFactoryProvider(
- localResourceMetadata, LocalResource.LSMBTreeResource);
-
+ @Override
+ public JobSpecification buildCreationJobSpec() throws AsterixException, AlgebricksException {
+ JobSpecification spec = JobSpecificationUtils.createJobSpecification();
+ AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
+ //prepare a LocalResourceMetadata which will be stored in NC's local resource repository
+ ILocalResourceMetadata localResourceMetadata = new LSMBTreeLocalResourceMetadata(
+ secondaryRecDesc.getTypeTraits(), secondaryComparatorFactories, secondaryBloomFilterKeyFields, true,
+ dataset.getDatasetId());
+ ILocalResourceFactoryProvider localResourceFactoryProvider = new PersistentLocalResourceFactoryProvider(
+ localResourceMetadata, LocalResource.LSMBTreeResource);
TreeIndexCreateOperatorDescriptor secondaryIndexCreateOp = new TreeIndexCreateOperatorDescriptor(spec,
AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER,
secondaryFileSplitProvider, secondaryRecDesc.getTypeTraits(), secondaryComparatorFactories,
@@ -78,30 +87,93 @@
return spec;
}
- @Override
- public JobSpecification buildLoadingJobSpec() throws AsterixException, AlgebricksException {
- JobSpecification spec = JobSpecificationUtils.createJobSpecification();
- // Create dummy key provider for feeding the primary index scan.
- AbstractOperatorDescriptor keyProviderOp = createDummyKeyProviderOp(spec);
+ @Override
+ public JobSpecification buildLoadingJobSpec() throws AsterixException, AlgebricksException{
+ if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
+ JobSpecification spec = JobSpecificationUtils.createJobSpecification();
+ Pair<ExternalDataIndexingOperatorDescriptor, AlgebricksPartitionConstraint> RIDScanOpAndConstraints;
+ AlgebricksMetaOperatorDescriptor asterixAssignOp;
+ try
+ {
+ //create external indexing scan operator
+ RIDScanOpAndConstraints = createExternalIndexingOp(spec);
- // Create primary index scan op.
- BTreeSearchOperatorDescriptor primaryScanOp = createPrimaryIndexScanOp(spec);
+ //create assign operator
+ asterixAssignOp = createExternalAssignOp(spec);
+ AlgebricksPartitionConstraintHelper.setPartitionConstraintInJobSpec(spec, asterixAssignOp,
+ RIDScanOpAndConstraints.second);
+ }
+ catch(Exception e)
+ {
+ throw new AsterixException("Failed to create external index scanning and loading job");
+ }
- // Assign op.
- AlgebricksMetaOperatorDescriptor asterixAssignOp = createAssignOp(spec, primaryScanOp, numSecondaryKeys);
+ // If any of the secondary fields are nullable, then add a select op that filters nulls.
+ AlgebricksMetaOperatorDescriptor selectOp = null;
+ if (anySecondaryKeyIsNullable) {
+ selectOp = createFilterNullsSelectOp(spec, numSecondaryKeys);
+ }
- // If any of the secondary fields are nullable, then add a select op that filters nulls.
- AlgebricksMetaOperatorDescriptor selectOp = null;
- if (anySecondaryKeyIsNullable) {
- selectOp = createFilterNullsSelectOp(spec, numSecondaryKeys);
- }
+ // Sort by secondary keys.
+ ExternalSortOperatorDescriptor sortOp = createSortOp(spec, secondaryComparatorFactories, secondaryRecDesc,RIDScanOpAndConstraints.second);
+ AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
+ // Create secondary BTree bulk load op.
+ TreeIndexBulkLoadOperatorDescriptor secondaryBulkLoadOp = createTreeIndexBulkLoadOp(
+ spec,
+ numSecondaryKeys,
+ new LSMBTreeDataflowHelperFactory(new AsterixVirtualBufferCacheProvider(dataset.getDatasetId()),
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, new SecondaryIndexOperationTrackerProvider(
+ LSMBTreeIOOperationCallbackFactory.INSTANCE, dataset.getDatasetId()),
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER,
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, storageProperties
+ .getBloomFilterFalsePositiveRate()), BTree.DEFAULT_FILL_FACTOR);
+ IBinaryHashFunctionFactory[] hashFactories = DatasetUtils.computeExternalDataKeysBinaryHashFunFactories(dataset, NonTaggedDataFormat.INSTANCE.getBinaryHashFunctionFactoryProvider());
- // Sort by secondary keys.
- ExternalSortOperatorDescriptor sortOp = createSortOp(spec, secondaryComparatorFactories, secondaryRecDesc);
+ //select partitioning keys (always the first 2 after secondary keys)
+ int[] keys = new int[2];
+ keys[0] = numSecondaryKeys;
+ keys[1] = numSecondaryKeys + 1;
- AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
- // Create secondary BTree bulk load op.
+ IConnectorDescriptor hashConn = new MToNPartitioningConnectorDescriptor(spec,
+ new ExternalIndexHashPartitionComputerFactory(keys, hashFactories));
+
+ spec.connect(new OneToOneConnectorDescriptor(spec), RIDScanOpAndConstraints.first, 0, asterixAssignOp, 0);
+ if (anySecondaryKeyIsNullable) {
+ spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, selectOp, 0);
+ spec.connect(hashConn, selectOp, 0, sortOp, 0);
+ } else {
+ spec.connect(hashConn, asterixAssignOp, 0, sortOp, 0);
+ }
+ spec.connect(new OneToOneConnectorDescriptor(spec), sortOp, 0, secondaryBulkLoadOp, 0);
+ spec.addRoot(secondaryBulkLoadOp);
+ spec.setConnectorPolicyAssignmentPolicy(new ConnectorPolicyAssignmentPolicy());
+ return spec;
+ }
+ else
+ {
+ JobSpecification spec = JobSpecificationUtils.createJobSpecification();
+
+ // Create dummy key provider for feeding the primary index scan.
+ AbstractOperatorDescriptor keyProviderOp = createDummyKeyProviderOp(spec);
+
+ // Create primary index scan op.
+ BTreeSearchOperatorDescriptor primaryScanOp = createPrimaryIndexScanOp(spec);
+
+ // Assign op.
+ AlgebricksMetaOperatorDescriptor asterixAssignOp = createAssignOp(spec, primaryScanOp, numSecondaryKeys);
+
+ // If any of the secondary fields are nullable, then add a select op that filters nulls.
+ AlgebricksMetaOperatorDescriptor selectOp = null;
+ if (anySecondaryKeyIsNullable) {
+ selectOp = createFilterNullsSelectOp(spec, numSecondaryKeys);
+ }
+
+ // Sort by secondary keys.
+ ExternalSortOperatorDescriptor sortOp = createSortOp(spec, secondaryComparatorFactories, secondaryRecDesc);
+
+ AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
+ // Create secondary BTree bulk load op.
TreeIndexBulkLoadOperatorDescriptor secondaryBulkLoadOp = createTreeIndexBulkLoadOp(
spec,
numSecondaryKeys,
@@ -112,18 +184,20 @@
AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, storageProperties
.getBloomFilterFalsePositiveRate()), BTree.DEFAULT_FILL_FACTOR);
- // Connect the operators.
- spec.connect(new OneToOneConnectorDescriptor(spec), keyProviderOp, 0, primaryScanOp, 0);
- spec.connect(new OneToOneConnectorDescriptor(spec), primaryScanOp, 0, asterixAssignOp, 0);
- if (anySecondaryKeyIsNullable) {
- spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, selectOp, 0);
- spec.connect(new OneToOneConnectorDescriptor(spec), selectOp, 0, sortOp, 0);
- } else {
- spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, sortOp, 0);
- }
- spec.connect(new OneToOneConnectorDescriptor(spec), sortOp, 0, secondaryBulkLoadOp, 0);
- spec.addRoot(secondaryBulkLoadOp);
- spec.setConnectorPolicyAssignmentPolicy(new ConnectorPolicyAssignmentPolicy());
- return spec;
- }
+ // Connect the operators.
+ spec.connect(new OneToOneConnectorDescriptor(spec), keyProviderOp, 0, primaryScanOp, 0);
+ spec.connect(new OneToOneConnectorDescriptor(spec), primaryScanOp, 0, asterixAssignOp, 0);
+ if (anySecondaryKeyIsNullable) {
+ spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, selectOp, 0);
+ spec.connect(new OneToOneConnectorDescriptor(spec), selectOp, 0, sortOp, 0);
+ } else {
+ spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, sortOp, 0);
+ }
+ spec.connect(new OneToOneConnectorDescriptor(spec), sortOp, 0, secondaryBulkLoadOp, 0);
+ spec.addRoot(secondaryBulkLoadOp);
+ spec.setConnectorPolicyAssignmentPolicy(new ConnectorPolicyAssignmentPolicy());
+ return spec;
+ }
+ }
}
+
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryIndexCreator.java b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryIndexCreator.java
index a9b3881..5da336f 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryIndexCreator.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryIndexCreator.java
@@ -19,6 +19,9 @@
import java.io.IOException;
import java.util.List;
+import edu.uci.ics.asterix.external.adapter.factory.HDFSAdapterFactory;
+import edu.uci.ics.asterix.external.data.operator.ExternalDataIndexingOperatorDescriptor;
+import edu.uci.ics.asterix.external.dataset.adapter.HiveAdapter;
import edu.uci.ics.asterix.common.config.AsterixStorageProperties;
import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.IAsterixPropertiesProvider;
@@ -32,11 +35,13 @@
import edu.uci.ics.asterix.formats.nontagged.AqlBinaryComparatorFactoryProvider;
import edu.uci.ics.asterix.formats.nontagged.AqlSerializerDeserializerProvider;
import edu.uci.ics.asterix.formats.nontagged.AqlTypeTraitProvider;
+import edu.uci.ics.asterix.metadata.entities.ExternalDatasetDetails;
import edu.uci.ics.asterix.metadata.MetadataException;
import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
import edu.uci.ics.asterix.metadata.entities.Dataset;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.utils.DatasetUtils;
+import edu.uci.ics.asterix.om.types.BuiltinType;
import edu.uci.ics.asterix.om.types.ARecordType;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.asterix.om.util.AsterixAppContextInfo;
@@ -78,6 +83,7 @@
import edu.uci.ics.hyracks.dataflow.std.file.IFileSplitProvider;
import edu.uci.ics.hyracks.dataflow.std.misc.ConstantTupleSourceOperatorDescriptor;
import edu.uci.ics.hyracks.dataflow.std.sort.ExternalSortOperatorDescriptor;
+import edu.uci.ics.asterix.runtime.formats.NonTaggedDataFormat;
import edu.uci.ics.hyracks.storage.am.btree.dataflow.BTreeSearchOperatorDescriptor;
import edu.uci.ics.hyracks.storage.am.common.api.ISearchOperationCallbackFactory;
import edu.uci.ics.hyracks.storage.am.common.dataflow.IIndexDataflowHelperFactory;
@@ -85,6 +91,7 @@
import edu.uci.ics.hyracks.storage.am.common.impls.NoOpOperationCallbackFactory;
import edu.uci.ics.hyracks.storage.am.lsm.btree.dataflow.LSMBTreeDataflowHelperFactory;
+
@SuppressWarnings("rawtypes")
// TODO: We should eventually have a hierarchy of classes that can create all
// possible index job specs,
@@ -102,11 +109,11 @@
protected ISerializerDeserializer payloadSerde;
protected IFileSplitProvider primaryFileSplitProvider;
protected AlgebricksPartitionConstraint primaryPartitionConstraint;
+ protected List<String> secondaryKeyFields;
protected IFileSplitProvider secondaryFileSplitProvider;
protected AlgebricksPartitionConstraint secondaryPartitionConstraint;
protected String secondaryIndexName;
protected boolean anySecondaryKeyIsNullable = false;
-
protected long numElementsHint;
protected IBinaryComparatorFactory[] primaryComparatorFactories;
protected int[] primaryBloomFilterKeyFields;
@@ -115,7 +122,6 @@
protected int[] secondaryBloomFilterKeyFields;
protected RecordDescriptor secondaryRecDesc;
protected ICopyEvaluatorFactory[] secondaryFieldAccessEvalFactories;
-
protected IAsterixPropertiesProvider propertiesProvider;
// Prevent public construction. Should be created via createIndexCreator().
@@ -159,37 +165,68 @@
public abstract JobSpecification buildLoadingJobSpec() throws AsterixException, AlgebricksException;
protected void init(CompiledCreateIndexStatement createIndexStmt, AqlMetadataProvider metadataProvider)
- throws AsterixException, AlgebricksException {
- this.metadataProvider = metadataProvider;
- dataverseName = createIndexStmt.getDataverseName() == null ? metadataProvider.getDefaultDataverseName()
- : createIndexStmt.getDataverseName();
- datasetName = createIndexStmt.getDatasetName();
- secondaryIndexName = createIndexStmt.getIndexName();
- dataset = metadataProvider.findDataset(dataverseName, datasetName);
- if (dataset == null) {
- throw new AsterixException("Unknown dataset " + datasetName);
+ throws AsterixException, AlgebricksException {
+ this.metadataProvider = metadataProvider;
+ dataverseName = createIndexStmt.getDataverseName() == null ? metadataProvider.getDefaultDataverseName()
+ : createIndexStmt.getDataverseName();
+ datasetName = createIndexStmt.getDatasetName();
+ secondaryIndexName = createIndexStmt.getIndexName();
+ dataset = metadataProvider.findDataset(dataverseName, datasetName);
+ if (dataset == null) {
+ throw new AsterixException("Unknown dataset " + datasetName);
+ }
+ if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
+ //get external dataset details
+ ExternalDatasetDetails edsd = ((ExternalDatasetDetails)dataset.getDatasetDetails());
+ //get adapter name
+ String adapter = edsd.getAdapter();
+ //if not an hdfs adapter, throw an exception
+ if(!adapter.equals(HDFSAdapterFactory.HDFS_ADAPTER_NAME) && !adapter.equals(HiveAdapter.class.getName()))
+ {
+ throw new AsterixException("Cannot index an external dataset with adapter type(" + adapter + ").");
+ }
+ //get the item type
+ ARecordType externalItemType = (ARecordType) metadataProvider.findType(dataset.getDataverseName(), dataset.getItemTypeName());
+ //number of primary keys here depends on the file input, 3 for rcfiles and 2 for text and sequence files.
+ numPrimaryKeys = DatasetUtils.getExternalRIDSize(dataset);
+ itemType = createExternalItemTypeWithRID(externalItemType);
+ payloadSerde = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(itemType);
+ numSecondaryKeys = createIndexStmt.getKeyFields().size();
+ //splits and constraints <--They don't exist-->
+ primaryFileSplitProvider = null;
+ primaryPartitionConstraint = null;
+ //create secondary split and constraints
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> secondarySplitsAndConstraint = metadataProvider
+ .splitProviderAndPartitionConstraintsForExternalDataset(dataverseName, datasetName,
+ secondaryIndexName);
+ secondaryFileSplitProvider = secondarySplitsAndConstraint.first;
+ secondaryPartitionConstraint = secondarySplitsAndConstraint.second;
+ // Must be called in this order.
+ setExternalRIDDescAndComparators();
+ setExternalSecondaryRecDescAndComparators(createIndexStmt, metadataProvider);
+ numElementsHint = metadataProvider.getCardinalityPerPartitionHint(dataset);
+ }
+ else
+ {
+ itemType = (ARecordType) metadataProvider.findType(dataset.getDataverseName(), dataset.getItemTypeName());
+ payloadSerde = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(itemType);
+ numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
+ numSecondaryKeys = createIndexStmt.getKeyFields().size();
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> primarySplitsAndConstraint = metadataProvider
+ .splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataverseName, datasetName, datasetName);
+ primaryFileSplitProvider = primarySplitsAndConstraint.first;
+ primaryPartitionConstraint = primarySplitsAndConstraint.second;
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> secondarySplitsAndConstraint = metadataProvider
+ .splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataverseName, datasetName,
+ secondaryIndexName);
+ secondaryFileSplitProvider = secondarySplitsAndConstraint.first;
+ secondaryPartitionConstraint = secondarySplitsAndConstraint.second;
+ // Must be called in this order.
+ setPrimaryRecDescAndComparators();
+ setSecondaryRecDescAndComparators(createIndexStmt, metadataProvider);
+ numElementsHint = metadataProvider.getCardinalityPerPartitionHint(dataset);
+ }
}
- if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
- throw new AsterixException("Cannot index an external dataset (" + datasetName + ").");
- }
- itemType = (ARecordType) metadataProvider.findType(dataset.getDataverseName(), dataset.getItemTypeName());
- payloadSerde = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(itemType);
- numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
- numSecondaryKeys = createIndexStmt.getKeyFields().size();
- Pair<IFileSplitProvider, AlgebricksPartitionConstraint> primarySplitsAndConstraint = metadataProvider
- .splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataverseName, datasetName, datasetName);
- primaryFileSplitProvider = primarySplitsAndConstraint.first;
- primaryPartitionConstraint = primarySplitsAndConstraint.second;
- Pair<IFileSplitProvider, AlgebricksPartitionConstraint> secondarySplitsAndConstraint = metadataProvider
- .splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataverseName, datasetName,
- secondaryIndexName);
- secondaryFileSplitProvider = secondarySplitsAndConstraint.first;
- secondaryPartitionConstraint = secondarySplitsAndConstraint.second;
- // Must be called in this order.
- setPrimaryRecDescAndComparators();
- setSecondaryRecDescAndComparators(createIndexStmt, metadataProvider);
- numElementsHint = metadataProvider.getCardinalityPerPartitionHint(dataset);
- }
protected void setPrimaryRecDescAndComparators() throws AlgebricksException {
List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
@@ -273,6 +310,210 @@
return keyProviderOp;
}
+protected ARecordType createExternalItemTypeWithRID(
+ ARecordType externalItemType) throws AsterixException {
+
+ String[] fieldsNames = new String[externalItemType.getFieldNames().length+numPrimaryKeys];
+ IAType[] fieldsTypes = new IAType[externalItemType.getFieldTypes().length+numPrimaryKeys];
+
+ //add RID fields names and types
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ fieldsNames[0] = "_file-number";
+ fieldsTypes[0] = BuiltinType.AINT32;
+ }
+ else
+ {
+ fieldsNames[0] = "_file-name";
+ fieldsTypes[0] = BuiltinType.ASTRING;
+ }
+ fieldsNames[1] = "_byte-location";
+ fieldsTypes[1] = BuiltinType.AINT64;
+ if(numPrimaryKeys == 3)
+ {
+ //add the row number for rc files
+ fieldsNames[2] = "_row-number";
+ fieldsTypes[2] = BuiltinType.AINT32;
+ }
+
+ //add the original fields names and types
+ for(int i=0; i < externalItemType.getFieldNames().length; i++)
+ {
+ fieldsNames[i+numPrimaryKeys] = externalItemType.getFieldNames()[i];
+ fieldsTypes[i+numPrimaryKeys] = externalItemType.getFieldTypes()[i];
+ }
+ return new ARecordType(externalItemType.getTypeName(), fieldsNames, fieldsTypes, externalItemType.isOpen());
+ }
+
+ protected void setExternalRIDDescAndComparators() throws AlgebricksException {
+
+ ISerializerDeserializer[] externalRecFields = new ISerializerDeserializer[itemType.getFieldNames().length];
+ ITypeTraits[] externalTypeTraits = new ITypeTraits[itemType.getFieldNames().length];
+
+ primaryComparatorFactories = new IBinaryComparatorFactory[numPrimaryKeys];
+ primaryBloomFilterKeyFields = new int[numPrimaryKeys];
+ ISerializerDeserializerProvider serdeProvider = metadataProvider.getFormat().getSerdeProvider();
+
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ primaryComparatorFactories[0] = AqlBinaryComparatorFactoryProvider.INSTANCE.getBinaryComparatorFactory(BuiltinType.AINT32, true);
+ }
+ else
+ {
+ primaryComparatorFactories[0] = AqlBinaryComparatorFactoryProvider.INSTANCE.getBinaryComparatorFactory(BuiltinType.ASTRING, true);
+ }
+ primaryComparatorFactories[1] = AqlBinaryComparatorFactoryProvider.INSTANCE.getBinaryComparatorFactory(BuiltinType.AINT64, true);
+
+ primaryBloomFilterKeyFields[0]=0;
+ primaryBloomFilterKeyFields[1]=1;
+
+ if(numPrimaryKeys == 3)
+ {
+ primaryComparatorFactories[2] = AqlBinaryComparatorFactoryProvider.INSTANCE.getBinaryComparatorFactory(BuiltinType.AINT32, true);
+ primaryBloomFilterKeyFields[2]=2;
+ }
+
+ for(int i=0; i < itemType.getFieldNames().length; i++)
+ {
+ externalRecFields[i] = serdeProvider.getSerializerDeserializer(itemType.getFieldTypes()[i]);
+ externalTypeTraits[i] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(itemType.getFieldTypes()[i]);
+ }
+ primaryRecDesc = new RecordDescriptor(externalRecFields, externalTypeTraits);
+ }
+
+protected void setExternalSecondaryRecDescAndComparators(CompiledCreateIndexStatement createIndexStmt,
+ AqlMetadataProvider metadataProvider) throws AlgebricksException, AsterixException {
+ secondaryKeyFields = createIndexStmt.getKeyFields();
+ secondaryFieldAccessEvalFactories = new ICopyEvaluatorFactory[numSecondaryKeys+ numPrimaryKeys];
+ secondaryComparatorFactories = new IBinaryComparatorFactory[numSecondaryKeys + numPrimaryKeys];
+ secondaryBloomFilterKeyFields = new int[numSecondaryKeys];
+ ISerializerDeserializer[] secondaryRecFields = new ISerializerDeserializer[numPrimaryKeys + numSecondaryKeys];
+ ITypeTraits[] secondaryTypeTraits = new ITypeTraits[numSecondaryKeys + numPrimaryKeys];
+ ISerializerDeserializerProvider serdeProvider = metadataProvider.getFormat().getSerdeProvider();
+ ITypeTraitProvider typeTraitProvider = metadataProvider.getFormat().getTypeTraitProvider();
+ IBinaryComparatorFactoryProvider comparatorFactoryProvider = metadataProvider.getFormat()
+ .getBinaryComparatorFactoryProvider();
+
+ for (int i = 0; i < numSecondaryKeys; i++) {
+ secondaryFieldAccessEvalFactories[i] = metadataProvider.getFormat().getFieldAccessEvaluatorFactory(
+ itemType, secondaryKeyFields.get(i), 0);
+ Pair<IAType, Boolean> keyTypePair = Index.getNonNullableKeyFieldType(secondaryKeyFields.get(i), itemType);
+ IAType keyType = keyTypePair.first;
+ anySecondaryKeyIsNullable = anySecondaryKeyIsNullable || keyTypePair.second;
+ ISerializerDeserializer keySerde = serdeProvider.getSerializerDeserializer(keyType);
+ secondaryRecFields[i] = keySerde;
+ secondaryComparatorFactories[i] = comparatorFactoryProvider.getBinaryComparatorFactory(keyType, true);
+ secondaryTypeTraits[i] = typeTraitProvider.getTypeTrait(keyType);
+ secondaryBloomFilterKeyFields[i] = i;
+ }
+
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ secondaryFieldAccessEvalFactories[numSecondaryKeys] = metadataProvider.getFormat().getFieldAccessEvaluatorFactory(
+ itemType, "_file-number", 0);
+ }
+ else
+ {
+ secondaryFieldAccessEvalFactories[numSecondaryKeys] = metadataProvider.getFormat().getFieldAccessEvaluatorFactory(
+ itemType, "_file-name", 0);
+ }
+ secondaryFieldAccessEvalFactories[numSecondaryKeys+1] = metadataProvider.getFormat().getFieldAccessEvaluatorFactory(
+ itemType, "_byte-location", 0);
+ if(numPrimaryKeys == 3)
+ {
+ secondaryFieldAccessEvalFactories[numSecondaryKeys+2] = metadataProvider.getFormat().getFieldAccessEvaluatorFactory(
+ itemType, "_row-number", 0);
+ }
+
+ for (int i = 0; i < numPrimaryKeys; i++) {
+ secondaryRecFields[numSecondaryKeys + i] = primaryRecDesc.getFields()[i];
+ secondaryTypeTraits[numSecondaryKeys + i] = primaryRecDesc.getTypeTraits()[i];
+ secondaryComparatorFactories[numSecondaryKeys + i] = primaryComparatorFactories[i];
+ }
+ secondaryRecDesc = new RecordDescriptor(secondaryRecFields, secondaryTypeTraits);
+ }
+
+protected Pair<ExternalDataIndexingOperatorDescriptor, AlgebricksPartitionConstraint> createExternalIndexingOp(JobSpecification spec) throws Exception {
+ Pair<ExternalDataIndexingOperatorDescriptor,AlgebricksPartitionConstraint> indexingOpAndConstraints = metadataProvider.buildExternalDataIndexingRuntime(spec, itemType, dataset, NonTaggedDataFormat.INSTANCE);
+ AlgebricksPartitionConstraintHelper.setPartitionConstraintInJobSpec(spec, indexingOpAndConstraints.first,
+ indexingOpAndConstraints.second);
+ return indexingOpAndConstraints;
+ }
+
+protected AlgebricksMetaOperatorDescriptor createExternalAssignOp(JobSpecification spec) throws AlgebricksException {
+ int[] outColumns = new int[numSecondaryKeys + numPrimaryKeys];
+ int[] projectionList = new int[numSecondaryKeys + numPrimaryKeys];
+ for (int i = 0; i < numSecondaryKeys + numPrimaryKeys; i++) {
+ outColumns[i] = i;
+ projectionList[i] = i;
+ }
+
+ IScalarEvaluatorFactory[] sefs = new IScalarEvaluatorFactory[secondaryFieldAccessEvalFactories.length];
+ for (int i = 0; i < secondaryFieldAccessEvalFactories.length; ++i) {
+ sefs[i] = new LogicalExpressionJobGenToExpressionRuntimeProviderAdapter.ScalarEvaluatorFactoryAdapter(
+ secondaryFieldAccessEvalFactories[i]);
+ }
+ AssignRuntimeFactory assign = new AssignRuntimeFactory(outColumns, sefs, projectionList);
+ AlgebricksMetaOperatorDescriptor asterixAssignOp = new AlgebricksMetaOperatorDescriptor(spec, 1, 1,
+ new IPushRuntimeFactory[] { assign }, new RecordDescriptor[] { secondaryRecDesc });
+ return asterixAssignOp;
+ }
+
+ protected ExternalSortOperatorDescriptor createSortOp(JobSpecification spec,
+ IBinaryComparatorFactory[] secondaryComparatorFactories, RecordDescriptor secondaryRecDesc,
+ AlgebricksPartitionConstraint partitionConstraints) {
+ int[] sortFields = new int[secondaryComparatorFactories.length];
+ for (int i = 0; i < secondaryComparatorFactories.length; i++) {
+ sortFields[i] = i;
+ }
+ ExternalSortOperatorDescriptor sortOp = new ExternalSortOperatorDescriptor(spec,
+ physOptConf.getMaxFramesExternalSort(), sortFields, secondaryComparatorFactories, secondaryRecDesc);
+ AlgebricksPartitionConstraintHelper.setPartitionConstraintInJobSpec(spec, sortOp, partitionConstraints);
+ return sortOp;
+ }
+
+protected ARecordType createSecondaryItemType(ARecordType externalItemType, boolean isRCFile) throws AsterixException
+ {
+
+ String[] fieldsNames = new String[numSecondaryKeys+numPrimaryKeys];
+ IAType[] fieldsTypes = new IAType[numSecondaryKeys+numPrimaryKeys];
+
+ //first create the secondary index fields
+ for(int i=0; i<numSecondaryKeys; i++)
+ {
+ fieldsNames[i] = secondaryKeyFields.get(i);
+ try {
+ fieldsTypes[i] = externalItemType.getFieldType(fieldsNames[i]);
+ } catch (IOException e) {
+ // TODO Auto-generated catch block
+ throw new AsterixException(e);
+ }
+ }
+
+ //second add RID fields (File name or number and byte location)
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ fieldsNames[numSecondaryKeys] = "_file-number";
+ fieldsTypes[numSecondaryKeys] = BuiltinType.ASTRING;
+ }
+ else
+ {
+ fieldsNames[numSecondaryKeys] = "_file-name";
+ fieldsTypes[numSecondaryKeys] = BuiltinType.ASTRING;
+ }
+ fieldsNames[numSecondaryKeys+1] = "_byte-location";
+ fieldsTypes[numSecondaryKeys+1] = BuiltinType.AINT64;
+
+ if(isRCFile)
+ {
+ fieldsNames[numSecondaryKeys+2] = "_row-Number";
+ fieldsTypes[numSecondaryKeys+2] = BuiltinType.AINT32;
+ }
+
+ //return type
+ return new ARecordType(externalItemType.getTypeName(), fieldsNames, fieldsTypes, externalItemType.isOpen());
+ }
+
protected BTreeSearchOperatorDescriptor createPrimaryIndexScanOp(JobSpecification spec) throws AlgebricksException {
// -Infinity
int[] lowKeyFields = null;
@@ -392,3 +633,4 @@
return asterixSelectOp;
}
}
+
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryInvertedIndexCreator.java b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryInvertedIndexCreator.java
index 40e0aa9..6f64aa2 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryInvertedIndexCreator.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryInvertedIndexCreator.java
@@ -296,4 +296,10 @@
storageProperties.getBloomFilterFalsePositiveRate());
}
}
+
+ @Override
+ protected void setExternalSecondaryRecDescAndComparators(CompiledCreateIndexStatement createIndexStmt,
+ AqlMetadataProvider metadataProvider) throws AlgebricksException, AsterixException {
+ throw new AsterixException("Cannot create inverted index on external dataset due to composite RID Fields.");
+ }
}
diff --git a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryRTreeCreator.java b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryRTreeCreator.java
index ec62068..6ce694c 100644
--- a/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryRTreeCreator.java
+++ b/asterix-app/src/main/java/edu/uci/ics/asterix/file/SecondaryRTreeCreator.java
@@ -18,36 +18,47 @@
import edu.uci.ics.asterix.common.api.ILocalResourceMetadata;
import edu.uci.ics.asterix.common.config.AsterixStorageProperties;
+import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.IAsterixPropertiesProvider;
import edu.uci.ics.asterix.common.context.AsterixVirtualBufferCacheProvider;
import edu.uci.ics.asterix.common.exceptions.AsterixException;
import edu.uci.ics.asterix.common.ioopcallbacks.LSMRTreeIOOperationCallbackFactory;
import edu.uci.ics.asterix.dataflow.data.nontagged.valueproviders.AqlPrimitiveValueProviderFactory;
+import edu.uci.ics.asterix.external.adapter.factory.HDFSAdapterFactory;
+import edu.uci.ics.asterix.external.data.operator.ExternalDataIndexingOperatorDescriptor;
+import edu.uci.ics.asterix.external.util.ExternalIndexHashPartitionComputerFactory;
import edu.uci.ics.asterix.formats.nontagged.AqlBinaryComparatorFactoryProvider;
import edu.uci.ics.asterix.formats.nontagged.AqlSerializerDeserializerProvider;
import edu.uci.ics.asterix.formats.nontagged.AqlTypeTraitProvider;
import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
+import edu.uci.ics.asterix.metadata.entities.ExternalDatasetDetails;
import edu.uci.ics.asterix.metadata.entities.Index;
+import edu.uci.ics.asterix.metadata.utils.DatasetUtils;
import edu.uci.ics.asterix.om.types.ATypeTag;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.asterix.om.util.NonTaggedFormatUtil;
+import edu.uci.ics.asterix.runtime.formats.NonTaggedDataFormat;
import edu.uci.ics.asterix.transaction.management.opcallbacks.SecondaryIndexOperationTrackerProvider;
import edu.uci.ics.asterix.transaction.management.resource.LSMRTreeLocalResourceMetadata;
import edu.uci.ics.asterix.transaction.management.resource.PersistentLocalResourceFactoryProvider;
import edu.uci.ics.asterix.transaction.management.service.transaction.AsterixRuntimeComponentsProvider;
import edu.uci.ics.asterix.translator.CompiledStatements.CompiledCreateIndexStatement;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraintHelper;
import edu.uci.ics.hyracks.algebricks.common.exceptions.AlgebricksException;
import edu.uci.ics.hyracks.algebricks.common.utils.Pair;
import edu.uci.ics.hyracks.algebricks.core.jobgen.impl.ConnectorPolicyAssignmentPolicy;
import edu.uci.ics.hyracks.algebricks.core.rewriter.base.PhysicalOptimizationConfig;
import edu.uci.ics.hyracks.algebricks.runtime.operators.meta.AlgebricksMetaOperatorDescriptor;
+import edu.uci.ics.hyracks.api.dataflow.IConnectorDescriptor;
import edu.uci.ics.hyracks.api.dataflow.value.IBinaryComparatorFactory;
+import edu.uci.ics.hyracks.api.dataflow.value.IBinaryHashFunctionFactory;
import edu.uci.ics.hyracks.api.dataflow.value.ISerializerDeserializer;
import edu.uci.ics.hyracks.api.dataflow.value.ITypeTraits;
import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
import edu.uci.ics.hyracks.api.job.JobSpecification;
import edu.uci.ics.hyracks.dataflow.std.base.AbstractOperatorDescriptor;
+import edu.uci.ics.hyracks.dataflow.std.connectors.MToNPartitioningConnectorDescriptor;
import edu.uci.ics.hyracks.dataflow.std.connectors.OneToOneConnectorDescriptor;
import edu.uci.ics.hyracks.storage.am.btree.dataflow.BTreeSearchOperatorDescriptor;
import edu.uci.ics.hyracks.storage.am.btree.impls.BTree;
@@ -150,8 +161,121 @@
}
@Override
+ protected void setExternalSecondaryRecDescAndComparators(CompiledCreateIndexStatement createIndexStmt,
+ AqlMetadataProvider metadataProvider) throws AlgebricksException, AsterixException {
+ secondaryKeyFields = createIndexStmt.getKeyFields();
+ if (numSecondaryKeys != 1) {
+ throw new AsterixException(
+ "Cannot use "
+ + numSecondaryKeys
+ + " fields as a key for the R-tree index. There can be only one field as a key for the R-tree index.");
+ }
+ Pair<IAType, Boolean> spatialTypePair = Index.getNonNullableKeyFieldType(secondaryKeyFields.get(0), itemType);
+ IAType spatialType = spatialTypePair.first;
+ anySecondaryKeyIsNullable = spatialTypePair.second;
+ if (spatialType == null) {
+ throw new AsterixException("Could not find field " + secondaryKeyFields.get(0) + " in the schema.");
+ }
+ int numDimensions = NonTaggedFormatUtil.getNumDimensions(spatialType.getTypeTag());
+ numNestedSecondaryKeyFields = numDimensions * 2;
+ secondaryFieldAccessEvalFactories = metadataProvider.getFormat().createMBRFactory(itemType, secondaryKeyFields.get(0),
+ numPrimaryKeys, numDimensions);
+ secondaryComparatorFactories = new IBinaryComparatorFactory[numNestedSecondaryKeyFields];
+ valueProviderFactories = new IPrimitiveValueProviderFactory[numNestedSecondaryKeyFields];
+ ISerializerDeserializer[] secondaryRecFields = new ISerializerDeserializer[numPrimaryKeys
+ + numNestedSecondaryKeyFields];
+ ITypeTraits[] secondaryTypeTraits = new ITypeTraits[numNestedSecondaryKeyFields + numPrimaryKeys];
+ IAType nestedKeyType = NonTaggedFormatUtil.getNestedSpatialType(spatialType.getTypeTag());
+ keyType = nestedKeyType.getTypeTag();
+ for (int i = 0; i < numNestedSecondaryKeyFields; i++) {
+ ISerializerDeserializer keySerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(nestedKeyType);
+ secondaryRecFields[i] = keySerde;
+ secondaryComparatorFactories[i] = AqlBinaryComparatorFactoryProvider.INSTANCE.getBinaryComparatorFactory(
+ nestedKeyType, true);
+ secondaryTypeTraits[i] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(nestedKeyType);
+ valueProviderFactories[i] = AqlPrimitiveValueProviderFactory.INSTANCE;
+ }
+
+ // Add serializers and comparators for primary index fields.
+ for (int i = 0; i < numPrimaryKeys; i++) {
+ secondaryRecFields[numNestedSecondaryKeyFields + i] = primaryRecDesc.getFields()[i];
+ secondaryTypeTraits[numNestedSecondaryKeyFields + i] = primaryRecDesc.getTypeTraits()[i];
+ }
+ secondaryRecDesc = new RecordDescriptor(secondaryRecFields, secondaryTypeTraits);
+ }
+
+ @Override
public JobSpecification buildLoadingJobSpec() throws AsterixException, AlgebricksException {
JobSpecification spec = JobSpecificationUtils.createJobSpecification();
+ if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
+ Pair<ExternalDataIndexingOperatorDescriptor, AlgebricksPartitionConstraint> RIDScanOpAndConstraints;
+ AlgebricksMetaOperatorDescriptor asterixAssignOp;
+ try
+ {
+ //create external indexing scan operator
+ RIDScanOpAndConstraints = createExternalIndexingOp(spec);
+ //create assign operator
+ asterixAssignOp = createExternalAssignOp(spec);
+ AlgebricksPartitionConstraintHelper.setPartitionConstraintInJobSpec(spec, asterixAssignOp,
+ RIDScanOpAndConstraints.second);
+ }
+ catch(Exception e)
+ {
+ throw new AsterixException("Failed to create external index scanning and loading job");
+ }
+
+ // If any of the secondary fields are nullable, then add a select op that filters nulls.
+ AlgebricksMetaOperatorDescriptor selectOp = null;
+ if (anySecondaryKeyIsNullable) {
+ selectOp = createFilterNullsSelectOp(spec, numSecondaryKeys);
+ }
+
+ // Create secondary RTree bulk load op.
+ AsterixStorageProperties storageProperties = propertiesProvider.getStorageProperties();
+ TreeIndexBulkLoadOperatorDescriptor secondaryBulkLoadOp = createTreeIndexBulkLoadOp(
+ spec,
+ numNestedSecondaryKeyFields,
+ new LSMRTreeDataflowHelperFactory(valueProviderFactories, RTreePolicyType.RTREE,
+ primaryComparatorFactories, new AsterixVirtualBufferCacheProvider(dataset.getDatasetId()),
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, new SecondaryIndexOperationTrackerProvider(
+ LSMRTreeIOOperationCallbackFactory.INSTANCE, dataset.getDatasetId()),
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER,
+ AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER, AqlMetadataProvider.proposeLinearizer(
+ keyType, secondaryComparatorFactories.length), storageProperties
+ .getBloomFilterFalsePositiveRate()), BTree.DEFAULT_FILL_FACTOR);
+ // Connect the operators.
+ // Create a hash partitioning connector
+ ExternalDatasetDetails edsd = (ExternalDatasetDetails)dataset.getDatasetDetails();
+ IBinaryHashFunctionFactory[] hashFactories = null;
+ if(edsd.getProperties().get(HDFSAdapterFactory.KEY_INPUT_FORMAT).trim().equals(HDFSAdapterFactory.INPUT_FORMAT_RC))
+ {
+ hashFactories = DatasetUtils.computeExternalDataKeysBinaryHashFunFactories(dataset, NonTaggedDataFormat.INSTANCE.getBinaryHashFunctionFactoryProvider());
+ }
+ else
+ {
+ hashFactories = DatasetUtils.computeExternalDataKeysBinaryHashFunFactories(dataset, NonTaggedDataFormat.INSTANCE.getBinaryHashFunctionFactoryProvider());
+ }
+ //select partitioning keys (always the first 2 after secondary keys)
+ int[] keys = new int[2];
+ keys[0] = numSecondaryKeys;
+ keys[1] = numSecondaryKeys + 1;
+
+ IConnectorDescriptor hashConn = new MToNPartitioningConnectorDescriptor(spec,
+ new ExternalIndexHashPartitionComputerFactory(keys, hashFactories));
+ spec.connect(new OneToOneConnectorDescriptor(spec), RIDScanOpAndConstraints.first, 0, asterixAssignOp, 0);
+ if (anySecondaryKeyIsNullable) {
+ spec.connect(new OneToOneConnectorDescriptor(spec), asterixAssignOp, 0, selectOp, 0);
+ spec.connect(hashConn, selectOp, 0, secondaryBulkLoadOp, 0);
+ } else {
+ spec.connect(hashConn, asterixAssignOp, 0, secondaryBulkLoadOp, 0);
+ }
+ spec.addRoot(secondaryBulkLoadOp);
+ spec.setConnectorPolicyAssignmentPolicy(new ConnectorPolicyAssignmentPolicy());
+ return spec;
+ }
+ else
+ {
// Create dummy key provider for feeding the primary index scan.
AbstractOperatorDescriptor keyProviderOp = createDummyKeyProviderOp(spec);
@@ -195,5 +319,7 @@
spec.addRoot(secondaryBulkLoadOp);
spec.setConnectorPolicyAssignmentPolicy(new ConnectorPolicyAssignmentPolicy());
return spec;
+ }
}
}
+
diff --git a/asterix-app/src/main/resources/asterix-build-configuration.xml b/asterix-app/src/main/resources/asterix-build-configuration.xml
index 6a6332d..9c4d15b 100644
--- a/asterix-app/src/main/resources/asterix-build-configuration.xml
+++ b/asterix-app/src/main/resources/asterix-build-configuration.xml
@@ -1,50 +1,40 @@
-<!--
- ! Copyright 2009-2013 by The Regents of the University of California
- ! Licensed 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 from
- !
- ! 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.
- !-->
+<!-- ! Copyright 2009-2013 by The Regents of the University of California
+ ! Licensed 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 from ! ! 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. ! -->
<asterixConfiguration xmlns="asterixconf">
- <metadataNode>nc1</metadataNode>
- <store>
- <ncId>nc1</ncId>
- <storeDirs>nc1data</storeDirs>
- </store>
- <store>
- <ncId>nc2</ncId>
- <storeDirs>nc2data</storeDirs>
- </store>
- <transactionLogDir>
- <ncId>nc1</ncId>
- <txnLogDirPath>target/txnLogDir/nc1</txnLogDirPath>
- </transactionLogDir>
- <transactionLogDir>
- <ncId>nc2</ncId>
- <txnLogDirPath>target/txnLogDir/nc2</txnLogDirPath>
- </transactionLogDir>
- <property>
- <name>log.level</name>
- <value>WARNING</value>
- <description>Log level for running tests/build</description>
- </property>
- <property>
+ <metadataNode>nc1</metadataNode>
+ <store>
+ <ncId>nc1</ncId>
+ <storeDirs>nc1data</storeDirs>
+ </store>
+ <store>
+ <ncId>nc2</ncId>
+ <storeDirs>nc2data</storeDirs>
+ </store>
+ <transactionLogDir>
+ <ncId>nc1</ncId>
+ <txnLogDirPath>target/txnLogDir/nc1</txnLogDirPath>
+ </transactionLogDir>
+ <transactionLogDir>
+ <ncId>nc2</ncId>
+ <txnLogDirPath>target/txnLogDir/nc2</txnLogDirPath>
+ </transactionLogDir>
+ <property>
+ <name>log.level</name>
+ <value>WARNING</value>
+ <description>Log level for running tests/build</description>
+ </property>
+ <property>
<name>storage.memorycomponent.numpages</name>
<value>8</value>
<description>The number of pages to allocate for a memory component.
(Default = 8)
</description>
</property>
- <property>
- <name>txn.log.groupcommitinterval</name>
- <value>1</value>
- <description>The group commit wait time in milliseconds.</description>
- </property>
</asterixConfiguration>
diff --git a/asterix-aql/src/main/java/edu/uci/ics/asterix/aql/expression/ExternalDetailsDecl.java b/asterix-aql/src/main/java/edu/uci/ics/asterix/aql/expression/ExternalDetailsDecl.java
index 4233225..2d97f0a 100644
--- a/asterix-aql/src/main/java/edu/uci/ics/asterix/aql/expression/ExternalDetailsDecl.java
+++ b/asterix-aql/src/main/java/edu/uci/ics/asterix/aql/expression/ExternalDetailsDecl.java
@@ -16,10 +16,20 @@
import java.util.Map;
+import edu.uci.ics.asterix.metadata.bootstrap.MetadataConstants;
+
public class ExternalDetailsDecl implements IDatasetDetailsDecl {
private Map<String, String> properties;
private String adapter;
+ private Identifier nodegroupName;
+ public ExternalDetailsDecl(Map<String, String> properties, String adapter, Identifier nodegroupName) {
+ this.properties = properties;
+ this.adapter = adapter;
+ this.nodegroupName = nodegroupName == null ? new Identifier(MetadataConstants.METADATA_DEFAULT_NODEGROUP_NAME)
+ : nodegroupName;
+ }
+
public void setAdapter(String adapter) {
this.adapter = adapter;
}
@@ -35,4 +45,12 @@
public Map<String, String> getProperties() {
return properties;
}
+
+ public void setNodegroupName(Identifier nodegroupName) {
+ this.nodegroupName = nodegroupName;
+ }
+
+ public Identifier getNodegroupName() {
+ return nodegroupName;
+ }
}
diff --git a/asterix-aql/src/main/javacc/AQL.jj b/asterix-aql/src/main/javacc/AQL.jj
index cb6336b..8dd7bfb 100644
--- a/asterix-aql/src/main/javacc/AQL.jj
+++ b/asterix-aql/src/main/javacc/AQL.jj
@@ -308,11 +308,11 @@
<LEFTPAREN> typeName = Identifier() <RIGHTPAREN>
ifNotExists = IfNotExists()
"using" adapterName = AdapterName() properties = Configuration()
+ ("on" nodeGroupName = Identifier() )?
( "hints" hints = Properties() )?
{
- ExternalDetailsDecl edd = new ExternalDetailsDecl();
- edd.setAdapter(adapterName);
- edd.setProperties(properties);
+ ExternalDetailsDecl edd = new ExternalDetailsDecl(properties,
+ adapterName,nodeGroupName != null? new Identifier(nodeGroupName): null);
dsetDecl = new DatasetDecl(nameComponents.first,
nameComponents.second,
new Identifier(typeName),
diff --git a/asterix-common/src/main/java/edu/uci/ics/asterix/common/api/AsterixThreadExecutor.java b/asterix-common/src/main/java/edu/uci/ics/asterix/common/api/AsterixThreadExecutor.java
index 14975ff..edd4b2a 100644
--- a/asterix-common/src/main/java/edu/uci/ics/asterix/common/api/AsterixThreadExecutor.java
+++ b/asterix-common/src/main/java/edu/uci/ics/asterix/common/api/AsterixThreadExecutor.java
@@ -14,12 +14,15 @@
*/
package edu.uci.ics.asterix.common.api;
+import java.util.concurrent.Callable;
import java.util.concurrent.Executor;
+import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
public class AsterixThreadExecutor implements Executor {
public final static AsterixThreadExecutor INSTANCE = new AsterixThreadExecutor();
- private final Executor executor = Executors.newCachedThreadPool(AsterixThreadFactory.INSTANCE);
+ private final ExecutorService executorService = Executors.newCachedThreadPool(AsterixThreadFactory.INSTANCE);
private AsterixThreadExecutor() {
@@ -27,6 +30,10 @@
@Override
public void execute(Runnable command) {
- executor.execute(command);
+ executorService.execute(command);
+ }
+
+ public Future<Object> submit(Callable command) {
+ return (Future<Object>) executorService.submit(command);
}
}
diff --git a/asterix-common/src/main/java/edu/uci/ics/asterix/common/config/AsterixTransactionProperties.java b/asterix-common/src/main/java/edu/uci/ics/asterix/common/config/AsterixTransactionProperties.java
index 5a40ece..a1dd52a 100644
--- a/asterix-common/src/main/java/edu/uci/ics/asterix/common/config/AsterixTransactionProperties.java
+++ b/asterix-common/src/main/java/edu/uci/ics/asterix/common/config/AsterixTransactionProperties.java
@@ -25,12 +25,6 @@
private static final String TXN_LOG_PARTITIONSIZE_KEY = "txn.log.partitionsize";
private static final long TXN_LOG_PARTITIONSIZE_DEFAULT = ((long)2 << 30); // 2GB
- private static final String TXN_LOG_DISKSECTORSIZE_KEY = "txn.log.disksectorsize";
- private static final int TXN_LOG_DISKSECTORSIZE_DEFAULT = 4096;
-
- private static final String TXN_LOG_GROUPCOMMITINTERVAL_KEY = "txn.log.groupcommitinterval";
- private static int TXN_LOG_GROUPCOMMITINTERVAL_DEFAULT = 10; // 0.1ms
-
private static final String TXN_LOG_CHECKPOINT_LSNTHRESHOLD_KEY = "txn.log.checkpoint.lsnthreshold";
private static final int TXN_LOG_CHECKPOINT_LSNTHRESHOLD_DEFAULT = (64 << 20); // 64M
@@ -75,16 +69,6 @@
PropertyInterpreters.getLongPropertyInterpreter());
}
- public int getLogDiskSectorSize() {
- return accessor.getProperty(TXN_LOG_DISKSECTORSIZE_KEY, TXN_LOG_DISKSECTORSIZE_DEFAULT,
- PropertyInterpreters.getIntegerPropertyInterpreter());
- }
-
- public int getGroupCommitInterval() {
- return accessor.getProperty(TXN_LOG_GROUPCOMMITINTERVAL_KEY, TXN_LOG_GROUPCOMMITINTERVAL_DEFAULT,
- PropertyInterpreters.getIntegerPropertyInterpreter());
- }
-
public int getCheckpointLSNThreshold() {
return accessor.getProperty(TXN_LOG_CHECKPOINT_LSNTHRESHOLD_KEY, TXN_LOG_CHECKPOINT_LSNTHRESHOLD_DEFAULT,
PropertyInterpreters.getIntegerPropertyInterpreter());
diff --git a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILockManager.java b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILockManager.java
index 54c86af..a752afa 100644
--- a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILockManager.java
+++ b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILockManager.java
@@ -70,17 +70,6 @@
throws ACIDException;
/**
- * @param datasetId
- * @param entityHashValue
- * @param txnContext
- * @throws ACIDException
- * TODO
- * @return
- */
- public void unlock(DatasetId datasetId, int entityHashValue, ITransactionContext txnContext, boolean commitFlag)
- throws ACIDException;
-
- /**
* Call to lock and unlock a specific resource in a specific lock mode
*
* @param datasetId
diff --git a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILogRecord.java b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILogRecord.java
index d810ebd..d13ef6c 100644
--- a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILogRecord.java
+++ b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/ILogRecord.java
@@ -20,16 +20,18 @@
public interface ILogRecord {
- public static final int COMMIT_LOG_SIZE = 21;
- public static final int UPDATE_LOG_BASE_SIZE = 56;
+ public static final int JOB_COMMIT_LOG_SIZE = 13;
+ public static final int ENTITY_COMMIT_LOG_BASE_SIZE = 29;
+ public static final int UPDATE_LOG_BASE_SIZE = 64;
public boolean readLogRecord(ByteBuffer buffer);
public void writeLogRecord(ByteBuffer buffer);
-
- public void formCommitLogRecord(ITransactionContext txnCtx, byte logType, int jobId, int datasetId, int PKHashValue);
- public void setUpdateLogSize();
+ public void formJobCommitLogRecord(ITransactionContext txnCtx);
+
+ public void formEntityCommitLogRecord(ITransactionContext txnCtx, int datasetId, int PKHashValue,
+ ITupleReference tupleReference, int[] primaryKeyFields);
public ITransactionContext getTxnCtx();
@@ -98,11 +100,23 @@
public long getChecksum();
public void setChecksum(long checksum);
-
+
public long getLSN();
public void setLSN(long LSN);
public String getLogRecordForDisplay();
+ public void computeAndSetLogSize();
+
+ public int getPKValueSize();
+
+ public ITupleReference getPKValue();
+
+ public void setPKFields(int[] primaryKeyFields);
+
+ public void computeAndSetPKValueSize();
+
+ public void setPKValue(ITupleReference PKValue);
+
}
diff --git a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/LogManagerProperties.java b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/LogManagerProperties.java
index 591d9b1..dd1e7b4 100644
--- a/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/LogManagerProperties.java
+++ b/asterix-common/src/main/java/edu/uci/ics/asterix/common/transactions/LogManagerProperties.java
@@ -36,14 +36,10 @@
private final int logPageSize;
// number of log pages in the log buffer.
private final int numLogPages;
- // time in milliseconds
- private final long groupCommitWaitPeriod;
// logBufferSize = logPageSize * numLogPages;
private final int logBufferSize;
// maximum size of each log file
private final long logPartitionSize;
- // default disk sector size
- private final int diskSectorSize;
public LogManagerProperties(AsterixTransactionProperties txnProperties, String nodeId) {
this.logDirKey = new String(nodeId + LOG_DIR_SUFFIX);
@@ -52,12 +48,9 @@
long logPartitionSize = txnProperties.getLogPartitionSize();
this.logDir = txnProperties.getLogDirectory(nodeId);
this.logFilePrefix = DEFAULT_LOG_FILE_PREFIX;
- this.groupCommitWaitPeriod = txnProperties.getGroupCommitInterval();
-
this.logBufferSize = logPageSize * numLogPages;
//make sure that the log partition size is the multiple of log buffer size.
this.logPartitionSize = (logPartitionSize / logBufferSize) * logBufferSize;
- this.diskSectorSize = txnProperties.getLogDiskSectorSize();
}
public long getLogPartitionSize() {
@@ -84,18 +77,10 @@
return logBufferSize;
}
- public long getGroupCommitWaitPeriod() {
- return groupCommitWaitPeriod;
- }
-
public String getLogDirKey() {
return logDirKey;
}
- public int getDiskSectorSize() {
- return diskSectorSize;
- }
-
public String toString() {
StringBuilder builder = new StringBuilder();
builder.append("log_dir_ : " + logDir + lineSeparator);
@@ -103,7 +88,6 @@
builder.append("log_page_size : " + logPageSize + lineSeparator);
builder.append("num_log_pages : " + numLogPages + lineSeparator);
builder.append("log_partition_size : " + logPartitionSize + lineSeparator);
- builder.append("group_commit_wait_period : " + groupCommitWaitPeriod + lineSeparator);
return builder.toString();
}
}
diff --git a/asterix-external-data/pom.xml b/asterix-external-data/pom.xml
index f8d5ea2..7966b52 100644
--- a/asterix-external-data/pom.xml
+++ b/asterix-external-data/pom.xml
@@ -153,6 +153,11 @@
<artifactId>jdom</artifactId>
<version>1.0</version>
</dependency>
+ <dependency>
+ <groupId>org.apache.hive</groupId>
+ <artifactId>hive-exec</artifactId>
+ <version>0.11.0</version>
+ </dependency>
</dependencies>
</project>
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HDFSAdapterFactory.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HDFSAdapterFactory.java
index 4ca3d72..e9d2175 100644
--- a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HDFSAdapterFactory.java
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HDFSAdapterFactory.java
@@ -20,12 +20,21 @@
import org.apache.hadoop.mapred.InputSplit;
import org.apache.hadoop.mapred.JobConf;
-
+import org.apache.hadoop.conf.Configuration;
+import edu.uci.ics.asterix.external.dataset.adapter.HDFSAccessByRIDAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.HDFSAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.HDFSIndexingAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
+import edu.uci.ics.asterix.formats.nontagged.AqlSerializerDeserializerProvider;
+import edu.uci.ics.asterix.formats.nontagged.AqlTypeTraitProvider;
+import edu.uci.ics.asterix.om.types.BuiltinType;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksAbsolutePartitionConstraint;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.api.dataflow.value.ISerializerDeserializer;
+import edu.uci.ics.hyracks.api.dataflow.value.ITypeTraits;
+import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
import edu.uci.ics.hyracks.hdfs.dataflow.ConfFactory;
import edu.uci.ics.hyracks.hdfs.dataflow.InputSplitsFactory;
import edu.uci.ics.hyracks.hdfs.scheduler.Scheduler;
@@ -35,76 +44,190 @@
*/
@SuppressWarnings("deprecation")
public class HDFSAdapterFactory implements IGenericDatasetAdapterFactory {
- private static final long serialVersionUID = 1L;
+ private static final long serialVersionUID = 1L;
- public static final String HDFS_ADAPTER_NAME = "hdfs";
- public static final String CLUSTER_LOCATIONS = "cluster-locations";
- public static transient String SCHEDULER = "hdfs-scheduler";
+ public static final String HDFS_ADAPTER_NAME = "hdfs";
+ public static final String CLUSTER_LOCATIONS = "cluster-locations";
+ public static transient String SCHEDULER = "hdfs-scheduler";
- public static final String KEY_HDFS_URL = "hdfs";
- public static final String KEY_PATH = "path";
- public static final String KEY_INPUT_FORMAT = "input-format";
- public static final String INPUT_FORMAT_TEXT = "text-input-format";
- public static final String INPUT_FORMAT_SEQUENCE = "sequence-input-format";
+ public static final String KEY_HDFS_URL = "hdfs";
+ public static final String KEY_PATH = "path";
+ public static final String KEY_INPUT_FORMAT = "input-format";
+ public static final String INPUT_FORMAT_TEXT = "text-input-format";
+ public static final String INPUT_FORMAT_SEQUENCE = "sequence-input-format";
+ public static final String INPUT_FORMAT_RC = "rc-input-format";
+ public static final String KEY_DELIMITER = "delimiter";
+ public static final String KEY_FORMAT = "format";
+ public static final String FORMAT_DELIMITED_TEXT = "delimited-text";
- private transient AlgebricksPartitionConstraint clusterLocations;
- private String[] readSchedule;
- private boolean executed[];
- private InputSplitsFactory inputSplitsFactory;
- private ConfFactory confFactory;
- private boolean setup = false;
+ private transient AlgebricksPartitionConstraint clusterLocations;
+ private String[] readSchedule;
+ private boolean executed[];
+ private InputSplitsFactory inputSplitsFactory;
+ private ConfFactory confFactory;
+ private boolean setup = false;
- private static final Map<String, String> formatClassNames = initInputFormatMap();
+ private static final Map<String, String> formatClassNames = initInputFormatMap();
- private static Map<String, String> initInputFormatMap() {
- Map<String, String> formatClassNames = new HashMap<String, String>();
- formatClassNames.put(INPUT_FORMAT_TEXT, "org.apache.hadoop.mapred.TextInputFormat");
- formatClassNames.put(INPUT_FORMAT_SEQUENCE, "org.apache.hadoop.mapred.SequenceFileInputFormat");
- return formatClassNames;
- }
+ private static Map<String, String> initInputFormatMap() {
+ Map<String, String> formatClassNames = new HashMap<String, String>();
+ formatClassNames.put(INPUT_FORMAT_TEXT, "org.apache.hadoop.mapred.TextInputFormat");
+ formatClassNames.put(INPUT_FORMAT_SEQUENCE, "org.apache.hadoop.mapred.SequenceFileInputFormat");
+ formatClassNames.put(INPUT_FORMAT_RC, "org.apache.hadoop.hive.ql.io.RCFileInputFormat");
+ return formatClassNames;
+ }
- @Override
- public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
- if (!setup) {
- /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
- configureJobConf(configuration);
- JobConf conf = configureJobConf(configuration);
- confFactory = new ConfFactory(conf);
+ @Override
+ public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
+ if (!setup) {
+ /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
+ configureJobConf(configuration);
+ JobConf conf = configureJobConf(configuration);
+ confFactory = new ConfFactory(conf);
- clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
- int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+ int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
- InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
- inputSplitsFactory = new InputSplitsFactory(inputSplits);
+ InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
+ inputSplitsFactory = new InputSplitsFactory(inputSplits);
- Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
- readSchedule = scheduler.getLocationConstraints(inputSplits);
- executed = new boolean[readSchedule.length];
- Arrays.fill(executed, false);
+ Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
+ readSchedule = scheduler.getLocationConstraints(inputSplits);
+ executed = new boolean[readSchedule.length];
+ Arrays.fill(executed, false);
- setup = true;
- }
- JobConf conf = confFactory.getConf();
- InputSplit[] inputSplits = inputSplitsFactory.getSplits();
- HDFSAdapter hdfsAdapter = new HDFSAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations);
- hdfsAdapter.configure(configuration);
- return hdfsAdapter;
- }
+ setup = true;
+ }
+ JobConf conf = confFactory.getConf();
+ InputSplit[] inputSplits = inputSplitsFactory.getSplits();
+ HDFSAdapter hdfsAdapter = new HDFSAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations);
- @Override
- public String getName() {
- return HDFS_ADAPTER_NAME;
- }
+ //If input format is rcfile, configure parser expected format to delimeted text with 0x01 (default ) as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ }
- private JobConf configureJobConf(Map<String, Object> configuration) throws Exception {
- JobConf conf = new JobConf();
- conf.set("fs.default.name", ((String) configuration.get(KEY_HDFS_URL)).trim());
- conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
- conf.setClassLoader(HDFSAdapter.class.getClassLoader());
- conf.set("mapred.input.dir", ((String) configuration.get(KEY_PATH)).trim());
- conf.set("mapred.input.format.class",
- (String) formatClassNames.get(((String) configuration.get(KEY_INPUT_FORMAT)).trim()));
- return conf;
- }
+ hdfsAdapter.configure(configuration);
+ return hdfsAdapter;
+ }
+
+ @Override
+ public IControlledAdapter createAccessByRIDAdapter(
+ Map<String, Object> configuration, IAType atype, HashMap<Integer, String> files) throws Exception {
+ Configuration conf = configureHadoopConnection(configuration);
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+
+ //Create RID record desc
+ RecordDescriptor ridRecordDesc = null;
+
+ //If input format is rcfile, configure parser expected format to delimeted text with control char 0x01 as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ ridRecordDesc = getRIDRecDesc(true, files != null);
+ }
+ else
+ {
+ ridRecordDesc = getRIDRecDesc(false, files != null);
+ }
+ HDFSAccessByRIDAdapter adapter = new HDFSAccessByRIDAdapter(atype, ((String)configuration.get(KEY_INPUT_FORMAT)), clusterLocations,ridRecordDesc, conf, files);
+ adapter.configure(configuration);
+ return adapter;
+ }
+
+ @Override
+ public IDatasourceAdapter createIndexingAdapter(Map<String, Object> configuration, IAType atype, Map<String,Integer> files) throws Exception {
+ if (!setup) {
+ /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
+ configureJobConf(configuration);
+ JobConf conf = configureJobConf(configuration);
+ confFactory = new ConfFactory(conf);
+
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+ int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
+
+ InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
+ inputSplitsFactory = new InputSplitsFactory(inputSplits);
+
+ Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
+ readSchedule = scheduler.getLocationConstraints(inputSplits);
+ executed = new boolean[readSchedule.length];
+ Arrays.fill(executed, false);
+
+ setup = true;
+ }
+ JobConf conf = confFactory.getConf();
+ InputSplit[] inputSplits = inputSplitsFactory.getSplits();
+ //If input format is rcfile, configure parser expected format to delimeted text with 0x01 (default) as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ }
+ HDFSIndexingAdapter hdfsIndexingAdapter = new HDFSIndexingAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations, files);
+ hdfsIndexingAdapter.configure(configuration);
+ return hdfsIndexingAdapter;
+ }
+
+ @Override
+ public String getName() {
+ return HDFS_ADAPTER_NAME;
+ }
+
+ private JobConf configureJobConf(Map<String, Object> configuration) throws Exception {
+ JobConf conf = new JobConf();
+ conf.set("fs.default.name", ((String) configuration.get(KEY_HDFS_URL)).trim());
+ conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
+ conf.setClassLoader(HDFSAdapter.class.getClassLoader());
+ conf.set("mapred.input.dir", ((String) configuration.get(KEY_PATH)).trim());
+ conf.set("mapred.input.format.class",
+ (String) formatClassNames.get(((String) configuration.get(KEY_INPUT_FORMAT)).trim()));
+ return conf;
+ }
+
+ public static Configuration configureHadoopConnection(Map<String, Object> configuration)
+ {
+ Configuration conf = new Configuration();
+ conf.set("fs.default.name", ((String) configuration.get(KEY_HDFS_URL)).trim());
+ conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
+ return conf;
+ }
+
+ public static RecordDescriptor getRIDRecDesc(boolean isRCFile, boolean optimize){
+ int numOfPrimaryKeys = 2;
+ if(isRCFile)
+ {
+ numOfPrimaryKeys++;
+ }
+ @SuppressWarnings("rawtypes")
+ ISerializerDeserializer[] serde = new ISerializerDeserializer[numOfPrimaryKeys];
+ ITypeTraits[] tt = new ITypeTraits[numOfPrimaryKeys];
+ if(optimize)
+ {
+ serde[0] = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(BuiltinType.AINT32);
+ tt[0] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(BuiltinType.AINT32);
+ }
+ else
+ {
+ serde[0] = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(BuiltinType.ASTRING);
+ tt[0] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(BuiltinType.ASTRING);
+ }
+ serde[1] = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(BuiltinType.AINT64);
+ tt[1] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(BuiltinType.AINT64);
+ if(isRCFile)
+ {
+ //we add the row number for rc-files
+ serde[2] = AqlSerializerDeserializerProvider.INSTANCE.getSerializerDeserializer(BuiltinType.AINT32);
+ tt[2] = AqlTypeTraitProvider.INSTANCE.getTypeTrait(BuiltinType.AINT32);
+ }
+ return new RecordDescriptor(serde, tt);
+ }
+
}
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HiveAdapterFactory.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HiveAdapterFactory.java
index 409eb7a..64c8153 100644
--- a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HiveAdapterFactory.java
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/HiveAdapterFactory.java
@@ -18,15 +18,20 @@
import java.util.HashMap;
import java.util.Map;
+import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.mapred.InputSplit;
import org.apache.hadoop.mapred.JobConf;
+import edu.uci.ics.asterix.external.dataset.adapter.HDFSAccessByRIDAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.HDFSAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.HiveAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.HiveIndexingAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksAbsolutePartitionConstraint;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
import edu.uci.ics.hyracks.hdfs.dataflow.ConfFactory;
import edu.uci.ics.hyracks.hdfs.dataflow.InputSplitsFactory;
import edu.uci.ics.hyracks.hdfs.scheduler.Scheduler;
@@ -36,109 +41,185 @@
*/
@SuppressWarnings("deprecation")
public class HiveAdapterFactory implements IGenericDatasetAdapterFactory {
- private static final long serialVersionUID = 1L;
+ private static final long serialVersionUID = 1L;
- public static final String HDFS_ADAPTER_NAME = "hdfs";
- public static final String CLUSTER_LOCATIONS = "cluster-locations";
- public static transient String SCHEDULER = "hdfs-scheduler";
+ public static final String HDFS_ADAPTER_NAME = "hdfs";
+ public static final String CLUSTER_LOCATIONS = "cluster-locations";
+ public static transient String SCHEDULER = "hdfs-scheduler";
- public static final String KEY_HDFS_URL = "hdfs";
- public static final String KEY_PATH = "path";
- public static final String KEY_INPUT_FORMAT = "input-format";
- public static final String INPUT_FORMAT_TEXT = "text-input-format";
- public static final String INPUT_FORMAT_SEQUENCE = "sequence-input-format";
+ public static final String KEY_HDFS_URL = "hdfs";
+ public static final String KEY_PATH = "path";
+ public static final String KEY_INPUT_FORMAT = "input-format";
+ public static final String INPUT_FORMAT_TEXT = "text-input-format";
+ public static final String INPUT_FORMAT_SEQUENCE = "sequence-input-format";
+ public static final String INPUT_FORMAT_RC = "rc-input-format";
- public static final String KEY_FORMAT = "format";
- public static final String KEY_PARSER_FACTORY = "parser";
- public static final String FORMAT_DELIMITED_TEXT = "delimited-text";
- public static final String FORMAT_ADM = "adm";
+ public static final String KEY_FORMAT = "format";
+ public static final String KEY_PARSER_FACTORY = "parser";
+ public static final String FORMAT_DELIMITED_TEXT = "delimited-text";
+ public static final String FORMAT_ADM = "adm";
+ public static final String KEY_DELIMITER = "delimiter";
- public static final String HIVE_DATABASE = "database";
- public static final String HIVE_TABLE = "table";
- public static final String HIVE_HOME = "hive-home";
- public static final String HIVE_METASTORE_URI = "metastore-uri";
- public static final String HIVE_WAREHOUSE_DIR = "warehouse-dir";
- public static final String HIVE_METASTORE_RAWSTORE_IMPL = "rawstore-impl";
+ public static final String HIVE_DATABASE = "database";
+ public static final String HIVE_TABLE = "table";
+ public static final String HIVE_HOME = "hive-home";
+ public static final String HIVE_METASTORE_URI = "metastore-uri";
+ public static final String HIVE_WAREHOUSE_DIR = "warehouse-dir";
+ public static final String HIVE_METASTORE_RAWSTORE_IMPL = "rawstore-impl";
- private String[] readSchedule;
- private boolean executed[];
- private InputSplitsFactory inputSplitsFactory;
- private ConfFactory confFactory;
- private transient AlgebricksPartitionConstraint clusterLocations;
- private boolean setup = false;
+ private String[] readSchedule;
+ private boolean executed[];
+ private InputSplitsFactory inputSplitsFactory;
+ private ConfFactory confFactory;
+ private transient AlgebricksPartitionConstraint clusterLocations;
+ private boolean setup = false;
- private static final Map<String, String> formatClassNames = initInputFormatMap();
+ private static final Map<String, String> formatClassNames = initInputFormatMap();
- private static Map<String, String> initInputFormatMap() {
- Map<String, String> formatClassNames = new HashMap<String, String>();
- formatClassNames.put(INPUT_FORMAT_TEXT, "org.apache.hadoop.mapred.TextInputFormat");
- formatClassNames.put(INPUT_FORMAT_SEQUENCE, "org.apache.hadoop.mapred.SequenceFileInputFormat");
- return formatClassNames;
- }
+ private static Map<String, String> initInputFormatMap() {
+ Map<String, String> formatClassNames = new HashMap<String, String>();
+ formatClassNames.put(INPUT_FORMAT_TEXT, "org.apache.hadoop.mapred.TextInputFormat");
+ formatClassNames.put(INPUT_FORMAT_SEQUENCE, "org.apache.hadoop.mapred.SequenceFileInputFormat");
+ formatClassNames.put(INPUT_FORMAT_RC, "org.apache.hadoop.hive.ql.io.RCFileInputFormat");
+ return formatClassNames;
+ }
- @Override
- public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
- if (!setup) {
- /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
- configureJobConf(configuration);
- JobConf conf = configureJobConf(configuration);
- confFactory = new ConfFactory(conf);
- clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
- int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
+ @Override
+ public IControlledAdapter createAccessByRIDAdapter(Map<String, Object> configuration, IAType atype, HashMap<Integer, String> files) throws Exception {
+ Configuration conf = HDFSAdapterFactory.configureHadoopConnection(configuration);
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+ //Create RID record desc
+ RecordDescriptor ridRecordDesc = null;
- InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
- inputSplitsFactory = new InputSplitsFactory(inputSplits);
+ //If input format is rcfile, configure parser expected format to delimeted text with control char 0x01 as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ ridRecordDesc = HDFSAdapterFactory.getRIDRecDesc(true, files != null);
+ }
+ else
+ {
+ ridRecordDesc = HDFSAdapterFactory.getRIDRecDesc(false, files != null);
+ }
+ HDFSAccessByRIDAdapter adapter = new HDFSAccessByRIDAdapter(atype, ((String)configuration.get(KEY_INPUT_FORMAT)), clusterLocations,ridRecordDesc, conf, files);
+ adapter.configure(configuration);
+ return adapter;
+ }
- Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
- readSchedule = scheduler.getLocationConstraints(inputSplits);
- executed = new boolean[readSchedule.length];
- Arrays.fill(executed, false);
+ @Override
+ public IDatasourceAdapter createIndexingAdapter(
+ Map<String, Object> configuration, IAType atype, Map<String,Integer> files) throws Exception {
+ if (!setup) {
+ /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
+ configureJobConf(configuration);
+ JobConf conf = configureJobConf(configuration);
+ confFactory = new ConfFactory(conf);
- setup = true;
- }
- JobConf conf = confFactory.getConf();
- InputSplit[] inputSplits = inputSplitsFactory.getSplits();
- HiveAdapter hiveAdapter = new HiveAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations);
- hiveAdapter.configure(configuration);
- return hiveAdapter;
- }
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+ int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
- @Override
- public String getName() {
- return "hive";
- }
+ InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
+ inputSplitsFactory = new InputSplitsFactory(inputSplits);
- private JobConf configureJobConf(Map<String, Object> configuration) throws Exception {
- JobConf conf = new JobConf();
+ Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
+ readSchedule = scheduler.getLocationConstraints(inputSplits);
+ executed = new boolean[readSchedule.length];
+ Arrays.fill(executed, false);
- /** configure hive */
- String database = (String) configuration.get(HIVE_DATABASE);
- String tablePath = null;
- if (database == null) {
- tablePath = configuration.get(HIVE_WAREHOUSE_DIR) + "/" + configuration.get(HIVE_TABLE);
- } else {
- tablePath = configuration.get(HIVE_WAREHOUSE_DIR) + "/" + tablePath + ".db" + "/"
- + configuration.get(HIVE_TABLE);
- }
- configuration.put(HDFSAdapter.KEY_PATH, tablePath);
- if (!configuration.get(KEY_FORMAT).equals(FORMAT_DELIMITED_TEXT)) {
- throw new IllegalArgumentException("format" + configuration.get(KEY_FORMAT) + " is not supported");
- }
+ setup = true;
+ }
+ JobConf conf = confFactory.getConf();
+ InputSplit[] inputSplits = inputSplitsFactory.getSplits();
+ HiveIndexingAdapter hiveIndexingAdapter = new HiveIndexingAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations, files);
- if (!(configuration.get(HDFSAdapterFactory.KEY_INPUT_FORMAT).equals(HDFSAdapterFactory.INPUT_FORMAT_TEXT) || configuration
- .get(HDFSAdapterFactory.KEY_INPUT_FORMAT).equals(HDFSAdapterFactory.INPUT_FORMAT_SEQUENCE))) {
- throw new IllegalArgumentException("file input format"
- + configuration.get(HDFSAdapterFactory.KEY_INPUT_FORMAT) + " is not supported");
- }
+ //If input format is rcfile, configure parser expected format to delimeted text with 0x01 (default ) as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ }
- /** configure hdfs */
- conf.set("fs.default.name", ((String) configuration.get(KEY_HDFS_URL)).trim());
- conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
- conf.setClassLoader(HDFSAdapter.class.getClassLoader());
- conf.set("mapred.input.dir", ((String) configuration.get(KEY_PATH)).trim());
- conf.set("mapred.input.format.class",
- (String) formatClassNames.get(((String) configuration.get(KEY_INPUT_FORMAT)).trim()));
- return conf;
- }
-}
+ hiveIndexingAdapter.configure(configuration);
+ return hiveIndexingAdapter;
+ }
+
+ @Override
+ public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
+ if (!setup) {
+ /** set up the factory --serializable stuff --- this if-block should be called only once for each factory instance */
+ configureJobConf(configuration);
+ JobConf conf = configureJobConf(configuration);
+ confFactory = new ConfFactory(conf);
+
+ clusterLocations = (AlgebricksPartitionConstraint) configuration.get(CLUSTER_LOCATIONS);
+ int numPartitions = ((AlgebricksAbsolutePartitionConstraint) clusterLocations).getLocations().length;
+
+ InputSplit[] inputSplits = conf.getInputFormat().getSplits(conf, numPartitions);
+ inputSplitsFactory = new InputSplitsFactory(inputSplits);
+
+ Scheduler scheduler = (Scheduler) configuration.get(SCHEDULER);
+ readSchedule = scheduler.getLocationConstraints(inputSplits);
+ executed = new boolean[readSchedule.length];
+ Arrays.fill(executed, false);
+
+ setup = true;
+ }
+ JobConf conf = confFactory.getConf();
+ InputSplit[] inputSplits = inputSplitsFactory.getSplits();
+ HiveAdapter hiveAdapter = new HiveAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations);
+
+ //If input format is rcfile, configure parser expected format to delimeted text with 0x01 (default ) as delimiter
+ if(((String)configuration.get(KEY_INPUT_FORMAT)).equals(INPUT_FORMAT_RC))
+ {
+ char delimeter = 0x01;
+ configuration.put(KEY_FORMAT, FORMAT_DELIMITED_TEXT);
+ configuration.put(KEY_DELIMITER, Character.toString(delimeter));
+ }
+
+ hiveAdapter.configure(configuration);
+ return hiveAdapter;
+ }
+
+ @Override
+ public String getName() {
+ return "hive";
+ }
+
+ private JobConf configureJobConf(Map<String, Object> configuration) throws Exception {
+ JobConf conf = new JobConf();
+
+ /** configure hive */
+ String database = (String) configuration.get(HIVE_DATABASE);
+ String tablePath = null;
+ if (database == null) {
+ tablePath = configuration.get(HIVE_WAREHOUSE_DIR) + "/" + configuration.get(HIVE_TABLE);
+ } else {
+ tablePath = configuration.get(HIVE_WAREHOUSE_DIR) + "/" + tablePath + ".db" + "/"
+ + configuration.get(HIVE_TABLE);
+ }
+ configuration.put(HDFSAdapter.KEY_PATH, tablePath);
+ if (!configuration.get(KEY_FORMAT).equals(FORMAT_DELIMITED_TEXT)) {
+ throw new IllegalArgumentException("format" + configuration.get(KEY_FORMAT) + " is not supported");
+ }
+
+ if (!(configuration.get(HDFSAdapterFactory.KEY_INPUT_FORMAT).equals(HDFSAdapterFactory.INPUT_FORMAT_TEXT) || configuration
+ .get(HDFSAdapterFactory.KEY_INPUT_FORMAT).equals(HDFSAdapterFactory.INPUT_FORMAT_SEQUENCE) || configuration
+ .get(HDFSAdapterFactory.KEY_INPUT_FORMAT).equals(HDFSAdapterFactory.INPUT_FORMAT_RC))) {
+ throw new IllegalArgumentException("file input format"
+ + configuration.get(HDFSAdapterFactory.KEY_INPUT_FORMAT) + " is not supported");
+ }
+
+ /** configure hdfs */
+ conf.set("fs.default.name", ((String) configuration.get(KEY_HDFS_URL)).trim());
+ conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
+ conf.setClassLoader(HDFSAdapter.class.getClassLoader());
+ conf.set("mapred.input.dir", ((String) configuration.get(KEY_PATH)).trim());
+ conf.set("mapred.input.format.class",
+ (String) formatClassNames.get(((String) configuration.get(KEY_INPUT_FORMAT)).trim()));
+ return conf;
+ }
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/IGenericDatasetAdapterFactory.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/IGenericDatasetAdapterFactory.java
index 0a178a7..f046f88 100644
--- a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/IGenericDatasetAdapterFactory.java
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/IGenericDatasetAdapterFactory.java
@@ -14,8 +14,10 @@
*/
package edu.uci.ics.asterix.external.adapter.factory;
+import java.util.HashMap;
import java.util.Map;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
import edu.uci.ics.asterix.om.types.IAType;
@@ -40,4 +42,29 @@
*/
public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception;
+ /**
+ * Creates an instance of IDatasourceAdapter that is used to read records and their RIDs.
+ *
+ * @param configuration
+ * The configuration parameters for the adapter that is instantiated.
+ * The passed-in configuration is used to configure the created instance of the adapter.
+ * @param atype
+ * The type for the ADM records that are returned by the adapter (contains both original fields and RID fields).
+ * @return An instance of IDatasourceAdapter.
+ * @throws Exception
+ */
+ public IDatasourceAdapter createIndexingAdapter(Map<String, Object> configuration, IAType atype, Map<String,Integer> files) throws Exception;
+
+ /**
+ * Creates an instance of IDatasourceAdapter that is used to read records using their RIDs.
+ *
+ * @param configuration
+ * The configuration parameters for the adapter that is instantiated.
+ * The passed-in configuration is used to configure the created instance of the adapter.
+ * @param atype
+ * The type for the ADM records that are returned by the adapter.
+ * @return An instance of IControlledAdapter.
+ * @throws Exception
+ */
+ public IControlledAdapter createAccessByRIDAdapter(Map<String, Object> configuration, IAType atype, HashMap<Integer, String> files) throws Exception;
}
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/NCFileSystemAdapterFactory.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/NCFileSystemAdapterFactory.java
index e680232..4fae7e7 100644
--- a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/NCFileSystemAdapterFactory.java
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/adapter/factory/NCFileSystemAdapterFactory.java
@@ -14,11 +14,14 @@
*/
package edu.uci.ics.asterix.external.adapter.factory;
+import java.util.HashMap;
import java.util.Map;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.NCFileSystemAdapter;
import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.algebricks.common.exceptions.NotImplementedException;
/**
* Factory class for creating an instance of NCFileSystemAdapter. An
@@ -26,18 +29,31 @@
* an NC.
*/
public class NCFileSystemAdapterFactory implements IGenericDatasetAdapterFactory {
- private static final long serialVersionUID = 1L;
- public static final String NC_FILE_SYSTEM_ADAPTER_NAME = "localfs";
+ private static final long serialVersionUID = 1L;
+ public static final String NC_FILE_SYSTEM_ADAPTER_NAME = "localfs";
- @Override
- public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
- NCFileSystemAdapter fsAdapter = new NCFileSystemAdapter(atype);
- fsAdapter.configure(configuration);
- return fsAdapter;
- }
+ @Override
+ public IDatasourceAdapter createAdapter(Map<String, Object> configuration, IAType atype) throws Exception {
+ NCFileSystemAdapter fsAdapter = new NCFileSystemAdapter(atype);
+ fsAdapter.configure(configuration);
+ return fsAdapter;
+ }
- @Override
- public String getName() {
- return NC_FILE_SYSTEM_ADAPTER_NAME;
- }
+ @Override
+ public String getName() {
+ return NC_FILE_SYSTEM_ADAPTER_NAME;
+ }
+
+ @Override
+ public IDatasourceAdapter createIndexingAdapter(
+ Map<String, Object> configuration, IAType atype, Map<String,Integer> files) throws Exception {
+ throw new NotImplementedException("Indexing Adapter is not implemented for NC FileSystem Data");
+ }
+
+
+
+ @Override
+ public IControlledAdapter createAccessByRIDAdapter(Map<String, Object> configuration, IAType atype, HashMap<Integer, String> files) throws Exception {
+ throw new NotImplementedException("Access by RID Adapter is not implemented for NC FileSystem Data");
+ }
}
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataAccessByRIDOperatorDescriptor.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataAccessByRIDOperatorDescriptor.java
new file mode 100644
index 0000000..aa91a56
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataAccessByRIDOperatorDescriptor.java
@@ -0,0 +1,78 @@
+package edu.uci.ics.asterix.external.data.operator;
+
+import java.nio.ByteBuffer;
+import java.util.HashMap;
+import java.util.Map;
+
+import edu.uci.ics.asterix.external.adapter.factory.IGenericDatasetAdapterFactory;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
+import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.dataflow.IOperatorNodePushable;
+import edu.uci.ics.hyracks.api.dataflow.value.IRecordDescriptorProvider;
+import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.api.job.IOperatorDescriptorRegistry;
+import edu.uci.ics.hyracks.dataflow.std.base.AbstractSingleActivityOperatorDescriptor;
+import edu.uci.ics.hyracks.dataflow.std.base.AbstractUnaryInputUnaryOutputOperatorNodePushable;
+
+public class ExternalDataAccessByRIDOperatorDescriptor extends AbstractSingleActivityOperatorDescriptor {
+
+ /**
+ * This operator is used to access external data residing in hdfs using record ids pushed in frame buffers
+ */
+ private static final long serialVersionUID = 1L;
+ private final Map<String, Object> adapterConfiguration;
+ private final IAType atype;
+ private IGenericDatasetAdapterFactory datasourceAdapterFactory;
+ private IControlledAdapter adapter;
+ private final HashMap<Integer, String> files;
+
+ public ExternalDataAccessByRIDOperatorDescriptor(
+ IOperatorDescriptorRegistry spec, Map<String, Object> arguments, IAType atype,
+ RecordDescriptor outRecDesc,IGenericDatasetAdapterFactory dataSourceAdapterFactory, HashMap<Integer, String> files) {
+ super(spec, 1, 1);
+ this.atype = atype;
+ this.adapterConfiguration = arguments;
+ this.datasourceAdapterFactory = dataSourceAdapterFactory;
+ this.recordDescriptors[0] = outRecDesc;
+ this.files = files;
+ }
+
+ @Override
+ public IOperatorNodePushable createPushRuntime(final IHyracksTaskContext ctx,
+ IRecordDescriptorProvider recordDescProvider, int partition,
+ int nPartitions) throws HyracksDataException {
+ return new AbstractUnaryInputUnaryOutputOperatorNodePushable() {
+ @Override
+ public void open() throws HyracksDataException {
+ //create the access by index adapter
+ try {
+ adapter = datasourceAdapterFactory.createAccessByRIDAdapter(adapterConfiguration, atype, files);
+ adapter.initialize(ctx);
+ } catch (Exception e) {
+ throw new HyracksDataException("error during creation of external read by RID adapter", e);
+ }
+ writer.open();
+ }
+
+ @Override
+ public void nextFrame(ByteBuffer buffer) throws HyracksDataException {
+ adapter.processNextFrame(buffer, writer);
+ }
+
+ @Override
+ public void close() throws HyracksDataException {
+ //close adapter and flush remaining frame if needed
+ adapter.close(writer);
+ //close writer
+ writer.close();
+ }
+
+ @Override
+ public void fail() throws HyracksDataException {
+ writer.fail();
+ }
+ };
+ }
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataIndexingOperatorDescriptor.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataIndexingOperatorDescriptor.java
new file mode 100644
index 0000000..9ff1f06
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/data/operator/ExternalDataIndexingOperatorDescriptor.java
@@ -0,0 +1,64 @@
+package edu.uci.ics.asterix.external.data.operator;
+
+import java.util.Map;
+
+import edu.uci.ics.asterix.external.adapter.factory.IGenericDatasetAdapterFactory;
+import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
+import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.dataflow.IOperatorNodePushable;
+import edu.uci.ics.hyracks.api.dataflow.value.IRecordDescriptorProvider;
+import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.api.job.JobSpecification;
+import edu.uci.ics.hyracks.dataflow.std.base.AbstractSingleActivityOperatorDescriptor;
+import edu.uci.ics.hyracks.dataflow.std.base.AbstractUnaryOutputSourceOperatorNodePushable;
+
+/*
+ * A single activity operator that provides the functionality of scanning data along
+ * with their RIDs using an instance of the configured adapter.
+ */
+
+public class ExternalDataIndexingOperatorDescriptor extends AbstractSingleActivityOperatorDescriptor{
+
+ private static final long serialVersionUID = 1L;
+
+ private final Map<String, Object> adapterConfiguration;
+ private final Map<String,Integer> files;
+ private final IAType atype;
+ private IGenericDatasetAdapterFactory datasourceAdapterFactory;
+
+ public ExternalDataIndexingOperatorDescriptor(JobSpecification spec, Map<String, Object> arguments, IAType atype,
+ RecordDescriptor rDesc, IGenericDatasetAdapterFactory dataSourceAdapterFactory, Map<String,Integer> files) {
+ super(spec, 0, 1);
+ recordDescriptors[0] = rDesc;
+ this.adapterConfiguration = arguments;
+ this.atype = atype;
+ this.datasourceAdapterFactory = dataSourceAdapterFactory;
+ this.files = files;
+ }
+
+ @Override
+ public IOperatorNodePushable createPushRuntime(final IHyracksTaskContext ctx,
+ IRecordDescriptorProvider recordDescProvider, final int partition, int nPartitions)
+ throws HyracksDataException {
+
+ return new AbstractUnaryOutputSourceOperatorNodePushable() {
+ @Override
+ public void initialize() throws HyracksDataException {
+ writer.open();
+ IDatasourceAdapter adapter = null;
+ try {
+ adapter = ((IGenericDatasetAdapterFactory) datasourceAdapterFactory).createIndexingAdapter(
+ adapterConfiguration, atype, files);
+ adapter.initialize(ctx);
+ adapter.start(partition, writer);
+ } catch (Exception e) {
+ throw new HyracksDataException("exception during reading from external data source", e);
+ } finally {
+ writer.close();
+ }
+ }
+ };
+ }
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAccessByRIDAdapter.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAccessByRIDAdapter.java
new file mode 100644
index 0000000..86a060c
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAccessByRIDAdapter.java
@@ -0,0 +1,1170 @@
+/*
+ * Copyright 2009-2013 by The Regents of the University of California
+ * Licensed 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 from
+ *
+ * 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 edu.uci.ics.asterix.external.dataset.adapter;
+
+import java.io.DataInputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.ByteBuffer;
+import java.util.HashMap;
+import java.util.Map;
+
+import org.apache.hadoop.io.LongWritable;
+import org.apache.hadoop.io.SequenceFile;
+import org.apache.hadoop.io.Text;
+import org.apache.hadoop.io.Writable;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hive.ql.io.RCFile;
+import org.apache.hadoop.hive.ql.io.RCFile.Reader;
+import org.apache.hadoop.hive.serde2.columnar.BytesRefArrayWritable;
+import edu.uci.ics.asterix.common.exceptions.AsterixException;
+import edu.uci.ics.asterix.external.adapter.factory.HDFSAdapterFactory;
+import edu.uci.ics.asterix.om.base.AInt32;
+import edu.uci.ics.asterix.om.base.AInt64;
+import edu.uci.ics.asterix.om.base.AString;
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.asterix.om.types.ATypeTag;
+import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.asterix.runtime.operators.file.ControlledADMTupleParserFactory;
+import edu.uci.ics.asterix.runtime.operators.file.ControlledDelimitedDataTupleParserFactory;
+import edu.uci.ics.asterix.runtime.operators.file.ControlledTupleParser;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.algebricks.common.exceptions.NotImplementedException;
+import edu.uci.ics.hyracks.api.comm.IFrameTupleAccessor;
+import edu.uci.ics.hyracks.api.comm.IFrameWriter;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.dataflow.value.RecordDescriptor;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.common.comm.io.FrameTupleAccessor;
+import edu.uci.ics.hyracks.dataflow.common.comm.util.ByteBufferInputStream;
+import edu.uci.ics.hyracks.dataflow.common.data.parsers.IValueParserFactory;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParserFactory;
+
+/**
+ * Provides functionality for fetching specific external data records stored in an HDFS instance
+ * using their RID.
+ */
+@SuppressWarnings({ "deprecation" })
+public class HDFSAccessByRIDAdapter extends FileSystemBasedAdapter implements IControlledAdapter{
+
+ private static final long serialVersionUID = 1L;
+ private boolean newFrame;
+ private transient ByteBuffer frameBuffer;
+ private String inputFormat;
+ private Configuration conf;
+ private transient FileSystem fs;
+ private RecordDescriptor inRecDesc;
+ private final HashMap<Integer, String> files;
+
+ public HDFSAccessByRIDAdapter(IAType atype, String inputFormat, AlgebricksPartitionConstraint clusterLocations, RecordDescriptor inRecDesc, Configuration conf, HashMap<Integer,String> files) {
+ super(atype);
+ this.inputFormat = inputFormat;
+ this.conf = conf;
+ this.inRecDesc = inRecDesc;
+ this.files = files;
+ }
+
+ @Override
+ public void configure(Map<String, Object> arguments) throws Exception {
+ this.configuration = arguments;
+ fs = FileSystem.get(conf);
+ String specifiedFormat = (String) configuration.get(KEY_FORMAT);
+ if (specifiedFormat == null) {
+ throw new IllegalArgumentException(" Unspecified data format");
+ } else if (FORMAT_DELIMITED_TEXT.equalsIgnoreCase(specifiedFormat)) {
+ parserFactory = getDelimitedDataTupleParserFactory((ARecordType) atype);
+ } else if (FORMAT_ADM.equalsIgnoreCase((String)configuration.get(KEY_FORMAT))) {
+ parserFactory = new ControlledADMTupleParserFactory((ARecordType) atype);
+ } else {
+ throw new IllegalArgumentException(" format " + configuration.get(KEY_FORMAT) + " not supported");
+ }
+ }
+
+ @Override
+ protected ITupleParserFactory getDelimitedDataTupleParserFactory(ARecordType recordType) throws AsterixException {
+ int n = recordType.getFieldTypes().length;
+ IValueParserFactory[] fieldParserFactories = new IValueParserFactory[n];
+ for (int i = 0; i < n; i++) {
+ ATypeTag tag = recordType.getFieldTypes()[i].getTypeTag();
+ IValueParserFactory vpf = typeToValueParserFactMap.get(tag);
+ if (vpf == null) {
+ throw new NotImplementedException("No value parser factory for delimited fields of type " + tag);
+ }
+ fieldParserFactories[i] = vpf;
+ }
+ String delimiterValue = (String) configuration.get(KEY_DELIMITER);
+ if (delimiterValue != null && delimiterValue.length() > 1) {
+ throw new AsterixException("improper delimiter");
+ }
+
+ Character delimiter = delimiterValue.charAt(0);
+ return new ControlledDelimitedDataTupleParserFactory(recordType, fieldParserFactories, delimiter);
+ }
+
+ @Override
+ public void start(int partition, IFrameWriter writer) throws Exception {
+ throw new NotImplementedException("Access by RID adapter doesn't support start function");
+ }
+
+ public void processNextFrame(ByteBuffer buffer, IFrameWriter writer) throws HyracksDataException
+ {
+ frameBuffer = buffer;
+ newFrame = true;
+ ((ControlledTupleParser)parser).parseNext(writer);
+ }
+
+ public void close(IFrameWriter writer) throws HyracksDataException
+ {
+ ((ControlledTupleParser)parser).close(writer);
+ }
+
+ public AdapterType getAdapterType() {
+ return AdapterType.READ;
+ }
+
+ @Override
+ public void initialize(IHyracksTaskContext ctx) throws Exception {
+ this.ctx = ctx;
+ //create parser and initialize it with an instance of the inputStream
+ parser = parserFactory.createTupleParser(ctx);
+ ((ControlledTupleParser)parser).initialize(getInputStream(0));
+ }
+
+ @Override
+ public InputStream getInputStream(int partition) throws IOException {
+
+ //if files map is not null, then it is optimized and we should return optimized inputStream, else return regular
+ if(files == null)
+ {
+
+ //different input stream implementation based on the input format
+ if(inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_RC))
+ {
+ return new InputStream() {
+ private RCFile.Reader reader;
+ private int rowDifference;
+ private String lastFileName = "";
+ private String newFileName;
+ private long lastByteLocation = 0;
+ private long newByteLocation = 0;
+ private int lastRowNumber = 0;
+ private int newRowNumber = 0;
+ private LongWritable key;
+ private BytesRefArrayWritable value;
+ private int EOL = "\n".getBytes()[0];
+ private byte delimiter = 0x01;
+ private boolean pendingValue = false;
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public void close()
+ {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ try {
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ pendingValue = false;
+ }
+
+ //check and see if there is a pending value
+ //Double check this
+ int numBytes = 0;
+ if (pendingValue) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = false;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get 3 things from the current tuple in the frame(File name, byte location and row number)
+ //get the fileName
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileName = ((AString) inRecDesc.getFields()[0].deserialize(dis)).getStringValue();
+ //check if it is a new file
+ if(!lastFileName.equals(newFileName))//stringBuilder.toString()))
+ {
+ //new file
+ lastFileName = newFileName;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+ //open new file
+ reader = new Reader(fs, new Path(lastFileName), conf);
+ //read and save byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ lastByteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(lastByteLocation);
+ //read and save rowNumber
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ lastRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+ //loop until row
+ for(int i=0; i < lastRowNumber; i++)
+ {
+ //this loop perform a single I/O and move to the next record in the block which is already in memory
+ //if no more records in the current block, it perform another I/O and get the next block
+ //<this should never happen here>
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file
+ //get the byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ newByteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+
+ //check if same block
+ if(lastByteLocation != newByteLocation)
+ {
+ //new block
+ lastByteLocation = newByteLocation;
+ //seek
+ reader.seek(lastByteLocation);
+ //read and save rowNumber
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ lastRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+ //loop until row
+ for(int i=0; i < lastRowNumber; i++)
+ {
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same block
+ //get the row number
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ newRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+
+ //calculate row difference
+ rowDifference = newRowNumber - lastRowNumber;
+
+ //update last row number
+ lastRowNumber = newRowNumber;
+
+ //move to the new row
+ for(int i=0; i < rowDifference; i++)
+ {
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ //move to next tuple
+ currentTupleIdx++;
+ }
+ //no more tuples in frame
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ private void copyCurrentTuple(byte[] buffer, int offset) throws IOException {
+ int rcOffset = 0;
+ for(int i=0; i< value.size(); i++)
+ {
+ System.arraycopy(value.get(i).getData(), value.get(i).getStart(), buffer, offset + rcOffset, value.get(i).getLength());
+ rcOffset += value.get(i).getLength() + 1;
+ buffer[rcOffset - 1] = delimiter;
+ }
+ }
+
+ private int getTupleSize(BytesRefArrayWritable value2) {
+ int size=0;
+ //loop over rc column and add lengths
+ for(int i=0; i< value.size(); i++)
+ {
+ size += value.get(i).getLength();
+ }
+ //add delimeters bytes sizes
+ size += value.size() -1;
+ return size;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+ };
+ }
+ else if (inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_TEXT))
+ {
+ return new InputStream() {
+ private FSDataInputStream reader;
+ private String lastFileName = "";
+ private String newFileName;
+ private int EOL = "\n".getBytes()[0];
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private long byteLocation;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private String value;
+ private String pendingValue = null;
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ }
+
+ //check and see if there is a pending value
+ int numBytes = 0;
+ if (pendingValue != null) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = pendingValue.length() + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ //there is enough space
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.length());
+ buffer[offset + numBytes + pendingValue.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = null;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get the fileName
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileName = ((AString) inRecDesc.getFields()[0].deserialize(dis)).getStringValue();
+ //check if it is a new file
+ if(!lastFileName.equals(newFileName))
+ {
+ //new file
+ lastFileName = newFileName;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+ //open new file
+ reader = fs.open(new Path(lastFileName));
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ value = reader.readLine();
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.length() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.length());
+ buffer[offset + numBytes + value.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file, just seek and read
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ value = reader.readLine();
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.length() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.length());
+ buffer[offset + numBytes + value.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ currentTupleIdx++;
+ }
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ @Override
+ public void close(){
+ try {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+
+ };
+ }
+ else if (inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_SEQUENCE))
+ {
+ return new InputStream() {
+ private SequenceFile.Reader reader;
+ private Writable key;
+ private Text value;
+ private String lastFileName = "";
+ private String newFileName;
+ private long byteLocation;
+ private int EOL = "\n".getBytes()[0];
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private Text pendingValue = null;
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ }
+
+ //check and see if there is a pending value
+ //Double check this
+ int numBytes = 0;
+ if (pendingValue != null) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = pendingValue.getLength() + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ //there is enough space
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
+ buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = null;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get the fileName]
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileName = ((AString) inRecDesc.getFields()[0].deserialize(dis)).getStringValue();
+ //check if it is a new file
+ if(!lastFileName.equals(newFileName))
+ {
+ //new file
+ lastFileName = newFileName;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+ //open new file
+ reader = new SequenceFile.Reader(fs,new Path(lastFileName),conf);
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ reader.next(key, value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.getLength() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file, just seek and read
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ reader.next(key, value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.getLength() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ currentTupleIdx++;
+ }
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ @Override
+ public void close(){
+ try {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+ };
+ }
+ //unknow format
+ throw new IOException("Unknown input format");
+ }
+ else
+ {
+ //optimized
+ //different input stream implementation based on the input format
+ if(inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_RC))
+ {
+ return new InputStream() {
+ private RCFile.Reader reader;
+ private int rowDifference;
+ private int lastFileNumber = -1;
+ private int newFileNumber = 0;
+ private long lastByteLocation = 0;
+ private long newByteLocation = 0;
+ private int lastRowNumber = 0;
+ private int newRowNumber = 0;
+ private LongWritable key;
+ private BytesRefArrayWritable value;
+ private int EOL = "\n".getBytes()[0];
+ private byte delimiter = 0x01;
+ private boolean pendingValue = false;
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public void close()
+ {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ try {
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ pendingValue = false;
+ }
+
+ //check and see if there is a pending value
+ //Double check this
+ int numBytes = 0;
+ if (pendingValue) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = false;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get 3 things from the current tuple in the frame(File name, byte location and row number)
+ //get the fileName
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileNumber = ((AInt32) inRecDesc.getFields()[0].deserialize(dis)).getIntegerValue();
+ //check if it is a new file
+ if(lastFileNumber != newFileNumber)
+ {
+ //new file
+ lastFileNumber = newFileNumber;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+ //open new file
+ reader = new Reader(fs, new Path(files.get(newFileNumber)), conf);
+ //read and save byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ lastByteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(lastByteLocation);
+ //read and save rowNumber
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ lastRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+ //loop until row
+ for(int i=0; i < lastRowNumber; i++)
+ {
+ //this loop perform a single I/O and move to the next record in the block which is already in memory
+ //if no more records in the current block, it perform another I/O and get the next block
+ //<this should never happen here>
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file
+ //get the byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ newByteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+
+ //check if same block
+ if(lastByteLocation != newByteLocation)
+ {
+ //new block
+ lastByteLocation = newByteLocation;
+ //seek
+ reader.seek(lastByteLocation);
+ //read and save rowNumber
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ lastRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+ //loop until row
+ for(int i=0; i < lastRowNumber; i++)
+ {
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same block
+ //get the row number
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 2));
+ newRowNumber = ((AInt32)(inRecDesc.getFields()[2].deserialize(dis))).getIntegerValue();
+
+ //calculate row difference
+ rowDifference = newRowNumber - lastRowNumber;
+
+ //update last row number
+ lastRowNumber = newRowNumber;
+
+ //move to the new row
+ for(int i=0; i < rowDifference; i++)
+ {
+ reader.next(key);
+ }
+ //read record
+ reader.getCurrentRow(value);
+
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = true;
+ return numBytes;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ //move to next tuple
+ currentTupleIdx++;
+ }
+ //no more tuples in frame
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ private void copyCurrentTuple(byte[] buffer, int offset) throws IOException {
+ int rcOffset = 0;
+ for(int i=0; i< value.size(); i++)
+ {
+ System.arraycopy(value.get(i).getData(), value.get(i).getStart(), buffer, offset + rcOffset, value.get(i).getLength());
+ rcOffset += value.get(i).getLength() + 1;
+ buffer[rcOffset - 1] = delimiter;
+ }
+ }
+
+ private int getTupleSize(BytesRefArrayWritable value2) {
+ int size=0;
+ //loop over rc column and add lengths
+ for(int i=0; i< value.size(); i++)
+ {
+ size += value.get(i).getLength();
+ }
+ //add delimeters bytes sizes
+ size += value.size() -1;
+ return size;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+ };
+ }
+ else if (inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_TEXT))
+ {
+ return new InputStream() {
+ private FSDataInputStream reader;
+ private int lastFileNumber = -1;
+ private int newFileNumber = 0;
+ private int EOL = "\n".getBytes()[0];
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private long byteLocation;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private String value;
+ private String pendingValue = null;
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ }
+
+ //check and see if there is a pending value
+ int numBytes = 0;
+ if (pendingValue != null) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = pendingValue.length() + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ //there is enough space
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.length());
+ buffer[offset + numBytes + pendingValue.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = null;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get the file number
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileNumber = ((AInt32) inRecDesc.getFields()[0].deserialize(dis)).getIntegerValue();
+ //check if it is a new file
+ if(lastFileNumber != newFileNumber)
+ {
+ //new file
+ lastFileNumber = newFileNumber;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+
+ //open new file
+ reader = fs.open(new Path(files.get(newFileNumber)));
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ value = reader.readLine();
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.length() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.length());
+ buffer[offset + numBytes + value.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file, just seek and read
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ value = reader.readLine();
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.length() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.length());
+ buffer[offset + numBytes + value.length()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ currentTupleIdx++;
+ }
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ @Override
+ public void close(){
+ try {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+
+ };
+ }
+ else if (inputFormat.equals(HDFSAdapterFactory.INPUT_FORMAT_SEQUENCE))
+ {
+ return new InputStream() {
+ private SequenceFile.Reader reader;
+ private Writable key;
+ private Text value;
+ private int lastFileNumber = -1;
+ private int newFileNumber = 0;
+ private long byteLocation;
+ private int EOL = "\n".getBytes()[0];
+ private int currentTupleIdx;
+ private int numberOfTuplesInCurrentFrame;
+ private IFrameTupleAccessor tupleAccessor = new FrameTupleAccessor(ctx.getFrameSize(),inRecDesc);
+ private Text pendingValue = null;
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+
+ if(newFrame)
+ {
+ //first time called with this frame
+ //reset frame buffer
+ tupleAccessor.reset(frameBuffer);
+ //get number of tuples in frame
+ numberOfTuplesInCurrentFrame = tupleAccessor.getTupleCount();
+ //set tuple index to first tuple
+ currentTupleIdx = 0;
+ //set new frame to false
+ newFrame = false;
+ }
+
+ //check and see if there is a pending value
+ //Double check this
+ int numBytes = 0;
+ if (pendingValue != null) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = pendingValue.getLength() + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ //there is enough space
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
+ buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = null;
+ //move to next tuple
+ currentTupleIdx++;
+ }
+
+ //No pending value or done with pending value
+ //check if there are more tuples in the frame
+ while(currentTupleIdx < numberOfTuplesInCurrentFrame)
+ {
+ //get the fileName]
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 0));
+ newFileNumber = ((AInt32) inRecDesc.getFields()[0].deserialize(dis)).getIntegerValue();
+ //check if it is a new file
+ if(lastFileNumber != newFileNumber)
+ {
+ //new file
+ lastFileNumber = newFileNumber;
+ //close old file
+ if(reader != null)
+ {
+ reader.close();
+ }
+ //open new file
+ reader = new SequenceFile.Reader(fs,new Path(files.get(newFileNumber)),conf);
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ reader.next(key, value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.getLength() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ else
+ {
+ //same file, just seek and read
+ //read byte location
+ bbis.setByteBuffer(frameBuffer, tupleAccessor.getTupleStartOffset(currentTupleIdx) + tupleAccessor.getFieldSlotsLength() + tupleAccessor.getFieldStartOffset(currentTupleIdx, 1));
+ byteLocation = ((AInt64) inRecDesc.getFields()[1].deserialize(dis)).getLongValue();
+ //seek
+ reader.seek(byteLocation);
+ //read record
+ reader.next(key, value);
+ //copy it to the buffer if there is enough space
+ int sizeOfNextTuple = value.getLength() + 1;
+ if(sizeOfNextTuple + numBytes > len)
+ {
+ //mark waiting value
+ pendingValue = value;
+ return numBytes;
+ }
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ currentTupleIdx++;
+ }
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ @Override
+ public void close(){
+ try {
+ if (reader != null)
+ {
+ reader.close();
+ }
+ super.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+ };
+ }
+ //unknow format
+ throw new IOException("Unknown input format");
+ }
+ }
+
+ @Override
+ public AlgebricksPartitionConstraint getPartitionConstraint()
+ throws Exception {
+ return partitionConstraint;
+ }
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAdapter.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAdapter.java
index f8b381b..2b355ff 100644
--- a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAdapter.java
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSAdapter.java
@@ -18,6 +18,7 @@
import java.io.InputStream;
import java.util.Map;
+import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.Counters.Counter;
import org.apache.hadoop.mapred.InputSplit;
@@ -26,6 +27,8 @@
import org.apache.hadoop.mapred.Reporter;
import org.apache.hadoop.mapred.SequenceFileInputFormat;
import org.apache.hadoop.mapred.TextInputFormat;
+import org.apache.hadoop.hive.ql.io.RCFileInputFormat;
+import org.apache.hadoop.hive.serde2.columnar.BytesRefArrayWritable;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
@@ -34,202 +37,348 @@
/**
* Provides functionality for fetching external data stored in an HDFS instance.
+ * Note: reader are never closed in adapters. should we make sure they are closed before returning or switching to a different reader?
*/
@SuppressWarnings({ "deprecation", "rawtypes" })
public class HDFSAdapter extends FileSystemBasedAdapter {
- private static final long serialVersionUID = 1L;
+ private static final long serialVersionUID = 1L;
- private transient String[] readSchedule;
- private transient boolean executed[];
- private transient InputSplit[] inputSplits;
- private transient JobConf conf;
- private transient AlgebricksPartitionConstraint clusterLocations;
+ private transient String[] readSchedule;
+ private transient boolean executed[];
+ private transient InputSplit[] inputSplits;
+ private transient JobConf conf;
+ private transient AlgebricksPartitionConstraint clusterLocations;
- private transient String nodeName;
+ private transient String nodeName;
- public HDFSAdapter(IAType atype, String[] readSchedule, boolean[] executed, InputSplit[] inputSplits, JobConf conf,
- AlgebricksPartitionConstraint clusterLocations) {
- super(atype);
- this.readSchedule = readSchedule;
- this.executed = executed;
- this.inputSplits = inputSplits;
- this.conf = conf;
- this.clusterLocations = clusterLocations;
- }
+ public HDFSAdapter(IAType atype, String[] readSchedule, boolean[] executed, InputSplit[] inputSplits, JobConf conf,
+ AlgebricksPartitionConstraint clusterLocations) {
+ super(atype);
+ this.readSchedule = readSchedule;
+ this.executed = executed;
+ this.inputSplits = inputSplits;
+ this.conf = conf;
+ this.clusterLocations = clusterLocations;
+ }
- @Override
- public void configure(Map<String, Object> arguments) throws Exception {
- this.configuration = arguments;
- configureFormat();
- }
+ @Override
+ public void configure(Map<String, Object> arguments) throws Exception {
+ this.configuration = arguments;
+ configureFormat();
+ }
- public AdapterType getAdapterType() {
- return AdapterType.READ_WRITE;
- }
+ public AdapterType getAdapterType() {
+ return AdapterType.READ_WRITE;
+ }
- @Override
- public void initialize(IHyracksTaskContext ctx) throws Exception {
- this.ctx = ctx;
- this.nodeName = ctx.getJobletContext().getApplicationContext().getNodeId();
- }
+ @Override
+ public void initialize(IHyracksTaskContext ctx) throws Exception {
+ this.ctx = ctx;
+ this.nodeName = ctx.getJobletContext().getApplicationContext().getNodeId();
+ }
- private Reporter getReporter() {
- Reporter reporter = new Reporter() {
+ private Reporter getReporter() {
+ Reporter reporter = new Reporter() {
- @Override
- public Counter getCounter(Enum<?> arg0) {
- return null;
- }
+ @Override
+ public Counter getCounter(Enum<?> arg0) {
+ return null;
+ }
- @Override
- public Counter getCounter(String arg0, String arg1) {
- return null;
- }
+ @Override
+ public Counter getCounter(String arg0, String arg1) {
+ return null;
+ }
- @Override
- public InputSplit getInputSplit() throws UnsupportedOperationException {
- return null;
- }
+ @Override
+ public InputSplit getInputSplit() throws UnsupportedOperationException {
+ return null;
+ }
- @Override
- public void incrCounter(Enum<?> arg0, long arg1) {
- }
+ @Override
+ public void incrCounter(Enum<?> arg0, long arg1) {
+ }
- @Override
- public void incrCounter(String arg0, String arg1, long arg2) {
- }
+ @Override
+ public void incrCounter(String arg0, String arg1, long arg2) {
+ }
- @Override
- public void setStatus(String arg0) {
- }
+ @Override
+ public void setStatus(String arg0) {
+ }
- @Override
- public void progress() {
- }
- };
+ @Override
+ public void progress() {
+ }
+ };
- return reporter;
- }
+ return reporter;
+ }
- @Override
- public InputStream getInputStream(int partition) throws IOException {
+ @Override
+ public InputStream getInputStream(int partition) throws IOException {
- return new InputStream() {
+ if(conf.getInputFormat() instanceof RCFileInputFormat)
+ {
+ //if hdfs input format is rc-input-format, we return a different InputStream
+ return new InputStream() {
- private RecordReader<Object, Text> reader;
- private Object key;
- private Text value;
- private boolean hasMore = false;
- private int EOL = "\n".getBytes()[0];
- private Text pendingValue = null;
- private int currentSplitIndex = 0;
+ private RecordReader<LongWritable, BytesRefArrayWritable> reader;
+ private LongWritable key;
+ private BytesRefArrayWritable value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private byte delimiter = 0x01;
+ private boolean pendingValue = false;
+ private int currentSplitIndex = 0;
- @SuppressWarnings("unchecked")
- private boolean moveToNext() throws IOException {
- for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
- /**
- * read all the partitions scheduled to the current node
- */
- if (readSchedule[currentSplitIndex].equals(nodeName)) {
- /**
- * pick an unread split to read
- * synchronize among simultaneous partitions in the same machine
- */
- synchronized (executed) {
- if (executed[currentSplitIndex] == false) {
- executed[currentSplitIndex] = true;
- } else {
- continue;
- }
- }
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
- /**
- * read the split
- */
- reader = getRecordReader(currentSplitIndex);
- key = reader.createKey();
- value = (Text) reader.createValue();
- return true;
- }
- }
- return false;
- }
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = reader.createValue();
+ return true;
+ }
+ }
+ return false;
+ }
- @Override
- public int read(byte[] buffer, int offset, int len) throws IOException {
- if (reader == null) {
- if (!moveToNext()) {
- //nothing to read
- return -1;
- }
- }
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
- int numBytes = 0;
- if (pendingValue != null) {
- System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
- buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
- numBytes += pendingValue.getLength() + 1;
- pendingValue = null;
- }
+ int numBytes = 0;
+ if (pendingValue) {
+ //last value didn't fit into buffer
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = false;
+ }
- while (numBytes < len) {
- hasMore = reader.next(key, value);
- if (!hasMore) {
- while (moveToNext()) {
- hasMore = reader.next(key, value);
- if (hasMore) {
- //move to the next non-empty split
- break;
- }
- }
- }
- if (!hasMore) {
- return (numBytes == 0) ? -1 : numBytes;
- }
- int sizeOfNextTuple = value.getLength() + 1;
- if (numBytes + sizeOfNextTuple > len) {
- // cannot add tuple to current buffer
- // but the reader has moved pass the fetched tuple
- // we need to store this for a subsequent read call.
- // and return this then.
- pendingValue = value;
- break;
- } else {
- System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
- buffer[offset + numBytes + value.getLength()] = (byte) EOL;
- numBytes += sizeOfNextTuple;
- }
- }
- return numBytes;
- }
+ while (numBytes < len) {
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if (numBytes + sizeOfNextTuple > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = true;
+ break;
+ } else {
+ //copy
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
- @Override
- public int read() throws IOException {
- throw new NotImplementedException("Use read(byte[], int, int");
- }
+ private void copyCurrentTuple(byte[] buffer, int offset) throws IOException {
+ int rcOffset = 0;
+ for(int i=0; i< value.size(); i++)
+ {
+ System.arraycopy(value.get(i).getData(), value.get(i).getStart(), buffer, offset + rcOffset, value.get(i).getLength());
+ rcOffset += value.get(i).getLength() + 1;
+ buffer[rcOffset - 1] = delimiter;
+ }
+ }
- private RecordReader getRecordReader(int slitIndex) throws IOException {
- if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
- SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
- RecordReader reader = format.getRecordReader(
- (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
- return reader;
- } else {
- TextInputFormat format = (TextInputFormat) conf.getInputFormat();
- RecordReader reader = format.getRecordReader(
- (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
- return reader;
- }
- }
+ private int getTupleSize(BytesRefArrayWritable value2) {
+ int size=0;
+ //loop over rc column and add lengths
+ for(int i=0; i< value.size(); i++)
+ {
+ size += value.get(i).getLength();
+ }
+ //add delimeters bytes sizes
+ size += value.size() -1;
+ return size;
+ }
- };
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
- }
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ RCFileInputFormat format = (RCFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
- @Override
- public AlgebricksPartitionConstraint getPartitionConstraint() throws Exception {
- return clusterLocations;
- }
+ };
+ }
+ else
+ {
+ return new InputStream() {
+
+ private RecordReader<Object, Text> reader;
+ private Object key;
+ private Text value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private Text pendingValue = null;
+ private int currentSplitIndex = 0;
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = (Text) reader.createValue();
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue != null) {
+ int sizeOfNextTuple = pendingValue.getLength() + 1;
+ if(sizeOfNextTuple > len)
+ {
+ return 0;
+ }
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
+ buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
+ numBytes += pendingValue.getLength() + 1;
+ pendingValue = null;
+ }
+
+ while (numBytes < len) {
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ int sizeOfNextTuple = value.getLength() + 1;
+ if (numBytes + sizeOfNextTuple > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = value;
+ break;
+ } else {
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
+ SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ } else {
+ TextInputFormat format = (TextInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+ }
+
+ };
+ }
+
+ }
+
+ @Override
+ public AlgebricksPartitionConstraint getPartitionConstraint() throws Exception {
+ return clusterLocations;
+ }
}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSIndexingAdapter.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSIndexingAdapter.java
new file mode 100644
index 0000000..59b39c5
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HDFSIndexingAdapter.java
@@ -0,0 +1,1208 @@
+package edu.uci.ics.asterix.external.dataset.adapter;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.util.Map;
+
+import org.apache.hadoop.hive.ql.io.RCFileInputFormat;
+import org.apache.hadoop.hive.serde2.columnar.BytesRefArrayWritable;
+import org.apache.hadoop.io.LongWritable;
+import org.apache.hadoop.io.Text;
+import org.apache.hadoop.mapred.InputSplit;
+import org.apache.hadoop.mapred.FileSplit;
+import org.apache.hadoop.mapred.JobConf;
+import org.apache.hadoop.mapred.RecordReader;
+import org.apache.hadoop.mapred.Reporter;
+import org.apache.hadoop.mapred.SequenceFileInputFormat;
+import org.apache.hadoop.mapred.TextInputFormat;
+import org.apache.hadoop.mapred.Counters.Counter;
+import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.algebricks.common.exceptions.NotImplementedException;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksException;
+
+
+/**
+ * Provides functionality for reading external files and attach RID info to them before they are sent to the parser
+ *
+ * Room for optimization: get numbers of indexing fields (or fields names for adm) and do a quick filtering before sending to parser
+ *
+ */
+@SuppressWarnings({ "deprecation", "rawtypes" })
+public class HDFSIndexingAdapter extends FileSystemBasedAdapter {
+
+ private static final long serialVersionUID = 1L;
+ private transient String[] readSchedule;
+ private transient boolean executed[];
+ private transient InputSplit[] inputSplits;
+ private transient JobConf conf;
+ private transient AlgebricksPartitionConstraint clusterLocations;
+ private final Map<String,Integer> files;
+ private transient String nodeName;
+ public static final byte[] fileNameFieldNameWithRecOpeningBraces = "{\"_file-name\":\"".getBytes();
+ public static final byte[] fileNameFieldClosingQuotation = "\"".getBytes();
+ public static final byte[] fileNumberFieldNameWithRecOpeningBraces = "{\"_file-number\":".getBytes();
+ public static final byte[] bytelocationFieldName = ",\"_byte-location\":".getBytes();
+ public static final byte[] bytelocationValueEnd = "i64,".getBytes();
+
+ public HDFSIndexingAdapter(IAType atype, String[] readSchedule, boolean[] executed, InputSplit[] inputSplits, JobConf conf,
+ AlgebricksPartitionConstraint clusterLocations, Map<String,Integer> files) {
+ super(atype);
+ this.readSchedule = readSchedule;
+ this.executed = executed;
+ this.inputSplits = inputSplits;
+ this.conf = conf;
+ this.clusterLocations = clusterLocations;
+ this.files = files;
+ }
+
+ @Override
+ public void configure(Map<String, Object> arguments) throws Exception {
+ this.configuration = arguments;
+ configureFormat();
+ }
+
+ public AdapterType getAdapterType() {
+ return AdapterType.READ;
+ }
+
+ @Override
+ public void initialize(IHyracksTaskContext ctx) throws Exception {
+ this.ctx = ctx;
+ this.nodeName = ctx.getJobletContext().getApplicationContext().getNodeId();
+ }
+
+ private Reporter getReporter() {
+ Reporter reporter = new Reporter() {
+
+ @Override
+ public Counter getCounter(Enum<?> arg0) {
+ return null;
+ }
+
+ @Override
+ public Counter getCounter(String arg0, String arg1) {
+ return null;
+ }
+
+ @Override
+ public InputSplit getInputSplit() throws UnsupportedOperationException {
+ return null;
+ }
+
+ @Override
+ public void incrCounter(Enum<?> arg0, long arg1) {
+ }
+
+ @Override
+ public void incrCounter(String arg0, String arg1, long arg2) {
+ }
+
+ @Override
+ public void setStatus(String arg0) {
+ }
+
+ @Override
+ public void progress() {
+ }
+ };
+
+ return reporter;
+ }
+
+ @Override
+ public InputStream getInputStream(int partition) throws IOException {
+ if(files == null)
+ {
+ if(conf.getInputFormat() instanceof RCFileInputFormat)
+ {
+ //indexing rc input format
+ return new InputStream() {
+
+ private RecordReader<LongWritable, BytesRefArrayWritable> reader;
+ private LongWritable key;
+ private BytesRefArrayWritable value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private byte delimiter = 0x01;
+ private boolean pendingValue = false;
+ private int currentSplitIndex = 0;
+ private byte[] fileName;
+ private byte[] byteLocation;
+ private byte[] rowNumberBytes;
+ private long blockByteLocation;
+ private long NextblockByteLocation;
+ private int rowNumber;
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = reader.createValue();
+ fileName = ((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath().getBytes();
+ blockByteLocation = reader.getPos();
+ pendingValue = reader.next(key, value);
+ NextblockByteLocation = reader.getPos();
+ rowNumber = 1;
+ byteLocation = String.valueOf(blockByteLocation).getBytes("UTF-8");
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue) {
+ //last value didn't fit into buffer
+ // 1 for EOL
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if (numBytes + sizeOfNextTuple + rowNumberBytes.length + byteLocation.length + fileName.length + 3 > len) {
+ return 0;
+ }
+
+ //copy filename
+ System.arraycopy(fileName, 0, buffer, offset + numBytes, fileName.length);
+ buffer[offset + numBytes + fileName.length] = delimiter;
+ numBytes += fileName.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy row number
+ System.arraycopy(rowNumberBytes, 0, buffer, offset + numBytes, rowNumberBytes.length);
+ buffer[offset + numBytes + rowNumberBytes.length] = delimiter;
+ numBytes += rowNumberBytes.length + 1;
+
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = false;
+ }
+
+ while (numBytes < len) {
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ //check if moved to next block
+ blockByteLocation = reader.getPos();
+ if(blockByteLocation != NextblockByteLocation)
+ {
+ //moved to a new block, reset stuff
+ //row number
+ rowNumber = 1;
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+
+ //block location
+ byteLocation = String.valueOf(NextblockByteLocation).getBytes("UTF-8");
+ NextblockByteLocation = blockByteLocation;
+ }
+ else
+ {
+ rowNumber += 1;
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+ }
+
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if (numBytes + sizeOfNextTuple + rowNumberBytes.length + byteLocation.length + fileName.length + 3 > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = true;
+ break;
+ } else {
+ //copy filename
+ System.arraycopy(fileName, 0, buffer, offset + numBytes, fileName.length);
+ buffer[offset + numBytes + fileName.length] = delimiter;
+ numBytes += fileName.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy row number
+ System.arraycopy(rowNumberBytes, 0, buffer, offset + numBytes, rowNumberBytes.length);
+ buffer[offset + numBytes + rowNumberBytes.length] = delimiter;
+ numBytes += rowNumberBytes.length + 1;
+
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
+
+ private void copyCurrentTuple(byte[] buffer, int offset) throws IOException {
+ int rcOffset = 0;
+ for(int i=0; i< value.size(); i++)
+ {
+ System.arraycopy(value.get(i).getData(), value.get(i).getStart(), buffer, offset + rcOffset, value.get(i).getLength());
+ rcOffset += value.get(i).getLength() + 1;
+ buffer[rcOffset - 1] = delimiter;
+ }
+ }
+
+ private int getTupleSize(BytesRefArrayWritable value2) {
+ int size=0;
+ //loop over rc column and add lengths
+ for(int i=0; i< value.size(); i++)
+ {
+ size += value.get(i).getLength();
+ }
+ //add delimeters bytes sizes
+ size += value.size() -1;
+ return size;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ RCFileInputFormat format = (RCFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+
+ };
+ }
+ else
+ {
+ //get content format
+ if(configuration.get(KEY_FORMAT).equals(FORMAT_DELIMITED_TEXT))
+ {
+ //reading data and RIDs for delimited text
+ return new InputStream() {
+
+ private RecordReader<Object, Text> reader;
+ private Object key;
+ private Text value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private Text pendingValue = null;
+ private int currentSplitIndex = 0;
+ private byte[] fileName;
+ private byte[] byteLocation;
+ private byte delimiter = ((String)configuration.get(KEY_DELIMITER)).getBytes()[0];
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = (Text) reader.createValue();
+ fileName = ((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath().getBytes();
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue != null) {
+ int sizeOfNextTuple = pendingValue.getLength() + 1;
+ if (numBytes + sizeOfNextTuple +byteLocation.length + fileName.length + 2> len)
+ {
+ return numBytes;
+ }
+ //copy filename
+ System.arraycopy(fileName, 0, buffer, offset + numBytes, fileName.length);
+ buffer[offset + numBytes + fileName.length] = delimiter;
+ numBytes += fileName.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy actual value
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
+ buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
+ numBytes += pendingValue.getLength() + 1;
+ pendingValue = null;
+ }
+
+ while (numBytes < len) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes();
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ int sizeOfNextTuple = value.getLength() + 1;
+ if (numBytes + sizeOfNextTuple +byteLocation.length + fileName.length + 2> len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = value;
+ break;
+ } else {
+ //copy filename
+ System.arraycopy(fileName, 0, buffer, offset + numBytes, fileName.length);
+ buffer[offset + numBytes + fileName.length] = delimiter;
+ numBytes += fileName.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //Copy actual value
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
+ SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ } else {
+ TextInputFormat format = (TextInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+ }
+
+ };
+ }
+ else if((configuration.get(KEY_FORMAT).equals(FORMAT_ADM)))
+ {
+ //reading data and RIDs for adm formatted data
+ return new InputStream() {
+
+ private RecordReader<Object, Text> reader;
+ private Object key;
+ private Text value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private Text pendingValue = null;
+ private int currentSplitIndex = 0;
+ private byte[] fileName;
+ private byte[] byteLocation;
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = (Text) reader.createValue();
+ fileName = ((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath().getBytes();
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue != null) {
+ int firstFieldLocation = value.find("\"");
+ int admValueSize = value.getLength();
+ if(firstFieldLocation >= 0)
+ {
+ int sizeOfNextTuple = value.getLength() - firstFieldLocation + 1;
+ int sizeOfNextTupleAndRID = fileNameFieldNameWithRecOpeningBraces.length + fileName.length + fileNameFieldClosingQuotation.length + bytelocationFieldName.length + byteLocation.length + bytelocationValueEnd.length + sizeOfNextTuple;
+ if (numBytes + sizeOfNextTupleAndRID > len) {
+ // still cannot add tuple to current buffer
+ // return 0 so parser would double the buffer size.
+ return 0;
+ } else {
+ //copy fileNameFieldNameWithRecOpeningBraces
+ System.arraycopy(fileNameFieldNameWithRecOpeningBraces, 0, buffer, offset + numBytes,fileNameFieldNameWithRecOpeningBraces.length);
+ numBytes += fileNameFieldNameWithRecOpeningBraces.length;
+ //copy fileName
+ System.arraycopy(fileName, 0, buffer, offset + numBytes,fileName.length);
+ numBytes += fileName.length;
+ //copy fileName closing quotation
+ System.arraycopy(fileNameFieldClosingQuotation, 0, buffer, offset + numBytes,fileNameFieldClosingQuotation.length);
+ numBytes += fileNameFieldClosingQuotation.length;
+ //copy bytelocationFieldName
+ System.arraycopy(bytelocationFieldName, 0, buffer, offset + numBytes,bytelocationFieldName.length);
+ numBytes += bytelocationFieldName.length;
+ //copy byte location value
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes,byteLocation.length);
+ numBytes += byteLocation.length;
+ //copy byte location field end
+ System.arraycopy(bytelocationValueEnd, 0, buffer, offset + numBytes,bytelocationValueEnd.length);
+ numBytes += bytelocationValueEnd.length;
+ //copy the actual adm instance
+ System.arraycopy(value.getBytes(), firstFieldLocation, buffer, offset + numBytes,admValueSize - firstFieldLocation);
+ buffer[offset + numBytes + admValueSize - firstFieldLocation] = (byte) EOL;
+ numBytes += admValueSize - firstFieldLocation +1;
+ }
+ }
+ pendingValue = null;
+ }
+
+ while (numBytes < len) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ //get the index of the first field name
+ int firstFieldLocation = value.find("\"");
+ int admValueSize = value.getLength();
+ if(firstFieldLocation >= 0)
+ {
+ int sizeOfNextTuple = value.getLength() - firstFieldLocation + 1;
+ int sizeOfNextTupleAndRID = fileNameFieldNameWithRecOpeningBraces.length + fileName.length + fileNameFieldClosingQuotation.length + bytelocationFieldName.length + byteLocation.length + bytelocationValueEnd.length + sizeOfNextTuple;
+ if (numBytes + sizeOfNextTupleAndRID > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = value;
+ break;
+ } else {
+ //copy fileNameFieldNameWithRecOpeningBraces
+ System.arraycopy(fileNameFieldNameWithRecOpeningBraces, 0, buffer, offset + numBytes,fileNameFieldNameWithRecOpeningBraces.length);
+ numBytes += fileNameFieldNameWithRecOpeningBraces.length;
+ //copy fileName
+ System.arraycopy(fileName, 0, buffer, offset + numBytes,fileName.length);
+ numBytes += fileName.length;
+ //copy fileName closing quotation
+ System.arraycopy(fileNameFieldClosingQuotation, 0, buffer, offset + numBytes,fileNameFieldClosingQuotation.length);
+ numBytes += fileNameFieldClosingQuotation.length;
+ //copy bytelocationFieldName
+ System.arraycopy(bytelocationFieldName, 0, buffer, offset + numBytes,bytelocationFieldName.length);
+ numBytes += bytelocationFieldName.length;
+ //copy byte location value
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes,byteLocation.length);
+ numBytes += byteLocation.length;
+ //copy byte location field end
+ System.arraycopy(bytelocationValueEnd, 0, buffer, offset + numBytes,bytelocationValueEnd.length);
+ numBytes += bytelocationValueEnd.length;
+ //copy the actual adm instance
+ System.arraycopy(value.getBytes(), firstFieldLocation, buffer, offset + numBytes,admValueSize - firstFieldLocation);
+ buffer[offset + numBytes + admValueSize - firstFieldLocation] = (byte) EOL;
+ numBytes += admValueSize - firstFieldLocation +1;
+ }
+ }
+ }
+ return numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
+ SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ } else {
+ TextInputFormat format = (TextInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+ }
+
+ };
+ }
+ else
+ {
+ throw new IOException("Can't index " +configuration.get(KEY_FORMAT)+" input");
+ }
+ }
+ }
+ else
+ {
+ if(conf.getInputFormat() instanceof RCFileInputFormat)
+ {
+ //indexing rc input format
+ return new InputStream() {
+
+ private RecordReader<LongWritable, BytesRefArrayWritable> reader;
+ private LongWritable key;
+ private BytesRefArrayWritable value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private byte delimiter = 0x01;
+ private boolean pendingValue = false;
+ private int currentSplitIndex = 0;
+ private byte[] fileNumber;
+ private byte[] byteLocation;
+ private byte[] rowNumberBytes;
+ private Integer file;
+ private long blockByteLocation;
+ private long NextblockByteLocation;
+ private int rowNumber;
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = reader.createValue();
+ //getting the file number
+ file = files.get(((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath());
+ if(file == null)
+ {
+ throw new HyracksException("a file was not found in the map while indexing");
+ }
+ fileNumber = String.valueOf(file).getBytes("UTF-8");
+ blockByteLocation = reader.getPos();
+ pendingValue = reader.next(key, value);
+ NextblockByteLocation = reader.getPos();
+ rowNumber = 1;
+ byteLocation = String.valueOf(blockByteLocation).getBytes("UTF-8");
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue) {
+ //last value didn't fit into buffer
+ // 1 for EOL
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if (numBytes + sizeOfNextTuple + rowNumberBytes.length + byteLocation.length + fileNumber.length + 3 > len) {
+ return 0;
+ }
+ //copy file number
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes, fileNumber.length);
+ buffer[offset + numBytes + fileNumber.length] = delimiter;
+ numBytes += fileNumber.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy row number
+ System.arraycopy(rowNumberBytes, 0, buffer, offset + numBytes, rowNumberBytes.length);
+ buffer[offset + numBytes + rowNumberBytes.length] = delimiter;
+ numBytes += rowNumberBytes.length + 1;
+
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ //set pending to false
+ pendingValue = false;
+ }
+
+ while (numBytes < len) {
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+
+ //check if moved to next block
+ blockByteLocation = reader.getPos();
+ if(blockByteLocation != NextblockByteLocation)
+ {
+ //moved to a new block, reset stuff
+ //row number
+ rowNumber = 1;
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+
+ //block location
+ byteLocation = String.valueOf(NextblockByteLocation).getBytes("UTF-8");
+ NextblockByteLocation = blockByteLocation;
+ }
+ else
+ {
+ rowNumber += 1;
+ rowNumberBytes = String.valueOf(rowNumber).getBytes("UTF-8");
+ }
+
+ int sizeOfNextTuple = getTupleSize(value) + 1;
+ if (numBytes + sizeOfNextTuple + rowNumberBytes.length + byteLocation.length + fileNumber.length + 3 > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = true;
+ break;
+ } else {
+ //copy file number
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes, fileNumber.length);
+ buffer[offset + numBytes + fileNumber.length] = delimiter;
+ numBytes += fileNumber.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy row number
+ System.arraycopy(rowNumberBytes, 0, buffer, offset + numBytes, rowNumberBytes.length);
+ buffer[offset + numBytes + rowNumberBytes.length] = delimiter;
+ numBytes += rowNumberBytes.length + 1;
+
+ copyCurrentTuple(buffer, offset + numBytes);
+ buffer[offset + numBytes + sizeOfNextTuple - 1] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
+
+ private void copyCurrentTuple(byte[] buffer, int offset) throws IOException {
+ int rcOffset = 0;
+ for(int i=0; i< value.size(); i++)
+ {
+ System.arraycopy(value.get(i).getData(), value.get(i).getStart(), buffer, offset + rcOffset, value.get(i).getLength());
+ rcOffset += value.get(i).getLength() + 1;
+ buffer[rcOffset - 1] = delimiter;
+ }
+ }
+
+ private int getTupleSize(BytesRefArrayWritable value2) {
+ int size=0;
+ //loop over rc column and add lengths
+ for(int i=0; i< value.size(); i++)
+ {
+ size += value.get(i).getLength();
+ }
+ //add delimeters bytes sizes
+ size += value.size() -1;
+ return size;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ RCFileInputFormat format = (RCFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+
+ };
+ }
+ else
+ {
+ //get content format
+ if(configuration.get(KEY_FORMAT).equals(FORMAT_DELIMITED_TEXT))
+ {
+ //reading data and RIDs for delimited text
+ return new InputStream() {
+
+ private RecordReader<Object, Text> reader;
+ private Object key;
+ private Text value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private Text pendingValue = null;
+ private int currentSplitIndex = 0;
+ private Integer file;
+ private byte[] fileNumber;
+ private byte[] byteLocation;
+ private byte delimiter = ((String)configuration.get(KEY_DELIMITER)).getBytes()[0];
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = (Text) reader.createValue();
+ file = files.get(((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath());
+ if(file == null)
+ {
+ throw new HyracksException("The file:"+((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath()+" was not found in the map while indexing");
+ }
+ fileNumber = String.valueOf(file).getBytes("UTF-8");
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue != null) {
+ int sizeOfNextTuple = pendingValue.getLength() + 1;
+ if (numBytes + sizeOfNextTuple +byteLocation.length + fileNumber.length + 2> len)
+ {
+ return numBytes;
+ }
+ //copy file number
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes, fileNumber.length);
+ buffer[offset + numBytes + fileNumber.length] = delimiter;
+ numBytes += fileNumber.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //copy actual value
+ System.arraycopy(pendingValue.getBytes(), 0, buffer, offset + numBytes, pendingValue.getLength());
+ buffer[offset + numBytes + pendingValue.getLength()] = (byte) EOL;
+ numBytes += pendingValue.getLength() + 1;
+ pendingValue = null;
+ }
+
+ while (numBytes < len) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes();
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ int sizeOfNextTuple = value.getLength() + 1;
+ if (numBytes + sizeOfNextTuple +byteLocation.length + fileNumber.length + 2> len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = value;
+ break;
+ } else {
+ //copy file number
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes, fileNumber.length);
+ buffer[offset + numBytes + fileNumber.length] = delimiter;
+ numBytes += fileNumber.length + 1;
+
+ //copy byte location
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes, byteLocation.length);
+ buffer[offset + numBytes + byteLocation.length] = delimiter;
+ numBytes += byteLocation.length + 1;
+
+ //Copy actual value
+ System.arraycopy(value.getBytes(), 0, buffer, offset + numBytes, value.getLength());
+ buffer[offset + numBytes + value.getLength()] = (byte) EOL;
+ numBytes += sizeOfNextTuple;
+ }
+ }
+ return numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
+ SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ } else {
+ TextInputFormat format = (TextInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+ }
+
+ };
+ }
+ else if((configuration.get(KEY_FORMAT).equals(FORMAT_ADM)))
+ {
+ //reading data and RIDs for adm formatted data
+ return new InputStream() {
+
+ private RecordReader<Object, Text> reader;
+ private Object key;
+ private Text value;
+ private boolean hasMore = false;
+ private int EOL = "\n".getBytes()[0];
+ private Text pendingValue = null;
+ private int currentSplitIndex = 0;
+ private Integer file;
+ private byte[] fileNumber;
+ private byte[] byteLocation;
+
+ @SuppressWarnings("unchecked")
+ private boolean moveToNext() throws IOException {
+ for (; currentSplitIndex < inputSplits.length; currentSplitIndex++) {
+ /**
+ * read all the partitions scheduled to the current node
+ */
+ if (readSchedule[currentSplitIndex].equals(nodeName)) {
+ /**
+ * pick an unread split to read
+ * synchronize among simultaneous partitions in the same machine
+ */
+ synchronized (executed) {
+ if (executed[currentSplitIndex] == false) {
+ executed[currentSplitIndex] = true;
+ } else {
+ continue;
+ }
+ }
+
+ /**
+ * read the split
+ */
+ reader = getRecordReader(currentSplitIndex);
+ key = reader.createKey();
+ value = (Text) reader.createValue();
+ file = files.get(((FileSplit)(inputSplits[currentSplitIndex])).getPath().toUri().getPath());
+ if(file == null)
+ {
+ throw new HyracksException("a file was not found in the map while indexing");
+ }
+ fileNumber = String.valueOf(file).getBytes("UTF-8");
+ return true;
+ }
+ }
+ return false;
+ }
+
+ @Override
+ public int read(byte[] buffer, int offset, int len) throws IOException {
+ if (reader == null) {
+ if (!moveToNext()) {
+ //nothing to read
+ return -1;
+ }
+ }
+
+ int numBytes = 0;
+ if (pendingValue != null) {
+ int firstFieldLocation = value.find("\"");
+ int admValueSize = value.getLength();
+ if(firstFieldLocation >= 0)
+ {
+ int sizeOfNextTuple = value.getLength() - firstFieldLocation + 1;
+ int sizeOfNextTupleAndRID = fileNumberFieldNameWithRecOpeningBraces.length + fileNumber.length + bytelocationFieldName.length + byteLocation.length + bytelocationValueEnd.length + sizeOfNextTuple;
+ if (numBytes + sizeOfNextTupleAndRID > len) {
+ // still cannot add tuple to current buffer
+ // return 0 so parser would double the buffer size.
+ return 0;
+ } else {
+ //copy fileNumberFieldNameWithRecOpeningBraces
+ System.arraycopy(fileNumberFieldNameWithRecOpeningBraces, 0, buffer, offset + numBytes,fileNumberFieldNameWithRecOpeningBraces.length);
+ numBytes += fileNumberFieldNameWithRecOpeningBraces.length;
+ //copy file Number
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes,fileNumber.length);
+ numBytes += fileNumber.length;
+ //copy bytelocationFieldName
+ System.arraycopy(bytelocationFieldName, 0, buffer, offset + numBytes,bytelocationFieldName.length);
+ numBytes += bytelocationFieldName.length;
+ //copy byte location value
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes,byteLocation.length);
+ numBytes += byteLocation.length;
+ //copy byte location field end
+ System.arraycopy(bytelocationValueEnd, 0, buffer, offset + numBytes,bytelocationValueEnd.length);
+ numBytes += bytelocationValueEnd.length;
+ //copy the actual adm instance
+ System.arraycopy(value.getBytes(), firstFieldLocation, buffer, offset + numBytes,admValueSize - firstFieldLocation);
+ buffer[offset + numBytes + admValueSize - firstFieldLocation] = (byte) EOL;
+ numBytes += admValueSize - firstFieldLocation +1;
+ }
+ }
+ pendingValue = null;
+ }
+
+ while (numBytes < len) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (!hasMore) {
+ while (moveToNext()) {
+ //get reader position before you actually read
+ byteLocation = String.valueOf(reader.getPos()).getBytes("UTF-8");
+ hasMore = reader.next(key, value);
+ if (hasMore) {
+ //move to the next non-empty split
+ break;
+ }
+ }
+ }
+ if (!hasMore) {
+ return (numBytes == 0) ? -1 : numBytes;
+ }
+ //get the index of the first field name
+ int firstFieldLocation = value.find("\"");
+ int admValueSize = value.getLength();
+ if(firstFieldLocation >= 0)
+ {
+ int sizeOfNextTuple = value.getLength() - firstFieldLocation + 1;
+ int sizeOfNextTupleAndRID = fileNumberFieldNameWithRecOpeningBraces.length + fileNumber.length + bytelocationFieldName.length + byteLocation.length + bytelocationValueEnd.length + sizeOfNextTuple;
+ if (numBytes + sizeOfNextTupleAndRID > len) {
+ // cannot add tuple to current buffer
+ // but the reader has moved pass the fetched tuple
+ // we need to store this for a subsequent read call.
+ // and return this then.
+ pendingValue = value;
+ break;
+ } else {
+ //copy fileNumberFieldNameWithRecOpeningBraces
+ System.arraycopy(fileNumberFieldNameWithRecOpeningBraces, 0, buffer, offset + numBytes,fileNumberFieldNameWithRecOpeningBraces.length);
+ numBytes += fileNumberFieldNameWithRecOpeningBraces.length;
+ //copy fileNumber
+ System.arraycopy(fileNumber, 0, buffer, offset + numBytes,fileNumber.length);
+ numBytes += fileNumber.length;
+ //copy bytelocationFieldName
+ System.arraycopy(bytelocationFieldName, 0, buffer, offset + numBytes,bytelocationFieldName.length);
+ numBytes += bytelocationFieldName.length;
+ //copy byte location value
+ System.arraycopy(byteLocation, 0, buffer, offset + numBytes,byteLocation.length);
+ numBytes += byteLocation.length;
+ //copy byte location field end
+ System.arraycopy(bytelocationValueEnd, 0, buffer, offset + numBytes,bytelocationValueEnd.length);
+ numBytes += bytelocationValueEnd.length;
+ //copy the actual adm instance
+ System.arraycopy(value.getBytes(), firstFieldLocation, buffer, offset + numBytes,admValueSize - firstFieldLocation);
+ buffer[offset + numBytes + admValueSize - firstFieldLocation] = (byte) EOL;
+ numBytes += admValueSize - firstFieldLocation +1;
+ }
+ }
+ }
+ return numBytes;
+ }
+
+ @Override
+ public int read() throws IOException {
+ throw new NotImplementedException("Use read(byte[], int, int");
+ }
+
+ private RecordReader getRecordReader(int slitIndex) throws IOException {
+ if (conf.getInputFormat() instanceof SequenceFileInputFormat) {
+ SequenceFileInputFormat format = (SequenceFileInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ } else {
+ TextInputFormat format = (TextInputFormat) conf.getInputFormat();
+ RecordReader reader = format.getRecordReader(
+ (org.apache.hadoop.mapred.FileSplit) inputSplits[slitIndex], conf, getReporter());
+ return reader;
+ }
+ }
+
+ };
+ }
+ else
+ {
+ throw new IOException("Can't index " +configuration.get(KEY_FORMAT)+" input");
+ }
+ }
+ }
+ }
+
+ @Override
+ public AlgebricksPartitionConstraint getPartitionConstraint() throws Exception {
+ return clusterLocations;
+ }
+}
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HiveIndexingAdapter.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HiveIndexingAdapter.java
new file mode 100644
index 0000000..178b106
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/HiveIndexingAdapter.java
@@ -0,0 +1,62 @@
+package edu.uci.ics.asterix.external.dataset.adapter;
+
+import java.util.Map;
+
+import org.apache.hadoop.mapred.InputSplit;
+import org.apache.hadoop.mapred.JobConf;
+
+import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint;
+import edu.uci.ics.hyracks.api.comm.IFrameWriter;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+
+/**
+ * Provides the functionality of fetching data in form of ADM records from a Hive dataset.
+ */
+@SuppressWarnings("deprecation")
+public class HiveIndexingAdapter extends AbstractDatasourceAdapter{
+
+ private static final long serialVersionUID = 1L;
+
+ public static final String HIVE_DATABASE = "database";
+ public static final String HIVE_TABLE = "table";
+ public static final String HIVE_HOME = "hive-home";
+ public static final String HIVE_METASTORE_URI = "metastore-uri";
+ public static final String HIVE_WAREHOUSE_DIR = "warehouse-dir";
+ public static final String HIVE_METASTORE_RAWSTORE_IMPL = "rawstore-impl";
+
+ private HDFSIndexingAdapter hdfsIndexingAdapter;
+
+ public HiveIndexingAdapter(IAType atype, String[] readSchedule, boolean[] executed, InputSplit[] inputSplits, JobConf conf,
+ AlgebricksPartitionConstraint clusterLocations, Map<String,Integer> files) {
+ this.hdfsIndexingAdapter = new HDFSIndexingAdapter(atype, readSchedule, executed, inputSplits, conf, clusterLocations, files);
+ this.atype = atype;
+ }
+
+ @Override
+ public AdapterType getAdapterType() {
+ return AdapterType.READ;
+ }
+
+ @Override
+ public void configure(Map<String, Object> arguments) throws Exception {
+ this.configuration = arguments;
+ this.hdfsIndexingAdapter.configure(arguments);
+ }
+
+ @Override
+ public void initialize(IHyracksTaskContext ctx) throws Exception {
+ hdfsIndexingAdapter.initialize(ctx);
+ }
+
+ @Override
+ public void start(int partition, IFrameWriter writer) throws Exception {
+ hdfsIndexingAdapter.start(partition, writer);
+ }
+
+ @Override
+ public AlgebricksPartitionConstraint getPartitionConstraint() throws Exception {
+ return hdfsIndexingAdapter.getPartitionConstraint();
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/IControlledAdapter.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/IControlledAdapter.java
new file mode 100644
index 0000000..253f675
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/dataset/adapter/IControlledAdapter.java
@@ -0,0 +1,17 @@
+package edu.uci.ics.asterix.external.dataset.adapter;
+
+import java.io.Serializable;
+import java.nio.ByteBuffer;
+
+import edu.uci.ics.hyracks.api.comm.IFrameWriter;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+
+public interface IControlledAdapter extends Serializable{
+
+ public void initialize(IHyracksTaskContext ctx) throws Exception;
+
+ public void processNextFrame(ByteBuffer buffer, IFrameWriter writer) throws HyracksDataException;
+
+ public void close(IFrameWriter writer) throws HyracksDataException;
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalDataFilesMetadataProvider.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalDataFilesMetadataProvider.java
new file mode 100644
index 0000000..47550a4
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalDataFilesMetadataProvider.java
@@ -0,0 +1,38 @@
+package edu.uci.ics.asterix.external.util;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.StringTokenizer;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+
+import edu.uci.ics.asterix.external.adapter.factory.HDFSAdapterFactory;
+import edu.uci.ics.asterix.external.dataset.adapter.AbstractDatasourceAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.HDFSAdapter;
+
+public class ExternalDataFilesMetadataProvider {
+ public static ArrayList<FileStatus> getHDFSFileStatus(AbstractDatasourceAdapter adapter) throws IOException
+ {
+ ArrayList<FileStatus> files = new ArrayList<FileStatus>();
+ //Configure hadoop connection
+ Configuration conf = HDFSAdapterFactory.configureHadoopConnection(adapter.getConfiguration());
+ FileSystem fs = FileSystem.get(conf);
+ //get the list of paths from the adapter
+ StringTokenizer tokenizer = new StringTokenizer(((String)adapter.getConfiguration().get(HDFSAdapter.KEY_PATH)),",");
+ Path inputPath = null;
+ FileStatus[] fileStatuses;
+ while(tokenizer.hasMoreTokens())
+ {
+ inputPath = new Path(tokenizer.nextToken().trim());
+ fileStatuses = fs.listStatus(inputPath);
+ for(int i=0; i < fileStatuses.length; i++)
+ {
+ files.add(fileStatuses[i]);
+ }
+ }
+ return files;
+ }
+}
\ No newline at end of file
diff --git a/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalIndexHashPartitionComputerFactory.java b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalIndexHashPartitionComputerFactory.java
new file mode 100644
index 0000000..29fcfb1
--- /dev/null
+++ b/asterix-external-data/src/main/java/edu/uci/ics/asterix/external/util/ExternalIndexHashPartitionComputerFactory.java
@@ -0,0 +1,94 @@
+package edu.uci.ics.asterix.external.util;
+
+import java.io.DataInputStream;
+import java.nio.ByteBuffer;
+
+import edu.uci.ics.asterix.formats.nontagged.AqlSerializerDeserializerProvider;
+import edu.uci.ics.asterix.om.base.AInt64;
+import edu.uci.ics.asterix.om.types.BuiltinType;
+import edu.uci.ics.hyracks.api.comm.IFrameTupleAccessor;
+import edu.uci.ics.hyracks.api.dataflow.value.IBinaryHashFunction;
+import edu.uci.ics.hyracks.api.dataflow.value.IBinaryHashFunctionFactory;
+import edu.uci.ics.hyracks.api.dataflow.value.ISerializerDeserializer;
+import edu.uci.ics.hyracks.api.dataflow.value.ITuplePartitionComputer;
+import edu.uci.ics.hyracks.api.dataflow.value.ITuplePartitionComputerFactory;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.common.comm.util.ByteBufferInputStream;
+
+/* A class that is used to partition external data tuples when building an index over them
+ * the computer it returns, computes the HDFS block value before using the actual hash partitioning
+ * function. this way we ensure that records within ranges of 64MB sizes are partitioned together to the same
+ * data node.
+ */
+
+public class ExternalIndexHashPartitionComputerFactory implements ITuplePartitionComputerFactory{
+ private static final long serialVersionUID = 1L;
+ private final int[] hashFields;
+ private final int bytesInHDFSBlock = 67108864;
+ private final IBinaryHashFunctionFactory[] hashFunctionFactories;
+ @SuppressWarnings("unchecked")
+ private final ISerializerDeserializer<AInt64> longSerde = AqlSerializerDeserializerProvider.INSTANCE.getNonTaggedSerializerDeserializer(BuiltinType.AINT64);
+
+ public ExternalIndexHashPartitionComputerFactory(int[] hashFields, IBinaryHashFunctionFactory[] hashFunctionFactories) {
+ this.hashFields = hashFields;
+ this.hashFunctionFactories = hashFunctionFactories;
+ }
+
+ @Override
+ public ITuplePartitionComputer createPartitioner() {
+ final IBinaryHashFunction[] hashFunctions = new IBinaryHashFunction[hashFunctionFactories.length];
+ for (int i = 0; i < hashFunctionFactories.length; ++i) {
+ hashFunctions[i] = hashFunctionFactories[i].createBinaryHashFunction();
+ }
+ return new ITuplePartitionComputer() {
+ private ByteBuffer serializedLong = ByteBuffer.allocate(8);;
+ private AInt64 byteLocation;
+ private ByteBufferInputStream bbis = new ByteBufferInputStream();
+ private DataInputStream dis = new DataInputStream(bbis);
+ @Override
+ public int partition(IFrameTupleAccessor accessor, int tIndex, int nParts){
+ if (nParts == 1) {
+ return 0;
+ }
+ int h = 0;
+ int startOffset = accessor.getTupleStartOffset(tIndex);
+ int slotLength = accessor.getFieldSlotsLength();
+ for (int j = 0; j < hashFields.length; ++j) {
+ int fIdx = hashFields[j];
+ IBinaryHashFunction hashFn = hashFunctions[j];
+ int fStart = accessor.getFieldStartOffset(tIndex, fIdx);
+ int fEnd = accessor.getFieldEndOffset(tIndex, fIdx);
+ if(j == 1)
+ {
+ //reset the buffer
+ serializedLong.clear();
+ //read byte location
+ bbis.setByteBuffer(accessor.getBuffer() , accessor.getTupleStartOffset(tIndex) + accessor.getFieldSlotsLength() + accessor.getFieldStartOffset(tIndex, hashFields[1]));
+ try {
+ byteLocation = ((AInt64) longSerde.deserialize(dis));
+ //compute the block number, serialize it into a new array and call the hash function
+ serializedLong.putLong(byteLocation.getLongValue()/bytesInHDFSBlock);
+ //call the hash function
+ int fh = hashFn
+ .hash(serializedLong.array(), 0,serializedLong.capacity());
+ h = h * 31 + fh;
+ } catch (HyracksDataException e) {
+ System.err.println("could not serialize the long byte position value!!!");
+ e.printStackTrace();
+ }
+ }
+ else
+ {
+ int fh = hashFn
+ .hash(accessor.getBuffer().array(), startOffset + slotLength + fStart, fEnd - fStart);
+ h = h * 31 + fh;
+ }
+ }
+ if (h < 0) {
+ h = -(h + 1);
+ }
+ return h % nParts;
+ }
+ };
+ }
+}
\ No newline at end of file
diff --git a/asterix-installer/src/main/resources/conf/asterix-configuration.xml b/asterix-installer/src/main/resources/conf/asterix-configuration.xml
index deeb5b0..084a3f8 100644
--- a/asterix-installer/src/main/resources/conf/asterix-configuration.xml
+++ b/asterix-installer/src/main/resources/conf/asterix-configuration.xml
@@ -126,21 +126,6 @@
</property>
<property>
- <name>txn.log.disksectorsize</name>
- <value>4096</value>
- <description>The size of a disk sector. (Default = "4096")
- </description>
- </property>
-
- <property>
- <name>txn.log.groupcommitinterval</name>
- <value>40</value>
- <description>The group commit wait time in milliseconds. (Default =
- "40" // 40ms)
- </description>
- </property>
-
- <property>
<name>txn.log.checkpoint.lsnthreshold</name>
<value>67108864</value>
<description>The size of the window that the maximum LSN is allowed to
diff --git a/asterix-maven-plugins/lexer-generator-maven-plugin/src/main/resources/Lexer.java b/asterix-maven-plugins/lexer-generator-maven-plugin/src/main/resources/Lexer.java
index dae1fb1..92d1c8e 100644
--- a/asterix-maven-plugins/lexer-generator-maven-plugin/src/main/resources/Lexer.java
+++ b/asterix-maven-plugins/lexer-generator-maven-plugin/src/main/resources/Lexer.java
@@ -19,215 +19,235 @@
public class [LEXER_NAME] {
- public static final int
- TOKEN_EOF = 0, TOKEN_AUX_NOT_FOUND = 1 [TOKENS_CONSTANTS];
+ public static final int
+ TOKEN_EOF = 0, TOKEN_AUX_NOT_FOUND = 1 [TOKENS_CONSTANTS];
- // Human representation of tokens. Useful for debug.
- // Is possible to convert a TOKEN_CONSTANT in its image through
- // [LEXER_NAME].tokenKindToString(TOKEN_CONSTANT);
- private static final String[] tokenImage = {
- "<EOF>", "<AUX_NOT_FOUND>" [TOKENS_IMAGES]
- };
-
- private static final char EOF_CHAR = 4;
- protected java.io.Reader inputStream;
- protected int column;
- protected int line;
- protected boolean prevCharIsCR;
- protected boolean prevCharIsLF;
- protected char[] buffer;
- protected int bufsize;
- protected int bufpos;
- protected int tokenBegin;
- protected int endOf_USED_Buffer;
- protected int endOf_UNUSED_Buffer;
- protected int maxUnusedBufferSize;
+ // Human representation of tokens. Useful for debug.
+ // Is possible to convert a TOKEN_CONSTANT in its image through
+ // [LEXER_NAME].tokenKindToString(TOKEN_CONSTANT);
+ private static final String[] tokenImage = {
+ "<EOF>", "<AUX_NOT_FOUND>" [TOKENS_IMAGES]
+ };
-// ================================================================================
-// Auxiliary functions. Can parse the tokens used in the grammar as partial/auxiliary
-// ================================================================================
+ private static final char EOF_CHAR = 4;
+ protected java.io.Reader inputStream;
+ protected int column;
+ protected int line;
+ protected boolean prevCharIsCR;
+ protected boolean prevCharIsLF;
+ protected char[] buffer;
+ protected int bufsize;
+ protected int bufpos;
+ protected int tokenBegin;
+ protected int endOf_USED_Buffer;
+ protected int endOf_UNUSED_Buffer;
+ protected int maxUnusedBufferSize;
- [LEXER_AUXFUNCTIONS]
+ // ================================================================================
+ // Auxiliary functions. Can parse the tokens used in the grammar as partial/auxiliary
+ // ================================================================================
-// ================================================================================
-// Main method. Return a TOKEN_CONSTANT
-// ================================================================================
-
- public int next() throws [LEXER_NAME]Exception, IOException{
- char currentChar = buffer[bufpos];
- while (currentChar == ' ' || currentChar=='\t' || currentChar == '\n' || currentChar=='\r')
- currentChar = readNextChar();
- tokenBegin = bufpos;
- if (currentChar==EOF_CHAR) return TOKEN_EOF;
+ [LEXER_AUXFUNCTIONS]
- [LEXER_LOGIC]
- }
+ // ================================================================================
+ // Main method. Return a TOKEN_CONSTANT
+ // ================================================================================
-// ================================================================================
-// Public interface
-// ================================================================================
-
- public [LEXER_NAME](java.io.Reader stream) throws IOException{
- reInit(stream);
- }
+ public int next() throws [LEXER_NAME]Exception, IOException{
+ if(bufpos < 0)
+ readNextChar();
+ char currentChar = buffer[bufpos];
+ while (currentChar == ' ' || currentChar=='\t' || currentChar == '\n' || currentChar=='\r')
+ currentChar = readNextChar();
+ tokenBegin = bufpos;
+ if (currentChar==EOF_CHAR) return TOKEN_EOF;
- public void reInit(java.io.Reader stream) throws IOException{
- done();
- inputStream = stream;
- bufsize = 4096;
- line = 1;
- column = 0;
- bufpos = -1;
- endOf_UNUSED_Buffer = bufsize;
- endOf_USED_Buffer = 0;
- prevCharIsCR = false;
- prevCharIsLF = false;
- buffer = new char[bufsize];
- tokenBegin = -1;
- maxUnusedBufferSize = 4096/2;
- readNextChar();
- }
+ [LEXER_LOGIC]
+ }
- public String getLastTokenImage() {
- if (bufpos >= tokenBegin)
- return new String(buffer, tokenBegin, bufpos - tokenBegin);
- else
- return new String(buffer, tokenBegin, bufsize - tokenBegin) +
- new String(buffer, 0, bufpos);
- }
-
- public static String tokenKindToString(int token) {
- return tokenImage[token];
- }
+ //used when done with stream, must be called exiplicitly now.
+ public void close()throws IOException
+ {
+ inputStream.close();
+ }
- public void done(){
- buffer = null;
- }
+ //used before processing a new patch in the inputStream
+ public void reset(){
+ line = 1;
+ column = 0;
+ bufpos = -1;
+ endOf_UNUSED_Buffer = bufsize;
+ endOf_USED_Buffer = 0;
+ prevCharIsCR = false;
+ prevCharIsLF = false;
+ tokenBegin = -1;
+ }
-// ================================================================================
-// Parse error management
-// ================================================================================
-
- protected int parseError(String reason) throws [LEXER_NAME]Exception {
- StringBuilder message = new StringBuilder();
- message.append(reason).append("\n");
- message.append("Line: ").append(line).append("\n");
- message.append("Row: ").append(column).append("\n");
- throw new [LEXER_NAME]Exception(message.toString());
- }
+ // ================================================================================
+ // Public interface
+ // ================================================================================
- protected int parseError(int ... tokens) throws [LEXER_NAME]Exception {
- StringBuilder message = new StringBuilder();
- message.append("Error while parsing. ");
- message.append(" Line: ").append(line);
- message.append(" Row: ").append(column);
- message.append(" Expecting:");
- for (int tokenId : tokens){
- message.append(" ").append([LEXER_NAME].tokenKindToString(tokenId));
- }
- throw new [LEXER_NAME]Exception(message.toString());
- }
-
- protected void updateLineColumn(char c){
- column++;
-
- if (prevCharIsLF)
- {
- prevCharIsLF = false;
- line += (column = 1);
- }
- else if (prevCharIsCR)
- {
- prevCharIsCR = false;
- if (c == '\n')
- {
- prevCharIsLF = true;
- }
- else
- {
- line += (column = 1);
- }
- }
-
- if (c=='\r') {
- prevCharIsCR = true;
- } else if(c == '\n') {
- prevCharIsLF = true;
- }
- }
-
-// ================================================================================
-// Read data, buffer management. It uses a circular (and expandable) buffer
-// ================================================================================
+ public [LEXER_NAME](java.io.Reader stream) throws IOException{
+ reInit(stream);
+ }
- protected char readNextChar() throws IOException {
- if (++bufpos >= endOf_USED_Buffer)
- fillBuff();
- char c = buffer[bufpos];
- updateLineColumn(c);
- return c;
- }
+ public void reInit(java.io.Reader stream) throws IOException{
+ done();
+ inputStream = stream;
+ bufsize = 4096;
+ line = 1;
+ column = 0;
+ bufpos = -1;
+ endOf_UNUSED_Buffer = bufsize;
+ endOf_USED_Buffer = 0;
+ prevCharIsCR = false;
+ prevCharIsLF = false;
+ buffer = new char[bufsize];
+ tokenBegin = -1;
+ maxUnusedBufferSize = 4096/2;
+ }
- protected boolean fillBuff() throws IOException {
- if (endOf_UNUSED_Buffer == endOf_USED_Buffer) // If no more unused buffer space
- {
- if (endOf_UNUSED_Buffer == bufsize) // -- If the previous unused space was
- { // -- at the end of the buffer
- if (tokenBegin > maxUnusedBufferSize) // -- -- If the first N bytes before
- { // the current token are enough
- bufpos = endOf_USED_Buffer = 0; // -- -- -- setup buffer to use that fragment
- endOf_UNUSED_Buffer = tokenBegin;
- }
- else if (tokenBegin < 0) // -- -- If no token yet
- bufpos = endOf_USED_Buffer = 0; // -- -- -- reuse the whole buffer
- else
- ExpandBuff(false); // -- -- Otherwise expand buffer after its end
- }
- else if (endOf_UNUSED_Buffer > tokenBegin) // If the endOf_UNUSED_Buffer is after the token
- endOf_UNUSED_Buffer = bufsize; // -- set endOf_UNUSED_Buffer to the end of the buffer
- else if ((tokenBegin - endOf_UNUSED_Buffer) < maxUnusedBufferSize)
- { // If between endOf_UNUSED_Buffer and the token
- ExpandBuff(true); // there is NOT enough space expand the buffer
- } // reorganizing it
- else
- endOf_UNUSED_Buffer = tokenBegin; // Otherwise there is enough space at the start
- } // so we set the buffer to use that fragment
- int i;
- if ((i = inputStream.read(buffer, endOf_USED_Buffer, endOf_UNUSED_Buffer - endOf_USED_Buffer)) == -1)
- {
- inputStream.close();
- buffer[endOf_USED_Buffer]=(char)EOF_CHAR;
- endOf_USED_Buffer++;
- return false;
- }
- else
- endOf_USED_Buffer += i;
- return true;
- }
+ public String getLastTokenImage() {
+ if (bufpos >= tokenBegin)
+ return new String(buffer, tokenBegin, bufpos - tokenBegin);
+ else
+ return new String(buffer, tokenBegin, bufsize - tokenBegin) +
+ new String(buffer, 0, bufpos);
+ }
+
+ public static String tokenKindToString(int token) {
+ return tokenImage[token];
+ }
+
+ public void done(){
+ buffer = null;
+ }
+
+ // ================================================================================
+ // Parse error management
+ // ================================================================================
+
+ protected int parseError(String reason) throws [LEXER_NAME]Exception {
+ StringBuilder message = new StringBuilder();
+ message.append(reason).append("\n");
+ message.append("Line: ").append(line).append("\n");
+ message.append("Row: ").append(column).append("\n");
+ throw new [LEXER_NAME]Exception(message.toString());
+ }
+
+ protected int parseError(int ... tokens) throws [LEXER_NAME]Exception {
+ StringBuilder message = new StringBuilder();
+ message.append("Error while parsing. ");
+ message.append(" Line: ").append(line);
+ message.append(" Row: ").append(column);
+ message.append(" Expecting:");
+ for (int tokenId : tokens){
+ message.append(" ").append([LEXER_NAME].tokenKindToString(tokenId));
+ }
+ throw new [LEXER_NAME]Exception(message.toString());
+ }
+
+ protected void updateLineColumn(char c){
+ column++;
+
+ if (prevCharIsLF)
+ {
+ prevCharIsLF = false;
+ line += (column = 1);
+ }
+ else if (prevCharIsCR)
+ {
+ prevCharIsCR = false;
+ if (c == '\n')
+ {
+ prevCharIsLF = true;
+ }
+ else
+ {
+ line += (column = 1);
+ }
+ }
+
+ if (c=='\r') {
+ prevCharIsCR = true;
+ } else if(c == '\n') {
+ prevCharIsLF = true;
+ }
+ }
+
+ // ================================================================================
+ // Read data, buffer management. It uses a circular (and expandable) buffer
+ // ================================================================================
+
+ protected char readNextChar() throws IOException {
+ if (++bufpos >= endOf_USED_Buffer)
+ fillBuff();
+ char c = buffer[bufpos];
+ updateLineColumn(c);
+ return c;
+ }
+
+ protected boolean fillBuff() throws IOException {
+ if (endOf_UNUSED_Buffer == endOf_USED_Buffer) // If no more unused buffer space
+ {
+ if (endOf_UNUSED_Buffer == bufsize) // -- If the previous unused space was
+ { // -- at the end of the buffer
+ if (tokenBegin > maxUnusedBufferSize) // -- -- If the first N bytes before
+ { // the current token are enough
+ bufpos = endOf_USED_Buffer = 0; // -- -- -- setup buffer to use that fragment
+ endOf_UNUSED_Buffer = tokenBegin;
+ }
+ else if (tokenBegin < 0) // -- -- If no token yet
+ bufpos = endOf_USED_Buffer = 0; // -- -- -- reuse the whole buffer
+ else
+ ExpandBuff(false); // -- -- Otherwise expand buffer after its end
+ }
+ else if (endOf_UNUSED_Buffer > tokenBegin) // If the endOf_UNUSED_Buffer is after the token
+ endOf_UNUSED_Buffer = bufsize; // -- set endOf_UNUSED_Buffer to the end of the buffer
+ else if ((tokenBegin - endOf_UNUSED_Buffer) < maxUnusedBufferSize)
+ { // If between endOf_UNUSED_Buffer and the token
+ ExpandBuff(true); // there is NOT enough space expand the buffer
+ } // reorganizing it
+ else
+ endOf_UNUSED_Buffer = tokenBegin; // Otherwise there is enough space at the start
+ } // so we set the buffer to use that fragment
+ int i;
+ if ((i = inputStream.read(buffer, endOf_USED_Buffer, endOf_UNUSED_Buffer - endOf_USED_Buffer)) == -1)
+ {
+ //moved outside
+ //inputStream.close();
+ buffer[endOf_USED_Buffer]=(char)EOF_CHAR;
+ endOf_USED_Buffer++;
+ return false;
+ }
+ else
+ endOf_USED_Buffer += i;
+ return true;
+ }
- protected void ExpandBuff(boolean wrapAround)
- {
- char[] newbuffer = new char[bufsize + maxUnusedBufferSize];
+ protected void ExpandBuff(boolean wrapAround)
+ {
+ char[] newbuffer = new char[bufsize + maxUnusedBufferSize];
- try {
- if (wrapAround) {
- System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
- System.arraycopy(buffer, 0, newbuffer, bufsize - tokenBegin, bufpos);
- buffer = newbuffer;
- endOf_USED_Buffer = (bufpos += (bufsize - tokenBegin));
- }
- else {
- System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
- buffer = newbuffer;
- endOf_USED_Buffer = (bufpos -= tokenBegin);
- }
- } catch (Throwable t) {
- throw new Error(t.getMessage());
- }
+ try {
+ if (wrapAround) {
+ System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
+ System.arraycopy(buffer, 0, newbuffer, bufsize - tokenBegin, bufpos);
+ buffer = newbuffer;
+ endOf_USED_Buffer = (bufpos += (bufsize - tokenBegin));
+ }
+ else {
+ System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
+ buffer = newbuffer;
+ endOf_USED_Buffer = (bufpos -= tokenBegin);
+ }
+ } catch (Throwable t) {
+ throw new Error(t.getMessage());
+ }
- bufsize += maxUnusedBufferSize;
- endOf_UNUSED_Buffer = bufsize;
- tokenBegin = 0;
- }
+ bufsize += maxUnusedBufferSize;
+ endOf_UNUSED_Buffer = bufsize;
+ tokenBegin = 0;
+ }
}
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataManager.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataManager.java
index 8f0eedb..2de5d78 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataManager.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataManager.java
@@ -31,6 +31,7 @@
import edu.uci.ics.asterix.metadata.entities.DatasourceAdapter;
import edu.uci.ics.asterix.metadata.entities.Datatype;
import edu.uci.ics.asterix.metadata.entities.Dataverse;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.metadata.entities.Function;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.entities.Node;
@@ -268,6 +269,39 @@
}
return dataset;
}
+
+ @Override
+ public List<ExternalFile> getDatasetExternalFiles(
+ MetadataTransactionContext mdTxnCtx, Dataset dataset)
+ throws MetadataException {
+ List<ExternalFile> externalFiles;
+ try {
+ externalFiles = metadataNode.getExternalDatasetFiles(mdTxnCtx.getJobId(), dataset);
+ } catch (RemoteException e) {
+ throw new MetadataException(e);
+ }
+ return externalFiles;
+ }
+
+ @Override
+ public void addExternalFile(MetadataTransactionContext mdTxnCtx,
+ ExternalFile externalFile) throws MetadataException {
+ try {
+ metadataNode.addExternalDatasetFile(mdTxnCtx.getJobId(), externalFile);
+ } catch (RemoteException e) {
+ throw new MetadataException(e);
+ }
+ }
+
+ @Override
+ public void dropExternalFile(MetadataTransactionContext mdTxnCtx,
+ ExternalFile externalFile) throws MetadataException {
+ try {
+ metadataNode.dropExternalFile(mdTxnCtx.getJobId(), externalFile.getDataverseName(), externalFile.getDatasetName(), externalFile.getFileNumber());
+ } catch (RemoteException e) {
+ throw new MetadataException(e);
+ }
+ }
@Override
public List<Index> getDatasetIndexes(MetadataTransactionContext ctx, String dataverseName, String datasetName)
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataNode.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataNode.java
index 8765aae..81ce4f8 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataNode.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/MetadataNode.java
@@ -41,6 +41,8 @@
import edu.uci.ics.asterix.metadata.entities.DatasourceAdapter;
import edu.uci.ics.asterix.metadata.entities.Datatype;
import edu.uci.ics.asterix.metadata.entities.Dataverse;
+import edu.uci.ics.asterix.metadata.entities.ExternalDatasetDetails;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.metadata.entities.Function;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.entities.InternalDatasetDetails;
@@ -50,6 +52,7 @@
import edu.uci.ics.asterix.metadata.entitytupletranslators.DatasourceAdapterTupleTranslator;
import edu.uci.ics.asterix.metadata.entitytupletranslators.DatatypeTupleTranslator;
import edu.uci.ics.asterix.metadata.entitytupletranslators.DataverseTupleTranslator;
+import edu.uci.ics.asterix.metadata.entitytupletranslators.ExternalFileTupleTranslator;
import edu.uci.ics.asterix.metadata.entitytupletranslators.FunctionTupleTranslator;
import edu.uci.ics.asterix.metadata.entitytupletranslators.IndexTupleTranslator;
import edu.uci.ics.asterix.metadata.entitytupletranslators.NodeGroupTupleTranslator;
@@ -59,6 +62,7 @@
import edu.uci.ics.asterix.metadata.valueextractors.MetadataEntityValueExtractor;
import edu.uci.ics.asterix.metadata.valueextractors.NestedDatatypeNameValueExtractor;
import edu.uci.ics.asterix.metadata.valueextractors.TupleCopyValueExtractor;
+import edu.uci.ics.asterix.om.base.AInt32;
import edu.uci.ics.asterix.om.base.AMutableString;
import edu.uci.ics.asterix.om.base.AString;
import edu.uci.ics.asterix.om.types.BuiltinType;
@@ -169,48 +173,56 @@
Index primaryIndex = new Index(dataset.getDataverseName(), dataset.getDatasetName(),
dataset.getDatasetName(), IndexType.BTREE, id.getPrimaryKey(), true, dataset.getPendingOp());
- addIndex(jobId, primaryIndex);
- ITupleReference nodeGroupTuple = createTuple(id.getNodeGroupName(), dataset.getDataverseName(),
- dataset.getDatasetName());
- insertTupleIntoIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, nodeGroupTuple);
- }
- // Add entry in datatype secondary index.
- ITupleReference dataTypeTuple = createTuple(dataset.getDataverseName(), dataset.getItemTypeName(),
- dataset.getDatasetName());
- insertTupleIntoIndex(jobId, MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeTuple);
- } catch (TreeIndexDuplicateKeyException e) {
- throw new MetadataException("A dataset with this name " + dataset.getDatasetName()
- + " already exists in dataverse '" + dataset.getDataverseName() + "'.", e);
- } catch (Exception e) {
- throw new MetadataException(e);
- }
- }
+ addIndex(jobId, primaryIndex);
+ ITupleReference nodeGroupTuple = createTuple(id.getNodeGroupName(), dataset.getDataverseName(),
+ dataset.getDatasetName());
+ insertTupleIntoIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, nodeGroupTuple);
+ }
+ else
+ {
+ //added for external data
+ ExternalDatasetDetails id = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ ITupleReference nodeGroupTuple = createTuple(id.getNodeGroupName(), dataset.getDataverseName(),
+ dataset.getDatasetName());
+ insertTupleIntoIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, nodeGroupTuple);
+ }
+ // Add entry in datatype secondary index.
+ ITupleReference dataTypeTuple = createTuple(dataset.getDataverseName(), dataset.getItemTypeName(),
+ dataset.getDatasetName());
+ insertTupleIntoIndex(jobId, MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeTuple);
+ } catch (TreeIndexDuplicateKeyException e) {
+ throw new MetadataException("A dataset with this name " + dataset.getDatasetName()
+ + " already exists in dataverse '" + dataset.getDataverseName() + "'.", e);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
- @Override
- public void addIndex(JobId jobId, Index index) throws MetadataException, RemoteException {
- try {
- IndexTupleTranslator tupleWriter = new IndexTupleTranslator(true);
- ITupleReference tuple = tupleWriter.getTupleFromMetadataEntity(index);
- insertTupleIntoIndex(jobId, MetadataPrimaryIndexes.INDEX_DATASET, tuple);
- } catch (TreeIndexDuplicateKeyException e) {
- throw new MetadataException("An index with name '" + index.getIndexName() + "' already exists.", e);
- } catch (Exception e) {
- throw new MetadataException(e);
- }
- }
+ @Override
+ public void addIndex(JobId jobId, Index index) throws MetadataException, RemoteException {
+ try {
+ IndexTupleTranslator tupleWriter = new IndexTupleTranslator(true);
+ ITupleReference tuple = tupleWriter.getTupleFromMetadataEntity(index);
+ insertTupleIntoIndex(jobId, MetadataPrimaryIndexes.INDEX_DATASET, tuple);
+ } catch (TreeIndexDuplicateKeyException e) {
+ throw new MetadataException("An index with name '" + index.getIndexName() + "' already exists.", e);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
- @Override
- public void addNode(JobId jobId, Node node) throws MetadataException, RemoteException {
- try {
- NodeTupleTranslator tupleReaderWriter = new NodeTupleTranslator(true);
- ITupleReference tuple = tupleReaderWriter.getTupleFromMetadataEntity(node);
- insertTupleIntoIndex(jobId, MetadataPrimaryIndexes.NODE_DATASET, tuple);
- } catch (TreeIndexDuplicateKeyException e) {
- throw new MetadataException("A node with name '" + node.getNodeName() + "' already exists.", e);
- } catch (Exception e) {
- throw new MetadataException(e);
- }
- }
+ @Override
+ public void addNode(JobId jobId, Node node) throws MetadataException, RemoteException {
+ try {
+ NodeTupleTranslator tupleReaderWriter = new NodeTupleTranslator(true);
+ ITupleReference tuple = tupleReaderWriter.getTupleFromMetadataEntity(node);
+ insertTupleIntoIndex(jobId, MetadataPrimaryIndexes.NODE_DATASET, tuple);
+ } catch (TreeIndexDuplicateKeyException e) {
+ throw new MetadataException("A node with name '" + node.getNodeName() + "' already exists.", e);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
@Override
public void addNodeGroup(JobId jobId, NodeGroup nodeGroup) throws MetadataException, RemoteException {
@@ -384,47 +396,60 @@
// artifacts.
}
- // Delete entry from secondary index 'group'.
- if (dataset.getDatasetType() == DatasetType.INTERNAL || dataset.getDatasetType() == DatasetType.FEED) {
- InternalDatasetDetails id = (InternalDatasetDetails) dataset.getDatasetDetails();
- ITupleReference groupNameSearchKey = createTuple(id.getNodeGroupName(), dataverseName, datasetName);
- // Searches the index for the tuple to be deleted. Acquires an S
- // lock on the GROUPNAME_ON_DATASET_INDEX index.
- try {
- ITupleReference groupNameTuple = getTupleToBeDeleted(jobId,
- MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameSearchKey);
- deleteTupleFromIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameTuple);
- } catch (TreeIndexException tie) {
- // ignore this exception and continue deleting all relevant
- // artifacts.
- }
- }
- // Delete entry from secondary index 'type'.
- ITupleReference dataTypeSearchKey = createTuple(dataverseName, dataset.getItemTypeName(), datasetName);
- // Searches the index for the tuple to be deleted. Acquires an S
- // lock on the DATATYPENAME_ON_DATASET_INDEX index.
- try {
- ITupleReference dataTypeTuple = getTupleToBeDeleted(jobId,
- MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeSearchKey);
- deleteTupleFromIndex(jobId, MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeTuple);
- } catch (TreeIndexException tie) {
- // ignore this exception and continue deleting all relevant
- // artifacts.
- }
+ // Delete entry from secondary index 'group'.
+ if (dataset.getDatasetType() == DatasetType.INTERNAL || dataset.getDatasetType() == DatasetType.FEED) {
+ InternalDatasetDetails id = (InternalDatasetDetails) dataset.getDatasetDetails();
+ ITupleReference groupNameSearchKey = createTuple(id.getNodeGroupName(), dataverseName, datasetName);
+ // Searches the index for the tuple to be deleted. Acquires an S
+ // lock on the GROUPNAME_ON_DATASET_INDEX index.
+ try {
+ ITupleReference groupNameTuple = getTupleToBeDeleted(jobId,
+ MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameSearchKey);
+ deleteTupleFromIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameTuple);
+ } catch (TreeIndexException tie) {
+ // ignore this exception and continue deleting all relevant
+ // artifacts.
+ }
+ }
+ else
+ {
+ ExternalDatasetDetails id = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ ITupleReference groupNameSearchKey = createTuple(id.getNodeGroupName(), dataverseName, datasetName);
+ // Searches the index for the tuple to be deleted. Acquires an S
+ // lock on the GROUPNAME_ON_DATASET_INDEX index.
+ try {
+ ITupleReference groupNameTuple = getTupleToBeDeleted(jobId,
+ MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameSearchKey);
+ deleteTupleFromIndex(jobId, MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX, groupNameTuple);
+ } catch (TreeIndexException tie) {
+ // ignore this exception and continue deleting all relevant
+ // artifacts.
+ }
+ }
+ // Delete entry from secondary index 'type'.
+ ITupleReference dataTypeSearchKey = createTuple(dataverseName, dataset.getItemTypeName(), datasetName);
+ // Searches the index for the tuple to be deleted. Acquires an S
+ // lock on the DATATYPENAME_ON_DATASET_INDEX index.
+ try {
+ ITupleReference dataTypeTuple = getTupleToBeDeleted(jobId,
+ MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeSearchKey);
+ deleteTupleFromIndex(jobId, MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX, dataTypeTuple);
+ } catch (TreeIndexException tie) {
+ // ignore this exception and continue deleting all relevant
+ // artifacts.
+ }
- // Delete entry(s) from the 'indexes' dataset.
- if (dataset.getDatasetType() == DatasetType.INTERNAL || dataset.getDatasetType() == DatasetType.FEED) {
- List<Index> datasetIndexes = getDatasetIndexes(jobId, dataverseName, datasetName);
- if (datasetIndexes != null) {
- for (Index index : datasetIndexes) {
- dropIndex(jobId, dataverseName, datasetName, index.getIndexName());
- }
- }
- }
- } catch (Exception e) {
- throw new MetadataException(e);
- }
- }
+ // Delete entry(s) from the 'indexes' dataset.
+ List<Index> datasetIndexes = getDatasetIndexes(jobId, dataverseName, datasetName);
+ if (datasetIndexes != null) {
+ for (Index index : datasetIndexes) {
+ dropIndex(jobId, dataverseName, datasetName, index.getIndexName());
+ }
+ }
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
@Override
public void dropIndex(JobId jobId, String dataverseName, String datasetName, String indexName)
@@ -1135,8 +1160,116 @@
}
}
- @Override
- public int getMostRecentDatasetId() throws MetadataException, RemoteException {
- return DatasetIdFactory.getMostRecentDatasetId();
+ @Override
+ public List<ExternalFile> getExternalDatasetFiles(JobId jobId,
+ Dataset dataset) throws MetadataException, RemoteException {
+ try {
+ ITupleReference searchKey = createTuple(dataset.getDataverseName(),dataset.getDatasetName());
+ ExternalFileTupleTranslator tupleReaderWriter = new ExternalFileTupleTranslator(false);
+ IValueExtractor<ExternalFile> valueExtractor = new MetadataEntityValueExtractor<ExternalFile>(
+ tupleReaderWriter);
+ List<ExternalFile> results = new ArrayList<ExternalFile>();
+ searchIndex(jobId, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET, searchKey, valueExtractor, results);
+ return results;
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
+
+ @SuppressWarnings("unchecked")
+ public ITupleReference createExternalFileSearchTuple(String dataverseName, String datasetName, int fileNumber) throws HyracksDataException {
+ ISerializerDeserializer<AString> stringSerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(BuiltinType.ASTRING);
+ ISerializerDeserializer<AInt32> intSerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(BuiltinType.AINT32);
+
+ AMutableString aString = new AMutableString("");
+ ArrayTupleBuilder tupleBuilder = new ArrayTupleBuilder(3);
+
+ //dataverse field
+ aString.setValue(dataverseName);
+ stringSerde.serialize(aString, tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+
+ //dataset field
+ aString.setValue(datasetName);
+ stringSerde.serialize(aString, tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+
+ //file number field
+ intSerde.serialize(new AInt32(fileNumber), tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+
+ ArrayTupleReference tuple = new ArrayTupleReference();
+ tuple.reset(tupleBuilder.getFieldEndOffsets(), tupleBuilder.getByteArray());
+ return tuple;
+ }
+
+ public ExternalFile getExternalDatasetFile(JobId jobId,String dataverseName, String datasetName,
+ int fileNumber) throws MetadataException, RemoteException {
+ try {
+ //create the search key
+ ITupleReference searchKey = createExternalFileSearchTuple(dataverseName, datasetName, fileNumber);
+ ExternalFileTupleTranslator tupleReaderWriter = new ExternalFileTupleTranslator(false);
+ IValueExtractor<ExternalFile> valueExtractor = new MetadataEntityValueExtractor<ExternalFile>(
+ tupleReaderWriter);
+ List<ExternalFile> results = new ArrayList<ExternalFile>();
+ searchIndex(jobId, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET, searchKey, valueExtractor, results);
+ return results.get(0);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
+
+ @Override
+ public void dropExternalFile(JobId jobId, String dataverseName,
+ String datasetName, int fileNumber) throws MetadataException,
+ RemoteException {
+ ExternalFile externalFile;
+ try {
+ externalFile = getExternalDatasetFile(jobId, dataverseName, datasetName,fileNumber);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ if (externalFile == null) {
+ throw new MetadataException("Cannot drop external file because it doesn't exist.");
+ }
+ try {
+ // Delete entry from the 'ExternalFile' dataset.
+ ITupleReference searchKey = createExternalFileSearchTuple(dataverseName, datasetName, fileNumber);
+ // Searches the index for the tuple to be deleted. Acquires an S
+ // lock on the 'ExternalFile' dataset.
+ ITupleReference datasetTuple = getTupleToBeDeleted(jobId,
+ MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET, searchKey);
+ deleteTupleFromIndex(jobId, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET, datasetTuple);
+
+ } catch (TreeIndexException e) {
+ throw new MetadataException("Couldn't drop externalFile.", e);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
+ }
+
+ @Override
+ public void addExternalDatasetFile(JobId jobId, ExternalFile externalFile)
+ throws MetadataException, RemoteException {
+ try {
+ // Insert into the 'externalFiles' dataset.
+ ExternalFileTupleTranslator tupleReaderWriter = new ExternalFileTupleTranslator(true);
+ ITupleReference externalFileTuple = tupleReaderWriter.getTupleFromMetadataEntity(externalFile);
+ insertTupleIntoIndex(jobId, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET, externalFileTuple);
+ } catch (TreeIndexDuplicateKeyException e) {
+ throw new MetadataException("An externalFile with this number " + externalFile.getFileNumber()
+ + " already exists in dataset '" + externalFile.getDatasetName() + "' in dataverse '"+externalFile.getDataverseName()+"'.", e);
+ } catch (Exception e) {
+ throw new MetadataException(e);
+ }
}
+
+
+ @Override
+ public int getMostRecentDatasetId() throws MetadataException, RemoteException {
+ return DatasetIdFactory.getMostRecentDatasetId();
+ }
}
+
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataManager.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataManager.java
index 22c5e46..53f72dd 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataManager.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataManager.java
@@ -26,6 +26,7 @@
import edu.uci.ics.asterix.metadata.entities.DatasourceAdapter;
import edu.uci.ics.asterix.metadata.entities.Datatype;
import edu.uci.ics.asterix.metadata.entities.Dataverse;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.metadata.entities.Function;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.entities.Node;
@@ -439,6 +440,36 @@
*/
public List<Function> getDataverseFunctions(MetadataTransactionContext ctx, String dataverseName)
throws MetadataException;
+
+ /**
+ * @param mdTxnCtx
+ * MetadataTransactionContext of an active metadata transaction.
+ * @param externalFile
+ * An instance of type ExternalFile that represents the external file being
+ * added
+ * @throws MetadataException
+ */
+ public void addExternalFile(MetadataTransactionContext mdTxnCtx, ExternalFile externalFile) throws MetadataException;
+
+ /**
+ * @param mdTxnCtx
+ * MetadataTransactionContext of an active metadata transaction.
+ * @param dataset
+ * An instance of type Dataset that represents the "external" dataset
+ * @return A list of external files belonging to the dataset
+ * @throws MetadataException
+ */
+ public List<ExternalFile> getDatasetExternalFiles(MetadataTransactionContext mdTxnCtx, Dataset dataset) throws MetadataException;
+
+ /**
+ * @param mdTxnCtx
+ * MetadataTransactionContext of an active metadata transaction.
+ * @param externalFile
+ * An instance of type ExternalFile that represents the external file being
+ * dropped
+ * @throws MetadataException
+ */
+ public void dropExternalFile(MetadataTransactionContext mdTxnCtx, ExternalFile externalFile) throws MetadataException;
public void initializeDatasetIdFactory(MetadataTransactionContext ctx) throws MetadataException;
@@ -453,4 +484,5 @@
public void releaseReadLatch();
+
}
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataNode.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataNode.java
index d1e63e1..206ef8a 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataNode.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/api/IMetadataNode.java
@@ -28,6 +28,7 @@
import edu.uci.ics.asterix.metadata.entities.DatasourceAdapter;
import edu.uci.ics.asterix.metadata.entities.Datatype;
import edu.uci.ics.asterix.metadata.entities.Dataverse;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.metadata.entities.Function;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.entities.Node;
@@ -471,6 +472,45 @@
* @throws RemoteException
*/
public void addAdapter(JobId jobId, DatasourceAdapter adapter) throws MetadataException, RemoteException;
+
+ /**
+ * @param jobId
+ * A globally unique id for an active metadata transaction.
+ * @param externalFile
+ * An object representing the external file entity
+ * @throws MetadataException
+ * for example, if the file already exists.
+ * @throws RemoteException
+ */
+ public void addExternalDatasetFile(JobId jobId, ExternalFile externalFile) throws MetadataException, RemoteException;
+
+ /**
+ * @param jobId
+ * A globally unique id for an active metadata transaction.
+ * @param dataset
+ * A dataset the files belongs to.
+ * @throws MetadataException
+ * @throws RemoteException
+ */
+ public List<ExternalFile> getExternalDatasetFiles(JobId jobId, Dataset dataset
+ ) throws MetadataException, RemoteException;
+
+ /**
+ * Deletes an externalFile , acquiring local locks on behalf of the given
+ * transaction id.
+ *
+ * @param jobId
+ * A globally unique id for an active metadata transaction.
+ * @param dataverseName
+ * dataverse asociated with the external dataset that owns the file to be deleted.
+ * @param datasetName
+ * Name of dataset owning the file to be deleted.
+ * @param fileNumber
+ * the id number for the file to be deleted
+ * @throws RemoteException
+ */
+ public void dropExternalFile(JobId jobId, String dataverseName, String datasetName, int fileNumber) throws MetadataException,
+ RemoteException;
public void initializeDatasetIdFactory(JobId jobId) throws MetadataException, RemoteException;
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataBootstrap.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataBootstrap.java
index 4a02bc5..ed89cfc 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataBootstrap.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataBootstrap.java
@@ -117,7 +117,7 @@
MetadataPrimaryIndexes.DATASET_DATASET, MetadataPrimaryIndexes.DATATYPE_DATASET,
MetadataPrimaryIndexes.INDEX_DATASET, MetadataPrimaryIndexes.NODE_DATASET,
MetadataPrimaryIndexes.NODEGROUP_DATASET, MetadataPrimaryIndexes.FUNCTION_DATASET,
- MetadataPrimaryIndexes.DATASOURCE_ADAPTER_DATASET };
+ MetadataPrimaryIndexes.DATASOURCE_ADAPTER_DATASET, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET };
secondaryIndexes = new IMetadataIndex[] { MetadataSecondaryIndexes.GROUPNAME_ON_DATASET_INDEX,
MetadataSecondaryIndexes.DATATYPENAME_ON_DATASET_INDEX,
MetadataSecondaryIndexes.DATATYPENAME_ON_DATATYPE_INDEX };
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataPrimaryIndexes.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataPrimaryIndexes.java
index 8bdd92b..100ec40 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataPrimaryIndexes.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataPrimaryIndexes.java
@@ -41,9 +41,11 @@
public static final int NODEGROUP_DATASET_ID = 6;
public static final int FUNCTION_DATASET_ID = 7;
public static final int DATASOURCE_ADAPTER_DATASET_ID = 8;
+ public static final int EXTERNAL_FILE_DATASET_ID = 9;
public static final int FIRST_AVAILABLE_USER_DATASET_ID = 100;
public static IMetadataIndex DATASOURCE_ADAPTER_DATASET;
+ public static IMetadataIndex EXTERNAL_FILE_DATASET;
/**
* Create all metadata primary index descriptors. MetadataRecordTypes must
@@ -92,5 +94,11 @@
BuiltinType.ASTRING, BuiltinType.ASTRING }, new String[] { "DataverseName", "Name" }, 0,
MetadataRecordTypes.DATASOURCE_ADAPTER_RECORDTYPE, DATASOURCE_ADAPTER_DATASET_ID, true, new int[] { 0,
1 });
+
+ EXTERNAL_FILE_DATASET = new MetadataIndex("ExternalFile", null, 4, new IAType[] {
+ BuiltinType.ASTRING, BuiltinType.ASTRING, BuiltinType.AINT32 }, new String[] { "DataverseName", "DatasetName",
+ "FileNumber"}, 0,
+ MetadataRecordTypes.EXTERNAL_FILE_RECORDTYPE, EXTERNAL_FILE_DATASET_ID, true, new int[] { 0,
+ 1, 2 });
}
}
\ No newline at end of file
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataRecordTypes.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataRecordTypes.java
index 50681ee..11f9c91 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataRecordTypes.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/bootstrap/MetadataRecordTypes.java
@@ -47,6 +47,7 @@
public static ARecordType NODEGROUP_RECORDTYPE;
public static ARecordType FUNCTION_RECORDTYPE;
public static ARecordType DATASOURCE_ADAPTER_RECORDTYPE;
+ public static ARecordType EXTERNAL_FILE_RECORDTYPE;
/**
* Create all metadata record types.
@@ -76,6 +77,7 @@
NODEGROUP_RECORDTYPE = createNodeGroupRecordType();
FUNCTION_RECORDTYPE = createFunctionRecordType();
DATASOURCE_ADAPTER_RECORDTYPE = createDatasourceAdapterRecordType();
+ EXTERNAL_FILE_RECORDTYPE = createExternalFileRecordType();
} catch (AsterixException e) {
throw new MetadataException(e);
}
@@ -125,12 +127,13 @@
// external details.
public static final int EXTERNAL_DETAILS_ARECORD_DATASOURCE_ADAPTER_FIELD_INDEX = 0;
public static final int EXTERNAL_DETAILS_ARECORD_PROPERTIES_FIELD_INDEX = 1;
-
+ public static final int EXTERNAL_DETAILS_ARECORD_GROUPNAME_FIELD_INDEX = 2;
+
private static final ARecordType createExternalDetailsRecordType() throws AsterixException {
AOrderedListType orderedPropertyListType = new AOrderedListType(DATASOURCE_ADAPTER_PROPERTIES_RECORDTYPE, null);
- String[] fieldNames = { "DatasourceAdapter", "Properties" };
- IAType[] fieldTypes = { BuiltinType.ASTRING, orderedPropertyListType };
+ String[] fieldNames = { "DatasourceAdapter", "Properties", "GroupName" };
+ IAType[] fieldTypes = { BuiltinType.ASTRING, orderedPropertyListType, BuiltinType.ASTRING };
return new ARecordType(null, fieldNames, fieldTypes, true);
}
@@ -356,5 +359,19 @@
BuiltinType.ASTRING };
return new ARecordType("DatasourceAdapterRecordType", fieldNames, fieldTypes, true);
}
+
+ public static final int EXTERNAL_FILE_ARECORD_DATAVERSENAME_FIELD_INDEX = 0;
+ public static final int EXTERNAL_FILE_ARECORD_DATASET_NAME_FIELD_INDEX = 1;
+ public static final int EXTERNAL_FILE_ARECORD_FILE_NUMBER_FIELD_INDEX = 2;
+ public static final int EXTERNAL_FILE_ARECORD_FILE_NAME_FIELD_INDEX = 3;
+ public static final int EXTERNAL_FILE_ARECORD_FILE_SIZE_FIELD_INDEX = 4;
+ public static final int EXTERNAL_FILE_ARECORD_FILE_MOD_DATE_FIELD_INDEX = 5;
+
+ private static ARecordType createExternalFileRecordType() throws AsterixException {
+ String[] fieldNames = { "DataverseName", "DatasetName", "FileNumber", "FileName", "FileSize", "FileModDate"};
+ IAType[] fieldTypes = { BuiltinType.ASTRING, BuiltinType.ASTRING, BuiltinType.AINT32, BuiltinType.ASTRING, BuiltinType.AINT64,
+ BuiltinType.ADATETIME};
+ return new ARecordType("ExternalFileRecordType", fieldNames, fieldTypes, true);
+ }
}
\ No newline at end of file
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/declared/AqlMetadataProvider.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/declared/AqlMetadataProvider.java
index 8b422c8..078b340d 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/declared/AqlMetadataProvider.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/declared/AqlMetadataProvider.java
@@ -18,11 +18,14 @@
import java.io.File;
import java.io.IOException;
import java.util.ArrayList;
+import java.util.Date;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.logging.Logger;
+import org.apache.hadoop.fs.FileStatus;
+
import edu.uci.ics.asterix.common.config.AsterixStorageProperties;
import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.common.config.DatasetConfig.IndexType;
@@ -44,13 +47,17 @@
import edu.uci.ics.asterix.external.adapter.factory.IAdapterFactory;
import edu.uci.ics.asterix.external.adapter.factory.IGenericDatasetAdapterFactory;
import edu.uci.ics.asterix.external.adapter.factory.ITypedDatasetAdapterFactory;
+import edu.uci.ics.asterix.external.data.operator.ExternalDataAccessByRIDOperatorDescriptor;
+import edu.uci.ics.asterix.external.data.operator.ExternalDataIndexingOperatorDescriptor;
import edu.uci.ics.asterix.external.data.operator.ExternalDataScanOperatorDescriptor;
import edu.uci.ics.asterix.external.data.operator.FeedIntakeOperatorDescriptor;
import edu.uci.ics.asterix.external.data.operator.FeedMessageOperatorDescriptor;
+import edu.uci.ics.asterix.external.dataset.adapter.AbstractDatasourceAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.ITypedDatasourceAdapter;
import edu.uci.ics.asterix.external.feed.lifecycle.FeedId;
import edu.uci.ics.asterix.external.feed.lifecycle.IFeedMessage;
+import edu.uci.ics.asterix.external.util.ExternalDataFilesMetadataProvider;
import edu.uci.ics.asterix.formats.base.IDataFormat;
import edu.uci.ics.asterix.formats.nontagged.AqlBinaryComparatorFactoryProvider;
import edu.uci.ics.asterix.formats.nontagged.AqlTypeTraitProvider;
@@ -64,6 +71,7 @@
import edu.uci.ics.asterix.metadata.entities.Datatype;
import edu.uci.ics.asterix.metadata.entities.Dataverse;
import edu.uci.ics.asterix.metadata.entities.ExternalDatasetDetails;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
import edu.uci.ics.asterix.metadata.entities.FeedDatasetDetails;
import edu.uci.ics.asterix.metadata.entities.Index;
import edu.uci.ics.asterix.metadata.entities.InternalDatasetDetails;
@@ -160,6 +168,7 @@
private boolean asyncResults;
private ResultSetId resultSetId;
private IResultSerializerFactoryProvider resultSerializerFactoryProvider;
+ private static boolean optimizeExternalIndexes = false;
private final Dataverse defaultDataverse;
private JobId jobId;
@@ -169,22 +178,6 @@
private static final Map<String, String> adapterFactoryMapping = initializeAdapterFactoryMapping();
private static Scheduler hdfsScheduler;
- public String getPropertyValue(String propertyName) {
- return config.get(propertyName);
- }
-
- public void setConfig(Map<String, String> config) {
- this.config = config;
- }
-
- public Map<String, String[]> getAllStores() {
- return stores;
- }
-
- public Map<String, String> getConfig() {
- return config;
- }
-
public AqlMetadataProvider(Dataverse defaultDataverse) {
this.defaultDataverse = defaultDataverse;
this.stores = AsterixAppContextInfo.getInstance().getMetadataProperties().getStores();
@@ -264,6 +257,30 @@
public IResultSerializerFactoryProvider getResultSerializerFactoryProvider() {
return resultSerializerFactoryProvider;
}
+
+ public String getPropertyValue(String propertyName) {
+ return config.get(propertyName);
+ }
+
+ public void setConfig(Map<String, String> config) {
+ this.config = config;
+ }
+
+ public Map<String, String[]> getAllStores() {
+ return stores;
+ }
+
+ public Map<String, String> getConfig() {
+ return config;
+ }
+
+ public static boolean isOptimizeExternalIndexes() {
+ return optimizeExternalIndexes;
+ }
+
+ public static void setOptimizeExternalIndexes(boolean optimizeExternalIndexes) {
+ AqlMetadataProvider.optimizeExternalIndexes = optimizeExternalIndexes;
+ }
@Override
public AqlDataSource findDataSource(AqlSourceId id) throws AlgebricksException {
@@ -402,6 +419,215 @@
return new Pair<IOperatorDescriptor, AlgebricksPartitionConstraint>(dataScanner, constraint);
}
+
+ @SuppressWarnings("rawtypes")
+ public Pair<ExternalDataIndexingOperatorDescriptor, AlgebricksPartitionConstraint> buildExternalDataIndexingRuntime(
+ JobSpecification jobSpec, IAType itemType, Dataset dataset, IDataFormat format)
+ throws AlgebricksException {
+ IGenericDatasetAdapterFactory adapterFactory;
+ IDatasourceAdapter adapter;
+ String adapterName;
+ DatasourceAdapter adapterEntity;
+ String adapterFactoryClassname;
+ ExternalDatasetDetails datasetDetails = null;
+ try {
+ datasetDetails = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ adapterName = datasetDetails.getAdapter();
+ adapterEntity = MetadataManager.INSTANCE.getAdapter(mdTxnCtx, MetadataConstants.METADATA_DATAVERSE_NAME,
+ adapterName);
+ if (adapterEntity != null) {
+ adapterFactoryClassname = adapterEntity.getClassname();
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ } else {
+ adapterFactoryClassname = adapterFactoryMapping.get(adapterName);
+ if (adapterFactoryClassname == null) {
+ throw new AlgebricksException(" Unknown adapter :" + adapterName);
+ }
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ }
+
+ adapter = ((IGenericDatasetAdapterFactory) adapterFactory).createIndexingAdapter(
+ wrapProperties(datasetDetails.getProperties()), itemType, null);
+ } catch (AlgebricksException ae) {
+ throw ae;
+ } catch (Exception e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to create adapter " + e);
+ }
+ if (!(adapter.getAdapterType().equals(IDatasourceAdapter.AdapterType.READ) || adapter.getAdapterType().equals(
+ IDatasourceAdapter.AdapterType.READ_WRITE))) {
+ throw new AlgebricksException("external dataset adapter does not support read operation");
+ }
+ ARecordType rt = (ARecordType) itemType;
+ ISerializerDeserializer payloadSerde = format.getSerdeProvider().getSerializerDeserializer(itemType);
+ RecordDescriptor indexerDesc = new RecordDescriptor(new ISerializerDeserializer[] { payloadSerde });
+ ExternalDataIndexingOperatorDescriptor dataIndexScanner = null;
+ List<ExternalFile> files = null;
+ HashMap<String, Integer> filesNumbers = null;
+ if(optimizeExternalIndexes)
+ {
+ try {
+ files = MetadataManager.INSTANCE.getDatasetExternalFiles(mdTxnCtx, dataset);
+ } catch (MetadataException e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to get list of external files from metadata " + e);
+ }
+
+ filesNumbers = new HashMap<String,Integer>();
+ for(int i=0; i< files.size(); i++)
+ {
+ filesNumbers.put(files.get(i).getFileName(), files.get(i).getFileNumber());
+ }
+
+ dataIndexScanner = new ExternalDataIndexingOperatorDescriptor(jobSpec,
+ wrapPropertiesEmpty(datasetDetails.getProperties()), rt, indexerDesc, adapterFactory,filesNumbers);
+ }
+ else
+ {
+ dataIndexScanner = new ExternalDataIndexingOperatorDescriptor(jobSpec,
+ wrapPropertiesEmpty(datasetDetails.getProperties()), rt, indexerDesc, adapterFactory,filesNumbers);
+ }
+ AlgebricksPartitionConstraint constraint;
+ try {
+ constraint = adapter.getPartitionConstraint();
+ } catch (Exception e) {
+ throw new AlgebricksException(e);
+ }
+ return new Pair<ExternalDataIndexingOperatorDescriptor, AlgebricksPartitionConstraint>(dataIndexScanner, constraint);
+ }
+
+ public ArrayList<ExternalFile> getExternalDatasetFiles(Dataset dataset) throws AlgebricksException
+ {
+ ArrayList<ExternalFile> files = new ArrayList<ExternalFile>();
+ if(dataset.getDatasetType() != DatasetType.EXTERNAL)
+ {
+ throw new AlgebricksException("Can only get external dataset files");
+ }
+ ExternalDatasetDetails datasetDetails = (ExternalDatasetDetails)dataset.getDatasetDetails();
+ IGenericDatasetAdapterFactory adapterFactory;
+ IDatasourceAdapter adapter;
+ String adapterName;
+ DatasourceAdapter adapterEntity;
+ String adapterFactoryClassname;
+ try {
+ adapterName = datasetDetails.getAdapter();
+ adapterEntity = MetadataManager.INSTANCE.getAdapter(mdTxnCtx, MetadataConstants.METADATA_DATAVERSE_NAME,
+ adapterName);
+ if (adapterEntity != null) {
+ adapterFactoryClassname = adapterEntity.getClassname();
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ } else {
+ adapterFactoryClassname = adapterFactoryMapping.get(adapterName);
+ if (adapterFactoryClassname == null) {
+ throw new AlgebricksException(" Unknown adapter :" + adapterName);
+ }
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ }
+
+ adapter = ((IGenericDatasetAdapterFactory) adapterFactory).createAdapter(
+ wrapProperties(datasetDetails.getProperties()), null);
+ }
+ catch (Exception e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to create adapter " + e);
+ }
+
+ try {
+ ArrayList<FileStatus> fileStatuses = ExternalDataFilesMetadataProvider.getHDFSFileStatus((AbstractDatasourceAdapter) adapter);
+ for(int i=0; i<fileStatuses.size(); i++)
+ {
+ files.add(new ExternalFile(dataset.getDataverseName(), dataset.getDatasetName(), new Date(fileStatuses.get(i).getModificationTime()),
+ fileStatuses.get(i).getLen(),
+ fileStatuses.get(i).getPath().toUri().getPath(),
+ i));
+ }
+ return files;
+ } catch (IOException e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to get list of HDFS files " + e);
+ }
+ }
+
+ @SuppressWarnings("rawtypes")
+ public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildExternalDataAccesByRIDRuntime(
+ JobSpecification jobSpec, Dataset dataset, Index secondaryIndex)
+ throws AlgebricksException {
+ IAType itemType = null;
+ try {
+ itemType = MetadataManager.INSTANCE.getDatatype(mdTxnCtx, dataset.getDataverseName(), dataset.getItemTypeName()).getDatatype();
+ } catch (MetadataException e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to get item type from metadata " + e);
+ }
+ if (itemType.getTypeTag() != ATypeTag.RECORD) {
+ throw new AlgebricksException("Can only scan datasets of records.");
+ }
+
+ ExternalDatasetDetails datasetDetails = (ExternalDatasetDetails)dataset.getDatasetDetails();
+ IGenericDatasetAdapterFactory adapterFactory;
+ IDatasourceAdapter adapter;
+ String adapterName;
+ DatasourceAdapter adapterEntity;
+ String adapterFactoryClassname;
+ try {
+ adapterName = datasetDetails.getAdapter();
+ adapterEntity = MetadataManager.INSTANCE.getAdapter(mdTxnCtx, MetadataConstants.METADATA_DATAVERSE_NAME,
+ adapterName);
+ if (adapterEntity != null) {
+ adapterFactoryClassname = adapterEntity.getClassname();
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ } else {
+ adapterFactoryClassname = adapterFactoryMapping.get(adapterName);
+ if (adapterFactoryClassname == null) {
+ throw new AlgebricksException(" Unknown adapter :" + adapterName);
+ }
+ adapterFactory = (IGenericDatasetAdapterFactory) Class.forName(adapterFactoryClassname).newInstance();
+ }
+
+ adapter = ((IGenericDatasetAdapterFactory) adapterFactory).createAdapter(
+ wrapProperties(datasetDetails.getProperties()), itemType);
+ } catch (AlgebricksException ae) {
+ throw ae;
+ } catch (Exception e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Unable to create adapter " + e);
+ }
+
+ if (!(adapter.getAdapterType().equals(IDatasourceAdapter.AdapterType.READ) || adapter.getAdapterType().equals(
+ IDatasourceAdapter.AdapterType.READ_WRITE))) {
+ throw new AlgebricksException("external dataset adapter does not support read operation");
+ }
+ IDataFormat format = NonTaggedDataFormat.INSTANCE;
+ ISerializerDeserializer payloadSerde = format.getSerdeProvider().getSerializerDeserializer(itemType);
+ RecordDescriptor outRecDesc = new RecordDescriptor(new ISerializerDeserializer[] { payloadSerde });
+
+ ExternalDataAccessByRIDOperatorDescriptor dataAccessOperator = null;
+ if(optimizeExternalIndexes)
+ {
+ //create the hashmap
+ List<ExternalFile> files=null;
+ try {
+ files = MetadataManager.INSTANCE.getDatasetExternalFiles(mdTxnCtx, dataset);
+ } catch (MetadataException e) {
+ e.printStackTrace();
+ throw new AlgebricksException("Couldn't get file names for access by optimized RIDs",e);
+ }
+ HashMap<Integer, String> filesMapping = new HashMap<Integer, String>();
+ for(int i=0; i < files.size(); i++)
+ {
+ filesMapping.put(files.get(i).getFileNumber(), files.get(i).getFileName());
+ }
+ dataAccessOperator = new ExternalDataAccessByRIDOperatorDescriptor(jobSpec, wrapPropertiesEmpty(datasetDetails.getProperties()),
+ itemType, outRecDesc, adapterFactory, filesMapping);
+ }
+ else
+ {
+ dataAccessOperator = new ExternalDataAccessByRIDOperatorDescriptor(jobSpec, wrapPropertiesEmpty(datasetDetails.getProperties()),
+ itemType, outRecDesc, adapterFactory, null);
+ }
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraints = splitProviderAndPartitionConstraintsForExternalDataset(dataset.getDataverseName(),dataset.getDatasetName(),secondaryIndex.getIndexName());
+ return new Pair<IOperatorDescriptor, AlgebricksPartitionConstraint>(dataAccessOperator, splitsAndConstraints.second);
+ }
@SuppressWarnings("rawtypes")
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildScannerRuntime(JobSpecification jobSpec,
@@ -504,64 +730,111 @@
}
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildBtreeRuntime(JobSpecification jobSpec,
- List<LogicalVariable> outputVars, IOperatorSchema opSchema, IVariableTypeEnvironment typeEnv,
- JobGenContext context, boolean retainInput, Dataset dataset, String indexName, int[] lowKeyFields,
- int[] highKeyFields, boolean lowKeyInclusive, boolean highKeyInclusive, Object implConfig)
- throws AlgebricksException {
- boolean isSecondary = true;
- try {
- Index primaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(),
- dataset.getDatasetName(), dataset.getDatasetName());
- if (primaryIndex != null) {
- isSecondary = !indexName.equals(primaryIndex.getIndexName());
- }
- int numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
- RecordDescriptor outputRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
- int numKeys = numPrimaryKeys;
- int keysStartIndex = outputRecDesc.getFieldCount() - numKeys - 1;
- ITypeTraits[] typeTraits = null;
- int[] bloomFilterKeyFields;
- if (isSecondary) {
- Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(),
- dataset.getDatasetName(), indexName);
- int numSecondaryKeys = secondaryIndex.getKeyFieldNames().size();
- numKeys += numSecondaryKeys;
- keysStartIndex = outputVars.size() - numKeys;
- typeTraits = JobGenHelper.variablesToTypeTraits(outputVars, keysStartIndex, numKeys, typeEnv, context);
- bloomFilterKeyFields = new int[numSecondaryKeys];
- for (int i = 0; i < numSecondaryKeys; i++) {
- bloomFilterKeyFields[i] = i;
- }
- } else {
- typeTraits = JobGenHelper.variablesToTypeTraits(outputVars, keysStartIndex, numKeys + 1, typeEnv,
- context);
- bloomFilterKeyFields = new int[numPrimaryKeys];
- for (int i = 0; i < numPrimaryKeys; i++) {
- bloomFilterKeyFields[i] = i;
- }
- }
- IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(
- outputVars, keysStartIndex, numKeys, typeEnv, context);
+ List<LogicalVariable> outputVars, IOperatorSchema opSchema, IVariableTypeEnvironment typeEnv,
+ JobGenContext context, boolean retainInput, Dataset dataset, String indexName, int[] lowKeyFields,
+ int[] highKeyFields, boolean lowKeyInclusive, boolean highKeyInclusive, Object implConfig)
+ throws AlgebricksException {
+ boolean isSecondary = true;
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL){
+ try {
+ int numPrimaryKeys = DatasetUtils.getExternalRIDSize(dataset);
+ RecordDescriptor outputRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
+ int numKeys = numPrimaryKeys;;
+ ITypeTraits[] typeTraits = null;
+ int[] bloomFilterKeyFields;
+ Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(),
+ dataset.getDatasetName(), indexName);
+ int numSecondaryKeys = secondaryIndex.getKeyFieldNames().size();
+ numKeys += numSecondaryKeys;
+ int keysStartIndex = outputVars.size() - numKeys;
+ typeTraits = JobGenHelper.variablesToTypeTraits(outputVars, keysStartIndex, numKeys, typeEnv, context);
+ bloomFilterKeyFields = new int[numSecondaryKeys];
+ for (int i = 0; i < numSecondaryKeys; i++) {
+ bloomFilterKeyFields[i] = i;
+ }
+ IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(
+ outputVars, keysStartIndex, numKeys, typeEnv, context);
+ IAsterixApplicationContextInfo appContext = (IAsterixApplicationContextInfo) context.getAppContext();
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> spPc;
+ try {
+ spPc = splitProviderAndPartitionConstraintsForExternalDataset(dataset.getDataverseName(),
+ dataset.getDatasetName(), indexName);
+ } catch (Exception e) {
+ throw new AlgebricksException(e);
+ }
+ ISearchOperationCallbackFactory searchCallbackFactory = null;
+ searchCallbackFactory = new SecondaryIndexSearchOperationCallbackFactory();
+ AsterixRuntimeComponentsProvider rtcProvider = AsterixRuntimeComponentsProvider.RUNTIME_PROVIDER;
+ BTreeSearchOperatorDescriptor btreeSearchOp = new BTreeSearchOperatorDescriptor(jobSpec, outputRecDesc,
+ appContext.getStorageManagerInterface(), appContext.getIndexLifecycleManagerProvider(), spPc.first,
+ typeTraits, comparatorFactories, bloomFilterKeyFields, lowKeyFields, highKeyFields,
+ lowKeyInclusive, highKeyInclusive, new LSMBTreeDataflowHelperFactory(
+ new AsterixVirtualBufferCacheProvider(dataset.getDatasetId()), rtcProvider,
+ isSecondary ? new SecondaryIndexOperationTrackerProvider(
+ LSMBTreeIOOperationCallbackFactory.INSTANCE, dataset.getDatasetId())
+ : new PrimaryIndexOperationTrackerProvider(dataset.getDatasetId()), rtcProvider,
+ rtcProvider, storageProperties.getBloomFilterFalsePositiveRate()), retainInput,
+ searchCallbackFactory);
+ return new Pair<IOperatorDescriptor, AlgebricksPartitionConstraint>(btreeSearchOp, spPc.second);
+ } catch (MetadataException me) {
+ throw new AlgebricksException(me);
+ }
+ }
+ else
+ {
+ try {
+ Index primaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(),
+ dataset.getDatasetName(), dataset.getDatasetName());
+ if (primaryIndex != null) {
+ isSecondary = !indexName.equals(primaryIndex.getIndexName());
+ }
+ int numPrimaryKeys = DatasetUtils.getPartitioningKeys(dataset).size();
+ RecordDescriptor outputRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
+ int numKeys = numPrimaryKeys;
+ int keysStartIndex = outputRecDesc.getFieldCount() - numKeys - 1;
+ ITypeTraits[] typeTraits = null;
+ int[] bloomFilterKeyFields;
+ if (isSecondary) {
+ Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(),
+ dataset.getDatasetName(), indexName);
+ int numSecondaryKeys = secondaryIndex.getKeyFieldNames().size();
+ numKeys += numSecondaryKeys;
+ keysStartIndex = outputVars.size() - numKeys;
+ typeTraits = JobGenHelper.variablesToTypeTraits(outputVars, keysStartIndex, numKeys, typeEnv, context);
+ bloomFilterKeyFields = new int[numSecondaryKeys];
+ for (int i = 0; i < numSecondaryKeys; i++) {
+ bloomFilterKeyFields[i] = i;
+ }
+ } else {
+ typeTraits = JobGenHelper.variablesToTypeTraits(outputVars, keysStartIndex, numKeys + 1, typeEnv,
+ context);
+ bloomFilterKeyFields = new int[numPrimaryKeys];
+ for (int i = 0; i < numPrimaryKeys; i++) {
+ bloomFilterKeyFields[i] = i;
+ }
+ }
+ IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(
+ outputVars, keysStartIndex, numKeys, typeEnv, context);
- IAsterixApplicationContextInfo appContext = (IAsterixApplicationContextInfo) context.getAppContext();
- Pair<IFileSplitProvider, AlgebricksPartitionConstraint> spPc;
- try {
- spPc = splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataset.getDataverseName(),
- dataset.getDatasetName(), indexName);
- } catch (Exception e) {
- throw new AlgebricksException(e);
- }
+ IAsterixApplicationContextInfo appContext = (IAsterixApplicationContextInfo) context.getAppContext();
+ Pair<IFileSplitProvider, AlgebricksPartitionConstraint> spPc;
+ try {
+ spPc = splitProviderAndPartitionConstraintsForInternalOrFeedDataset(dataset.getDataverseName(),
+ dataset.getDatasetName(), indexName);
+ } catch (Exception e) {
+ throw new AlgebricksException(e);
+ }
- ISearchOperationCallbackFactory searchCallbackFactory = null;
- if (isSecondary) {
- searchCallbackFactory = new SecondaryIndexSearchOperationCallbackFactory();
- } else {
- JobId jobId = ((JobEventListenerFactory) jobSpec.getJobletEventListenerFactory()).getJobId();
- int datasetId = dataset.getDatasetId();
- int[] primaryKeyFields = new int[numPrimaryKeys];
- for (int i = 0; i < numPrimaryKeys; i++) {
- primaryKeyFields[i] = i;
- }
+ ISearchOperationCallbackFactory searchCallbackFactory = null;
+ if (isSecondary) {
+ searchCallbackFactory = new SecondaryIndexSearchOperationCallbackFactory();
+ } else {
+ JobId jobId = ((JobEventListenerFactory) jobSpec.getJobletEventListenerFactory()).getJobId();
+ int datasetId = dataset.getDatasetId();
+ int[] primaryKeyFields = new int[numPrimaryKeys];
+ for (int i = 0; i < numPrimaryKeys; i++) {
+ primaryKeyFields[i] = i;
+ }
AqlMetadataImplConfig aqlMetadataImplConfig = (AqlMetadataImplConfig) implConfig;
ITransactionSubsystemProvider txnSubsystemProvider = new TransactionSubsystemProvider();
@@ -585,13 +858,14 @@
rtcProvider, storageProperties.getBloomFilterFalsePositiveRate()), retainInput,
searchCallbackFactory);
- return new Pair<IOperatorDescriptor, AlgebricksPartitionConstraint>(btreeSearchOp, spPc.second);
+ return new Pair<IOperatorDescriptor, AlgebricksPartitionConstraint>(btreeSearchOp, spPc.second);
- } catch (MetadataException me) {
- throw new AlgebricksException(me);
- }
- }
-
+ } catch (MetadataException me) {
+ throw new AlgebricksException(me);
+ }
+ }
+ }
+
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildRtreeRuntime(JobSpecification jobSpec,
List<LogicalVariable> outputVars, IOperatorSchema opSchema, IVariableTypeEnvironment typeEnv,
JobGenContext context, boolean retainInput, Dataset dataset, String indexName, int[] keyFields)
@@ -1347,9 +1621,20 @@
}
int numPartitions = 0;
- InternalDatasetDetails datasetDetails = (InternalDatasetDetails) dataset.getDatasetDetails();
- List<String> nodeGroup = MetadataManager.INSTANCE.getNodegroup(mdTxnCtx, datasetDetails.getNodeGroupName())
- .getNodeNames();
+ List<String> nodeGroup = null;
+ if(dataset.getDatasetType() == DatasetType.EXTERNAL)
+ {
+ ExternalDatasetDetails datasetDetails = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ nodeGroup = MetadataManager.INSTANCE.getNodegroup(mdTxnCtx, datasetDetails.getNodeGroupName())
+ .getNodeNames();
+ }
+ else
+ {
+ InternalDatasetDetails datasetDetails = (InternalDatasetDetails) dataset.getDatasetDetails();
+ nodeGroup = MetadataManager.INSTANCE.getNodegroup(mdTxnCtx, datasetDetails.getNodeGroupName())
+ .getNodeNames();
+ }
+
for (String nd : nodeGroup) {
numPartitions += AsterixClusterProperties.INSTANCE.getNumberOfIODevices(nd);
}
@@ -1367,6 +1652,12 @@
return splitProviderAndPartitionConstraints(splits);
}
+ public Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitProviderAndPartitionConstraintsForExternalDataset(
+ String dataverseName, String datasetName, String targetIdxName) throws AlgebricksException {
+ FileSplit[] splits = splitsForExternalDataset(mdTxnCtx, dataverseName, datasetName, targetIdxName);
+ return splitProviderAndPartitionConstraints(splits);
+ }
+
public Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitProviderAndPartitionConstraintsForDataverse(
String dataverse) {
FileSplit[] splits = splitsForDataverse(mdTxnCtx, dataverse);
@@ -1457,6 +1748,56 @@
}
}
+ private FileSplit[] splitsForExternalDataset(MetadataTransactionContext mdTxnCtx, String dataverseName,
+ String datasetName, String targetIdxName) throws AlgebricksException {
+
+ try {
+ File relPathFile = new File(getRelativePath(dataverseName, datasetName + "_idx_" + targetIdxName));
+ Dataset dataset = MetadataManager.INSTANCE.getDataset(mdTxnCtx, dataverseName, datasetName);
+ if (dataset.getDatasetType() != DatasetType.EXTERNAL) {
+ throw new AlgebricksException("Not an external dataset");
+ }
+ ExternalDatasetDetails datasetDetails = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ List<String> nodeGroup = MetadataManager.INSTANCE.getNodegroup(mdTxnCtx, datasetDetails.getNodeGroupName())
+ .getNodeNames();
+ if (nodeGroup == null) {
+ throw new AlgebricksException("Couldn't find node group " + datasetDetails.getNodeGroupName());
+ }
+
+ List<FileSplit> splitArray = new ArrayList<FileSplit>();
+ for (String nd : nodeGroup) {
+ String[] nodeStores = stores.get(nd);
+ if (nodeStores == null) {
+ LOGGER.warning("Node " + nd + " has no stores.");
+ throw new AlgebricksException("Node " + nd + " has no stores.");
+ } else {
+ int numIODevices;
+ if (datasetDetails.getNodeGroupName().compareTo(MetadataConstants.METADATA_NODEGROUP_NAME) == 0) {
+ numIODevices = 1;
+ } else {
+ numIODevices = AsterixClusterProperties.INSTANCE.getNumberOfIODevices(nd);
+ }
+ String[] ioDevices = AsterixClusterProperties.INSTANCE.getIODevices(nd);
+ for (int j = 0; j < nodeStores.length; j++) {
+ for (int k = 0; k < numIODevices; k++) {
+ File f = new File(ioDevices[k] + File.separator + nodeStores[j] + File.separator
+ + relPathFile);
+ splitArray.add(new FileSplit(nd, new FileReference(f), k));
+ }
+ }
+ }
+ }
+ FileSplit[] splits = new FileSplit[splitArray.size()];
+ int i = 0;
+ for (FileSplit fs : splitArray) {
+ splits[i++] = fs;
+ }
+ return splits;
+ } catch (MetadataException me) {
+ throw new AlgebricksException(me);
+ }
+ }
+
private static Map<String, String> initializeAdapterFactoryMapping() {
Map<String, String> adapterFactoryMapping = new HashMap<String, String>();
adapterFactoryMapping.put("edu.uci.ics.asterix.external.dataset.adapter.NCFileSystemAdapter",
@@ -1469,6 +1810,8 @@
"edu.uci.ics.asterix.external.dataset.adapter..RSSFeedAdapterFactory");
adapterFactoryMapping.put("edu.uci.ics.asterix.external.dataset.adapter.CNNFeedAdapter",
"edu.uci.ics.asterix.external.dataset.adapter.CNNFeedAdapterFactory");
+ adapterFactoryMapping.put("edu.uci.ics.asterix.external.dataset.adapter.HiveAdapter",
+ "edu.uci.ics.asterix.external.adapter.factory.HiveAdapterFactory");
return adapterFactoryMapping;
}
@@ -1567,3 +1910,4 @@
}
}
+
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalDatasetDetails.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalDatasetDetails.java
index 18cef340..849c6bd 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalDatasetDetails.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalDatasetDetails.java
@@ -40,17 +40,25 @@
private static final long serialVersionUID = 1L;
private final String adapter;
- private final Map<String, String> properties;
+ private final Map<String,String> properties;
+ private final String nodeGroupName;
- private final static ARecordType externalRecordType = MetadataRecordTypes.EXTERNAL_DETAILS_RECORDTYPE;
+ private final static ARecordType externalRecordType = MetadataRecordTypes.EXTERNAL_DETAILS_RECORDTYPE;
private final static ARecordType propertyRecordType = MetadataRecordTypes.DATASOURCE_ADAPTER_PROPERTIES_RECORDTYPE;
- public ExternalDatasetDetails(String adapter, Map<String, String> properties) {
+ public ExternalDatasetDetails(String adapter, Map<String,String> properties, String nodeGroupName) {
this.properties = properties;
this.adapter = adapter;
+ this.nodeGroupName = nodeGroupName;
+ }
+
+ public ExternalDatasetDetails(String adapter, Map<String,String> properties) {
+ this.properties = properties;
+ this.adapter = adapter;
+ this.nodeGroupName = null;
}
- public String getAdapter() {
+ public String getAdapter() {
return adapter;
}
@@ -58,6 +66,10 @@
return properties;
}
+ public String getNodeGroupName() {
+ return nodeGroupName;
+ }
+
@Override
public DatasetType getDatasetType() {
return DatasetType.EXTERNAL;
@@ -93,6 +105,12 @@
fieldValue.reset();
listBuilder.write(fieldValue.getDataOutput(), true);
externalRecordBuilder.addField(MetadataRecordTypes.EXTERNAL_DETAILS_ARECORD_PROPERTIES_FIELD_INDEX, fieldValue);
+
+ //write field 2
+ fieldValue.reset();
+ aString.setValue(getNodeGroupName());
+ stringSerde.serialize(aString, fieldValue.getDataOutput());
+ externalRecordBuilder.addField(MetadataRecordTypes.EXTERNAL_DETAILS_ARECORD_GROUPNAME_FIELD_INDEX, fieldValue);
try {
externalRecordBuilder.write(out, true);
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalFile.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalFile.java
new file mode 100644
index 0000000..0128783
--- /dev/null
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entities/ExternalFile.java
@@ -0,0 +1,110 @@
+package edu.uci.ics.asterix.metadata.entities;
+
+import java.util.Date;
+
+import edu.uci.ics.asterix.metadata.MetadataCache;
+import edu.uci.ics.asterix.metadata.api.IMetadataEntity;
+
+public class ExternalFile implements IMetadataEntity{
+
+ /**
+ * A class for metadata entity externalFile
+ * This class represents an external dataset file and is intended for use with external data indexes
+ */
+ private static final long serialVersionUID = 1L;
+
+ private String dataverseName;
+ private String datasetName;
+ private Date lastModefiedTime;
+ private long size;
+ private String fileName;
+ private int fileNumber;
+
+
+ public ExternalFile(String dataverseName, String datasetName, Date lastModefiedTime, long size, String fileName,
+ int fileNumber) {
+ this.dataverseName = dataverseName;
+ this.datasetName = datasetName;
+ this.lastModefiedTime = lastModefiedTime;
+ this.size = size;
+ this.fileName = fileName;
+ this.fileNumber = fileNumber;
+ }
+
+ public String getDataverseName() {
+ return dataverseName;
+ }
+
+ public void setDataverseName(String dataverseName) {
+ this.dataverseName = dataverseName;
+ }
+
+ public String getDatasetName() {
+ return datasetName;
+ }
+
+ public void setDatasetName(String datasetName) {
+ this.datasetName = datasetName;
+ }
+ public Date getLastModefiedTime() {
+ return lastModefiedTime;
+ }
+
+ public void setLastModefiedTime(Date lastModefiedTime) {
+ this.lastModefiedTime = lastModefiedTime;
+ }
+
+ public long getSize() {
+ return size;
+ }
+
+ public void setSize(long size) {
+ this.size = size;
+ }
+
+ public String getFileName() {
+ return fileName;
+ }
+
+ public void setFileName(String fileName) {
+ this.fileName = fileName;
+ }
+
+ public int getFileNumber() {
+ return fileNumber;
+ }
+
+ public void setFileNumber(int fileNumber) {
+ this.fileNumber = fileNumber;
+ }
+
+ @Override
+ public Object addToCache(MetadataCache cache) {
+ //return cache.addExternalFileIfNotExists(this);
+ return null;
+ }
+
+ @Override
+ public Object dropFromCache(MetadataCache cache) {
+ //cache.dropExternalFile(this);
+ return null;
+ }
+
+ @Override
+ public boolean equals(Object obj)
+ {
+ if (obj == null)
+ return false;
+ if (obj == this)
+ return true;
+ if (!(obj instanceof ExternalFile))
+ return false;
+ ExternalFile anotherFile = (ExternalFile) obj;
+ if(fileNumber != anotherFile.fileNumber)
+ return false;
+ if(!fileName.equals(anotherFile.fileName))
+ return false;
+ return true;
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/DatasetTupleTranslator.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/DatasetTupleTranslator.java
index 61f856a..d346a6c 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/DatasetTupleTranslator.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/DatasetTupleTranslator.java
@@ -219,7 +219,10 @@
.getStringValue();
properties.put(key, value);
}
- datasetDetails = new ExternalDatasetDetails(adapter, properties);
+ String groupName = ((AString) datasetDetailsRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_DETAILS_ARECORD_GROUPNAME_FIELD_INDEX ))
+ .getStringValue();
+ datasetDetails = new ExternalDatasetDetails(adapter, properties,groupName);
}
Map<String, String> hints = getDatasetHints(datasetRecord);
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/ExternalFileTupleTranslator.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/ExternalFileTupleTranslator.java
new file mode 100644
index 0000000..6837c72
--- /dev/null
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/entitytupletranslators/ExternalFileTupleTranslator.java
@@ -0,0 +1,158 @@
+package edu.uci.ics.asterix.metadata.entitytupletranslators;
+
+import java.io.ByteArrayInputStream;
+import java.io.DataInput;
+import java.io.DataInputStream;
+import java.io.IOException;
+import java.util.Date;
+
+import edu.uci.ics.asterix.common.exceptions.AsterixException;
+import edu.uci.ics.asterix.formats.nontagged.AqlSerializerDeserializerProvider;
+import edu.uci.ics.asterix.metadata.MetadataException;
+import edu.uci.ics.asterix.metadata.bootstrap.MetadataPrimaryIndexes;
+import edu.uci.ics.asterix.metadata.bootstrap.MetadataRecordTypes;
+import edu.uci.ics.asterix.metadata.entities.ExternalFile;
+import edu.uci.ics.asterix.om.base.ADateTime;
+import edu.uci.ics.asterix.om.base.AInt32;
+import edu.uci.ics.asterix.om.base.AInt64;
+import edu.uci.ics.asterix.om.base.AMutableDateTime;
+import edu.uci.ics.asterix.om.base.AMutableInt32;
+import edu.uci.ics.asterix.om.base.AMutableInt64;
+import edu.uci.ics.asterix.om.base.ARecord;
+import edu.uci.ics.asterix.om.base.AString;
+import edu.uci.ics.asterix.om.types.BuiltinType;
+import edu.uci.ics.hyracks.api.dataflow.value.ISerializerDeserializer;
+import edu.uci.ics.hyracks.dataflow.common.data.accessors.ITupleReference;
+
+public class ExternalFileTupleTranslator extends AbstractTupleTranslator<ExternalFile>{
+
+ // Field indexes of serialized ExternalFile in a tuple.
+ // First key field.
+ public static final int EXTERNAL_FILE_DATAVERSENAME_TUPLE_FIELD_INDEX = 0;
+ // Second key field.
+ public static final int EXTERNAL_FILE_DATASETNAME_TUPLE_FIELD_INDEX = 1;
+ // Third key field
+ public static final int EXTERNAL_FILE_NUMBER_TUPLE_FIELD_INDEX = 2;
+ // Payload field containing serialized ExternalFile.
+ public static final int EXTERNAL_FILE_PAYLOAD_TUPLE_FIELD_INDEX = 3;
+
+ protected AMutableInt32 aInt32 = new AMutableInt32(0);
+ protected AMutableDateTime aDateTime = new AMutableDateTime(0);
+ protected AMutableInt64 aInt64 = new AMutableInt64(0);
+
+ @SuppressWarnings("unchecked")
+ protected ISerializerDeserializer<AInt32> intSerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(BuiltinType.AINT32);
+ @SuppressWarnings("unchecked")
+ protected ISerializerDeserializer<ADateTime> dateTimeSerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(BuiltinType.ADATETIME);
+ @SuppressWarnings("unchecked")
+ protected ISerializerDeserializer<AInt64> longSerde = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(BuiltinType.AINT64);
+ @SuppressWarnings("unchecked")
+ private ISerializerDeserializer<ARecord> recordSerDes = AqlSerializerDeserializerProvider.INSTANCE
+ .getSerializerDeserializer(MetadataRecordTypes.EXTERNAL_FILE_RECORDTYPE);
+
+ public ExternalFileTupleTranslator(boolean getTuple) {
+ super(getTuple, MetadataPrimaryIndexes.EXTERNAL_FILE_DATASET.getFieldCount());
+ }
+
+ @Override
+ public ExternalFile getMetadataEntytiFromTuple(ITupleReference tuple)
+ throws MetadataException, IOException {
+ byte[] serRecord = tuple.getFieldData(EXTERNAL_FILE_PAYLOAD_TUPLE_FIELD_INDEX);
+ int recordStartOffset = tuple.getFieldStart(EXTERNAL_FILE_PAYLOAD_TUPLE_FIELD_INDEX);
+ int recordLength = tuple.getFieldLength(EXTERNAL_FILE_PAYLOAD_TUPLE_FIELD_INDEX);
+ ByteArrayInputStream stream = new ByteArrayInputStream(serRecord, recordStartOffset, recordLength);
+ DataInput in = new DataInputStream(stream);
+ ARecord externalFileRecord = (ARecord) recordSerDes.deserialize(in);
+ return createExternalFileFromARecord(externalFileRecord);
+ }
+
+ private ExternalFile createExternalFileFromARecord(ARecord externalFileRecord) {
+ String dataverseName = ((AString) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_DATAVERSENAME_FIELD_INDEX))
+ .getStringValue();
+ String datasetName = ((AString) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_DATASET_NAME_FIELD_INDEX)).getStringValue();
+ String FileName = ((AString) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_NAME_FIELD_INDEX)).getStringValue();
+ int fileNumber = ((AInt32) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_NUMBER_FIELD_INDEX)).getIntegerValue();
+ Date lastMoDifiedDate = new Date(((ADateTime) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_MOD_DATE_FIELD_INDEX)).getChrononTime());
+ long fileSize = ((AInt64) externalFileRecord
+ .getValueByPos(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_SIZE_FIELD_INDEX)).getLongValue();
+
+ return new ExternalFile(dataverseName, datasetName,lastMoDifiedDate,fileSize,FileName,fileNumber);
+ }
+
+ @Override
+ public ITupleReference getTupleFromMetadataEntity(
+ ExternalFile externalFile) throws MetadataException, IOException {
+ // write the key in the first 3 fields of the tuple
+ tupleBuilder.reset();
+ //dataverse name
+ aString.setValue(externalFile.getDataverseName());
+ stringSerde.serialize(aString, tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+ //dataset name
+ aString.setValue(externalFile.getDatasetName());
+ stringSerde.serialize(aString, tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+ //file number
+ aInt32.setValue(externalFile.getFileNumber());
+ intSerde.serialize(aInt32, tupleBuilder.getDataOutput());
+ tupleBuilder.addFieldEndOffset();
+
+ // write the pay-load in the fourth field of the tuple
+ recordBuilder.reset(MetadataRecordTypes.EXTERNAL_FILE_RECORDTYPE);
+
+ // write field 0
+ fieldValue.reset();
+ aString.setValue(externalFile.getDataverseName());
+ stringSerde.serialize(aString, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_DATAVERSENAME_FIELD_INDEX, fieldValue);
+
+ // write field 1
+ fieldValue.reset();
+ aString.setValue(externalFile.getDatasetName());
+ stringSerde.serialize(aString, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_DATASET_NAME_FIELD_INDEX, fieldValue);
+
+ // write field 2
+ fieldValue.reset();
+ aInt32.setValue(externalFile.getFileNumber());
+ intSerde.serialize(aInt32, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_NUMBER_FIELD_INDEX, fieldValue);
+
+ // write field 3
+ fieldValue.reset();
+ aString.setValue(externalFile.getFileName());
+ stringSerde.serialize(aString, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_NAME_FIELD_INDEX, fieldValue);
+
+ // write field 4
+ fieldValue.reset();
+ aInt64.setValue(externalFile.getSize());
+ longSerde.serialize(aInt64, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_SIZE_FIELD_INDEX, fieldValue);
+
+ //write field 5
+ fieldValue.reset();
+ aDateTime.setValue(externalFile.getLastModefiedTime().getTime());
+ dateTimeSerde.serialize(aDateTime, fieldValue.getDataOutput());
+ recordBuilder.addField(MetadataRecordTypes.EXTERNAL_FILE_ARECORD_FILE_MOD_DATE_FIELD_INDEX, fieldValue);
+
+ // write record
+ try {
+ recordBuilder.write(tupleBuilder.getDataOutput(), true);
+ } catch (AsterixException e) {
+ throw new MetadataException(e);
+ }
+ tupleBuilder.addFieldEndOffset();
+
+ tuple.reset(tupleBuilder.getFieldEndOffsets(), tupleBuilder.getByteArray());
+ return tuple;
+ }
+}
\ No newline at end of file
diff --git a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/utils/DatasetUtils.java b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/utils/DatasetUtils.java
index 820f277..95d26d9 100644
--- a/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/utils/DatasetUtils.java
+++ b/asterix-metadata/src/main/java/edu/uci/ics/asterix/metadata/utils/DatasetUtils.java
@@ -20,9 +20,12 @@
import edu.uci.ics.asterix.common.config.DatasetConfig.DatasetType;
import edu.uci.ics.asterix.formats.nontagged.AqlTypeTraitProvider;
+import edu.uci.ics.asterix.metadata.declared.AqlMetadataProvider;
import edu.uci.ics.asterix.metadata.entities.Dataset;
+import edu.uci.ics.asterix.metadata.entities.ExternalDatasetDetails;
import edu.uci.ics.asterix.metadata.entities.InternalDatasetDetails;
import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.asterix.om.types.BuiltinType;
import edu.uci.ics.asterix.om.types.IAType;
import edu.uci.ics.hyracks.algebricks.common.exceptions.AlgebricksException;
import edu.uci.ics.hyracks.algebricks.data.IBinaryComparatorFactoryProvider;
@@ -32,6 +35,8 @@
import edu.uci.ics.hyracks.api.dataflow.value.ITypeTraits;
public class DatasetUtils {
+ public static final String KEY_INPUT_FORMAT = "input-format";
+ public static final String INPUT_FORMAT_RC = "rc-input-format";
public static IBinaryComparatorFactory[] computeKeysBinaryComparatorFactories(Dataset dataset,
ARecordType itemType, IBinaryComparatorFactoryProvider comparatorFactoryProvider)
throws AlgebricksException {
@@ -82,6 +87,46 @@
}
return bhffs;
}
+
+ public static IBinaryHashFunctionFactory[] computeExternalDataKeysBinaryHashFunFactories(Dataset dataset,
+ IBinaryHashFunctionFactoryProvider hashFunProvider) throws AlgebricksException {
+ if (dataset.getDatasetType() != DatasetType.EXTERNAL) {
+ throw new AlgebricksException("not implemented");
+ }
+ //get dataset details
+ ExternalDatasetDetails edd = (ExternalDatasetDetails) dataset.getDatasetDetails();
+ if (edd.getProperties().get(KEY_INPUT_FORMAT).trim().equals(INPUT_FORMAT_RC))
+ {
+ //RID: <String(File name) OR Int32(File number), Int64(Block byte location), Int32(row number)>
+ IBinaryHashFunctionFactory[] bhffs = new IBinaryHashFunctionFactory[3];
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ bhffs[0] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.AINT32);
+ }
+ else
+ {
+ bhffs[0] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.ASTRING);
+ }
+ bhffs[1] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.AINT64);
+ bhffs[2] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.AINT32);
+ return bhffs;
+ }
+ else
+ {
+ //RID: <String(File name) OR Int32(File number), Int64(Record byte location)>
+ IBinaryHashFunctionFactory[] bhffs = new IBinaryHashFunctionFactory[2];
+ if(AqlMetadataProvider.isOptimizeExternalIndexes())
+ {
+ bhffs[0] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.AINT32);
+ }
+ else
+ {
+ bhffs[0] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.ASTRING);
+ }
+ bhffs[1] = hashFunProvider.getBinaryHashFunctionFactory(BuiltinType.AINT64);
+ return bhffs;
+ }
+ }
public static ITypeTraits[] computeTupleTypeTraits(Dataset dataset, ARecordType itemType)
throws AlgebricksException {
@@ -112,6 +157,17 @@
return (((InternalDatasetDetails) dataset.getDatasetDetails())).getNodeGroupName();
}
+ public static int getExternalRIDSize(Dataset dataset) {
+ ExternalDatasetDetails dsd = ((ExternalDatasetDetails) dataset.getDatasetDetails());
+ if (dsd.getProperties().get(KEY_INPUT_FORMAT).equals(INPUT_FORMAT_RC))
+ {
+ return 3;
+ }
+ else{
+ return 2;
+ }
+ }
+
public static int getPositionOfPartitioningKeyField(Dataset dataset, String fieldExpr) {
List<String> partitioningKeys = DatasetUtils.getPartitioningKeys(dataset);
for (int i = 0; i < partitioningKeys.size(); i++) {
diff --git a/asterix-om/src/main/java/edu/uci/ics/asterix/om/base/AInt64.java b/asterix-om/src/main/java/edu/uci/ics/asterix/om/base/AInt64.java
index cc3fd04..00b7459 100644
--- a/asterix-om/src/main/java/edu/uci/ics/asterix/om/base/AInt64.java
+++ b/asterix-om/src/main/java/edu/uci/ics/asterix/om/base/AInt64.java
@@ -33,6 +33,10 @@
public long getLongValue() {
return value;
}
+
+ public void setValue(long value) {
+ this.value = value;
+ }
@Override
public IAType getType() {
diff --git a/asterix-om/src/main/java/edu/uci/ics/asterix/om/functions/AsterixBuiltinFunctions.java b/asterix-om/src/main/java/edu/uci/ics/asterix/om/functions/AsterixBuiltinFunctions.java
index 3c8a0c4..b6bdb4b 100644
--- a/asterix-om/src/main/java/edu/uci/ics/asterix/om/functions/AsterixBuiltinFunctions.java
+++ b/asterix-om/src/main/java/edu/uci/ics/asterix/om/functions/AsterixBuiltinFunctions.java
@@ -243,6 +243,8 @@
public final static FunctionIdentifier INDEX_SEARCH = new FunctionIdentifier(FunctionConstants.ASTERIX_NS,
"index-search", FunctionIdentifier.VARARGS);
+ public final static FunctionIdentifier EXTERNAL_ACCESS_BY_RID = new FunctionIdentifier(FunctionConstants.ASTERIX_NS,
+ "external-access-by-rid", FunctionIdentifier.VARARGS);
public final static FunctionIdentifier MAKE_FIELD_INDEX_HANDLE = new FunctionIdentifier(
FunctionConstants.ASTERIX_NS, "make-field-index-handle", 2);
@@ -672,6 +674,14 @@
return BuiltinType.ANY; // TODO
}
});
+ addPrivateFunction(EXTERNAL_ACCESS_BY_RID, new IResultTypeComputer() {
+
+ @Override
+ public IAType computeType(ILogicalExpression expression, IVariableTypeEnvironment env,
+ IMetadataProvider<?, ?> mp) throws AlgebricksException {
+ return BuiltinType.ANY;
+ }
+ });
addFunction(INT8_CONSTRUCTOR, OptionalAInt8TypeComputer.INSTANCE);
addFunction(INT16_CONSTRUCTOR, OptionalAInt16TypeComputer.INSTANCE);
addFunction(INT32_CONSTRUCTOR, OptionalAInt32TypeComputer.INSTANCE);
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ADMDataParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ADMDataParser.java
index b8c8659..e08c509 100644
--- a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ADMDataParser.java
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ADMDataParser.java
@@ -92,6 +92,16 @@
throw new AsterixException(e);
}
}
+
+ @Override
+ public void close() throws IOException{
+ admLexer.close();
+ }
+
+ @Override
+ public void reset(){
+ admLexer.reset();
+ }
protected boolean parseAdmInstance(IAType objectType, boolean datasetRec, DataOutput out) throws AsterixException,
IOException {
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractControlledTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractControlledTupleParser.java
new file mode 100644
index 0000000..a4f1691
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractControlledTupleParser.java
@@ -0,0 +1,123 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+import java.io.DataOutput;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.ByteBuffer;
+import edu.uci.ics.asterix.common.exceptions.AsterixException;
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.api.comm.IFrameWriter;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.common.comm.io.ArrayTupleBuilder;
+import edu.uci.ics.hyracks.dataflow.common.comm.io.FrameTupleAppender;
+import edu.uci.ics.hyracks.dataflow.common.comm.util.FrameUtils;
+
+/**
+ * An Abstract class implementation for IControlledTupleParser. It provides common
+ * functionality involved in parsing data in an external format in a pipelined manner and packing
+ * frames with formed tuples.
+ * (DONE)
+ */
+public abstract class AbstractControlledTupleParser extends ControlledTupleParser{
+
+ protected ArrayTupleBuilder tb = new ArrayTupleBuilder(1);
+ protected transient DataOutput dos;
+ protected final FrameTupleAppender appender;
+ protected final ByteBuffer frame;
+ protected final ARecordType recType;
+ protected final IHyracksTaskContext ctx;
+ protected IDataParser parser;
+
+ public AbstractControlledTupleParser(IHyracksTaskContext ctx, ARecordType recType) throws HyracksDataException {
+ appender = new FrameTupleAppender(ctx.getFrameSize());
+ frame = ctx.allocateFrame();
+ this.recType = recType;
+ this.ctx = ctx;
+ dos = tb.getDataOutput();
+ }
+
+ public abstract IDataParser getDataParser();
+
+ @Override
+ public void parse(InputStream in, IFrameWriter writer)
+ throws HyracksDataException {
+ //This function when used works as non-pipelined parser
+ //This whole parser interface needs revisiting.
+ appender.reset(frame, true);
+ parser = getDataParser();
+ try {
+ parser.initialize(in, recType, true);
+ while (true) {
+ tb.reset();
+ if (!parser.parse(tb.getDataOutput())) {
+ parser.reset();
+ break;
+ }
+ tb.addFieldEndOffset();
+ addTupleToFrame(writer);
+ }
+ parser.close();
+ if (appender.getTupleCount() > 0) {
+ FrameUtils.flushFrame(frame, writer);
+ }
+ } catch (Exception e) {
+ throw new HyracksDataException("Failed to initialize data parser");
+ }
+ }
+
+ @Override
+ public void initialize(InputStream in) throws HyracksDataException {
+ appender.reset(frame, true);
+ parser = getDataParser();
+ try {
+ parser.initialize(in, recType, true);
+
+ } catch (Exception e) {
+ throw new HyracksDataException("Failed to initialize data parser");
+ }
+ }
+
+ @Override
+ public void parseNext(IFrameWriter writer) throws HyracksDataException {
+ try {
+ while (true) {
+ tb.reset();
+ if (!parser.parse(tb.getDataOutput())) {
+ parser.reset();
+ break;
+ }
+ tb.addFieldEndOffset();
+ addTupleToFrame(writer);
+ }
+ } catch (AsterixException ae) {
+ throw new HyracksDataException(ae);
+ } catch (IOException ioe) {
+ throw new HyracksDataException(ioe);
+ }
+ }
+
+ @Override
+ public void close(IFrameWriter writer) throws HyracksDataException {
+ try{
+ parser.close();
+ if (appender.getTupleCount() > 0) {
+ FrameUtils.flushFrame(frame, writer);
+ }
+ } catch (IOException ioe) {
+ throw new HyracksDataException(ioe);
+ }
+ }
+
+ protected void addTupleToFrame(IFrameWriter writer) throws HyracksDataException {
+ if (!appender.append(tb.getFieldEndOffsets(), tb.getByteArray(), 0, tb.getSize())) {
+ FrameUtils.flushFrame(frame, writer);
+ appender.reset(frame, true);
+ if (!appender.append(tb.getFieldEndOffsets(), tb.getByteArray(), 0, tb.getSize())) {
+ throw new IllegalStateException();
+ }
+ }
+
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractTupleParser.java
index 2322338..78159f5 100644
--- a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractTupleParser.java
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AbstractTupleParser.java
@@ -62,6 +62,7 @@
while (true) {
tb.reset();
if (!parser.parse(tb.getDataOutput())) {
+ parser.close();
break;
}
tb.addFieldEndOffset();
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AdmTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AdmTupleParser.java
index 7cb9bb0..34f75e6 100644
--- a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AdmTupleParser.java
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/AdmTupleParser.java
@@ -20,7 +20,7 @@
/**
* An extension of AbstractTupleParser that provides functionality for
- * parsing delimited files.
+ * parsing adm formatted input files.
*/
public class AdmTupleParser extends AbstractTupleParser {
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParser.java
new file mode 100644
index 0000000..aa2d9ba
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParser.java
@@ -0,0 +1,23 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+
+/**
+ * An extension of AbstractControlledTupleParser that provides functionality for
+ * parsing Adm formatted input.
+ */
+public class ControlledADMTupleParser extends AbstractControlledTupleParser{
+
+ public ControlledADMTupleParser(IHyracksTaskContext ctx, ARecordType recType)
+ throws HyracksDataException {
+ super(ctx, recType);
+ }
+
+ @Override
+ public IDataParser getDataParser() {
+ return new ADMDataParser();
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParserFactory.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParserFactory.java
new file mode 100644
index 0000000..d72366c
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledADMTupleParserFactory.java
@@ -0,0 +1,28 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParser;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParserFactory;
+
+/**
+ * A Controlled tuple parser factory for creating a tuple parser capable of parsing
+ * ADM data.
+ */
+public class ControlledADMTupleParserFactory implements ITupleParserFactory{
+ private static final long serialVersionUID = 1L;
+
+ protected ARecordType recType;
+
+ public ControlledADMTupleParserFactory(ARecordType recType){
+ this.recType = recType;
+ }
+
+ @Override
+ public ITupleParser createTupleParser(IHyracksTaskContext ctx)
+ throws HyracksDataException {
+ return new ControlledADMTupleParser(ctx, recType);
+ }
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParser.java
new file mode 100644
index 0000000..15643d4
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParser.java
@@ -0,0 +1,23 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.common.data.parsers.IValueParserFactory;
+
+public class ControlledDelimitedDataTupleParser extends AbstractControlledTupleParser{
+
+ private final DelimitedDataParser dataParser;
+
+ public ControlledDelimitedDataTupleParser(IHyracksTaskContext ctx,
+ ARecordType recType, IValueParserFactory[] valueParserFactories, char fieldDelimter) throws HyracksDataException {
+ super(ctx, recType);
+ dataParser = new DelimitedDataParser(recType, valueParserFactories, fieldDelimter);
+ }
+
+ @Override
+ public IDataParser getDataParser() {
+ return dataParser;
+ }
+
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParserFactory.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParserFactory.java
new file mode 100644
index 0000000..ced33ef
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledDelimitedDataTupleParserFactory.java
@@ -0,0 +1,29 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+import edu.uci.ics.asterix.om.types.ARecordType;
+import edu.uci.ics.hyracks.api.context.IHyracksTaskContext;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.common.data.parsers.IValueParserFactory;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParser;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParserFactory;
+
+public class ControlledDelimitedDataTupleParserFactory implements ITupleParserFactory{
+ private static final long serialVersionUID = 1L;
+ private IValueParserFactory[] valueParserFactories;
+ private char fieldDelimiter;
+ protected ARecordType recordType;
+
+
+ public ControlledDelimitedDataTupleParserFactory(ARecordType recordType, IValueParserFactory[] fieldParserFactories, char fieldDelimiter) {
+ this.recordType = recordType;
+ this.valueParserFactories = fieldParserFactories;
+ this.fieldDelimiter = fieldDelimiter;
+ }
+
+
+ @Override
+ public ITupleParser createTupleParser(IHyracksTaskContext ctx)
+ throws HyracksDataException {
+ return new ControlledDelimitedDataTupleParser(ctx, recordType, valueParserFactories, fieldDelimiter);
+ }
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledTupleParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledTupleParser.java
new file mode 100644
index 0000000..10b09f5
--- /dev/null
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/ControlledTupleParser.java
@@ -0,0 +1,29 @@
+package edu.uci.ics.asterix.runtime.operators.file;
+
+import java.io.InputStream;
+
+import edu.uci.ics.hyracks.api.comm.IFrameWriter;
+import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
+import edu.uci.ics.hyracks.dataflow.std.file.ITupleParser;
+
+/**
+ * This interface is to be implemented by parsers used in a pipelined hyracks job where input is not ready all at once
+ */
+public abstract class ControlledTupleParser implements ITupleParser{
+
+ /**
+ * This function associate an input stream with the parser
+ */
+ public abstract void initialize(InputStream in) throws HyracksDataException;
+
+ /**
+ * This function should flush the tuples setting in the frame writer buffer
+ * and free all resources
+ */
+ public abstract void close(IFrameWriter writer) throws HyracksDataException;
+
+ /**
+ * This function is called when there are more data ready for parsing in the input stream
+ */
+ public abstract void parseNext(IFrameWriter writer) throws HyracksDataException;
+}
\ No newline at end of file
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/DelimitedDataParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/DelimitedDataParser.java
index 5a639dc..22c43ec 100644
--- a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/DelimitedDataParser.java
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/DelimitedDataParser.java
@@ -100,6 +100,17 @@
}
@Override
+ public void close() {
+ cursor.close();
+ }
+
+ @Override
+ public void reset()
+ {
+ cursor.reset();
+ }
+
+ @Override
public boolean parse(DataOutput out) throws AsterixException, IOException {
while (cursor.nextRecord()) {
recBuilder.reset(recordType);
@@ -165,7 +176,7 @@
private static final int INITIAL_BUFFER_SIZE = 4096;
private static final int INCREMENT = 4096;
- private final Reader in;
+ private Reader in;
private char[] buffer;
private int start;
@@ -182,6 +193,21 @@
end = 0;
state = State.INIT;
}
+
+ public void close(){
+ try {
+ in.close();
+ } catch (IOException e) {
+ // TODO Auto-generated catch block
+ e.printStackTrace();
+ }
+ }
+
+ public void reset(){
+ start = 0;
+ end = 0;
+ state = State.INIT;
+ }
public boolean nextRecord() throws IOException {
while (true) {
diff --git a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/IDataParser.java b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/IDataParser.java
index 46d281b..cd9ae2e 100644
--- a/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/IDataParser.java
+++ b/asterix-runtime/src/main/java/edu/uci/ics/asterix/runtime/operators/file/IDataParser.java
@@ -52,4 +52,17 @@
* @throws IOException
*/
public boolean parse(DataOutput out) throws AsterixException, IOException;
+
+ /**
+ * Close the underlying inputStream object.
+ *
+ */
+ public void close() throws IOException;
+
+ /**
+ * Reset the parser before processing a new patch of input in the input stream
+ */
+ public void reset();
}
+
+
diff --git a/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapter.java b/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapter.java
index e037ec6..ffd9edd 100644
--- a/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapter.java
+++ b/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapter.java
@@ -224,7 +224,8 @@
while (continueIngestion) {
tb.reset();
if (!parser.parse(tb.getDataOutput())) {
- break;
+ parser.close();
+ break;
}
tb.addFieldEndOffset();
if (delayConfigured) {
diff --git a/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapterFactory.java b/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapterFactory.java
index bf3c086..c576cf8 100644
--- a/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapterFactory.java
+++ b/asterix-tools/src/main/java/edu/uci/ics/asterix/tools/external/data/RateControlledFileSystemBasedAdapterFactory.java
@@ -14,14 +14,17 @@
*/
package edu.uci.ics.asterix.tools.external.data;
+import java.util.HashMap;
import java.util.Map;
import edu.uci.ics.asterix.common.exceptions.AsterixException;
import edu.uci.ics.asterix.external.adapter.factory.IGenericDatasetAdapterFactory;
import edu.uci.ics.asterix.external.dataset.adapter.FileSystemBasedAdapter;
+import edu.uci.ics.asterix.external.dataset.adapter.IControlledAdapter;
import edu.uci.ics.asterix.external.dataset.adapter.IDatasourceAdapter;
import edu.uci.ics.asterix.om.types.ARecordType;
import edu.uci.ics.asterix.om.types.IAType;
+import edu.uci.ics.hyracks.algebricks.common.exceptions.NotImplementedException;
/**
* Factory class for creating @see{RateControllerFileSystemBasedAdapter} The
@@ -67,6 +70,12 @@
public String getName() {
return "file_feed";
}
+
+ @Override
+ public IDatasourceAdapter createIndexingAdapter(
+ Map<String, Object> configuration, IAType atype, Map<String,Integer> files) throws Exception {
+ throw new NotImplementedException("Rate Controlled Indexing Adapter is not implemented for feeds");
+ }
private void checkRequiredArgs(Map<String, Object> configuration) throws Exception {
if (configuration.get(KEY_FILE_SYSTEM) == null) {
@@ -83,4 +92,10 @@
}
}
+ @Override
+ public IControlledAdapter createAccessByRIDAdapter(
+ Map<String, Object> configuration, IAType atype, HashMap<Integer,String> files) throws Exception {
+ throw new NotImplementedException("Rate Controlled Access by RID Adapter is not implemented for feeds");
+ }
+
}
\ No newline at end of file
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/opcallbacks/AbstractIndexModificationOperationCallback.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/opcallbacks/AbstractIndexModificationOperationCallback.java
index dac3e95..031a26e 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/opcallbacks/AbstractIndexModificationOperationCallback.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/opcallbacks/AbstractIndexModificationOperationCallback.java
@@ -58,6 +58,9 @@
protected void log(int PKHash, ITupleReference newValue, IndexOperation oldOp, ITupleReference oldValue)
throws ACIDException {
logRecord.setPKHashValue(PKHash);
+ logRecord.setPKFields(primaryKeyFields);
+ logRecord.setPKValue(newValue);
+ logRecord.computeAndSetPKValueSize();
if (newValue != null) {
logRecord.setNewValueSize(tupleWriter.bytesRequired(newValue));
logRecord.setNewValue(newValue);
@@ -73,7 +76,7 @@
logRecord.setOldValueSize(0);
}
}
- logRecord.setUpdateLogSize();
+ logRecord.computeAndSetLogSize();
txnSubsystem.getLogManager().log(logRecord);
}
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/locking/LockManager.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/locking/LockManager.java
index e0baa3f..6d86f70 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/locking/LockManager.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/locking/LockManager.java
@@ -29,7 +29,6 @@
import edu.uci.ics.asterix.common.exceptions.ACIDException;
import edu.uci.ics.asterix.common.transactions.DatasetId;
import edu.uci.ics.asterix.common.transactions.ILockManager;
-import edu.uci.ics.asterix.common.transactions.ILogRecord;
import edu.uci.ics.asterix.common.transactions.ITransactionContext;
import edu.uci.ics.asterix.common.transactions.ITransactionManager;
import edu.uci.ics.asterix.common.transactions.JobId;
@@ -86,7 +85,6 @@
private LockRequestTracker lockRequestTracker; //for debugging
private ConsecutiveWakeupContext consecutiveWakeupContext;
- private final ILogRecord logRecord;
public LockManager(TransactionSubsystem txnSubsystem) throws ACIDException {
this.txnSubsystem = txnSubsystem;
@@ -104,7 +102,6 @@
this.tempDatasetIdObj = new DatasetId(0);
this.tempJobIdObj = new JobId(0);
this.consecutiveWakeupContext = new ConsecutiveWakeupContext();
- this.logRecord = new LogRecord();
if (IS_DEBUG_MODE) {
this.lockRequestTracker = new LockRequestTracker();
}
@@ -642,22 +639,16 @@
@Override
public void unlock(DatasetId datasetId, int entityHashValue, ITransactionContext txnContext) throws ACIDException {
- internalUnlock(datasetId, entityHashValue, txnContext, false, false);
- }
-
- @Override
- public void unlock(DatasetId datasetId, int entityHashValue, ITransactionContext txnContext, boolean commitFlag)
- throws ACIDException {
- internalUnlock(datasetId, entityHashValue, txnContext, false, commitFlag);
+ internalUnlock(datasetId, entityHashValue, txnContext, false);
}
private void instantUnlock(DatasetId datasetId, int entityHashValue, ITransactionContext txnContext)
throws ACIDException {
- internalUnlock(datasetId, entityHashValue, txnContext, true, false);
+ internalUnlock(datasetId, entityHashValue, txnContext, true);
}
private void internalUnlock(DatasetId datasetId, int entityHashValue, ITransactionContext txnContext,
- boolean isInstant, boolean commitFlag) throws ACIDException {
+ boolean isInstant) throws ACIDException {
JobId jobId = txnContext.getJobId();
int eLockInfo = -1;
DatasetLockInfo dLockInfo = null;
@@ -701,22 +692,6 @@
+ "," + entityHashValue + "]: Corresponding lock info doesn't exist.");
}
- //////////////////////////////////////////////////////////
- //[Notice]
- //If both EntityLockCount and DatasetLockCount are 1,
- //then write entity-commit log and return without releasing the lock.
- //The lock will be released when the entity-commit log is flushed.
- if (commitFlag && entityInfoManager.getEntityLockCount(entityInfo) == 1
- && entityInfoManager.getDatasetLockCount(entityInfo) == 1) {
- if (txnContext.isWriteTxn()) {
- logRecord.formCommitLogRecord(txnContext, LogType.ENTITY_COMMIT, jobId.getId(), datasetId.getId(),
- entityHashValue);
- txnSubsystem.getLogManager().log(logRecord);
- }
- return;
- }
- //////////////////////////////////////////////////////////
-
datasetLockMode = entityInfoManager.getDatasetLockMode(entityInfo) == LockMode.S ? LockMode.IS
: LockMode.IX;
@@ -758,11 +733,6 @@
waiterObjId = waiterObj.getNextWaiterObjId();
}
if (threadCount == 0) {
- if (entityInfoManager.getEntityLockMode(entityInfo) == LockMode.X) {
- //TODO
- //write a commit log for the unlocked resource
- //need to figure out that instantLock() also needs to write a commit log.
- }
entityInfoManager.deallocate(entityInfo);
}
}
@@ -2243,17 +2213,11 @@
tempDatasetIdObj.setId(logRecord.getDatasetId());
tempJobIdObj.setId(logRecord.getJobId());
txnCtx = txnSubsystem.getTransactionManager().getTransactionContext(tempJobIdObj);
- if (txnCtx == null) {
- throw new IllegalStateException("TransactionContext[" + tempJobIdObj + "] doesn't exist.");
- }
unlock(tempDatasetIdObj, logRecord.getPKHashValue(), txnCtx);
txnCtx.notifyOptracker(false);
} else if (logRecord.getLogType() == LogType.JOB_COMMIT) {
tempJobIdObj.setId(logRecord.getJobId());
txnCtx = txnSubsystem.getTransactionManager().getTransactionContext(tempJobIdObj);
- if (txnCtx == null) {
- throw new IllegalStateException("TransactionContext[" + tempJobIdObj + "] doesn't exist.");
- }
txnCtx.notifyOptracker(true);
((LogPage) logPage).notifyJobCommitter();
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogManager.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogManager.java
index 7fce48d..4f0bb59 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogManager.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogManager.java
@@ -24,7 +24,11 @@
import java.util.Collections;
import java.util.Comparator;
import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.Future;
import java.util.concurrent.LinkedBlockingQueue;
+import java.util.concurrent.atomic.AtomicBoolean;
import java.util.logging.Level;
import java.util.logging.Logger;
@@ -58,6 +62,7 @@
private FileChannel appendChannel;
private LogPage appendPage;
private LogFlusher logFlusher;
+ private Future<Object> futureLogFlusher;
public LogManager(TransactionSubsystem txnSubsystem) throws ACIDException {
this.txnSubsystem = txnSubsystem;
@@ -86,8 +91,7 @@
appendChannel = getFileChannel(appendLSN, false);
getAndInitNewPage();
logFlusher = new LogFlusher(this, emptyQ, flushQ);
- logFlusher.setDaemon(true);
- AsterixThreadExecutor.INSTANCE.execute(logFlusher);
+ futureLogFlusher = AsterixThreadExecutor.INSTANCE.submit(logFlusher);
}
@Override
@@ -174,6 +178,7 @@
@Override
public void stop(boolean dumpState, OutputStream os) {
+ terminateLogFlusher();
if (dumpState) {
// #. dump Configurable Variables
dumpConfVars(os);
@@ -267,15 +272,32 @@
}
private void terminateLogFlusher() {
+ if (LOGGER.isLoggable(Level.INFO)) {
+ LOGGER.info("Terminating LogFlusher thread ...");
+ }
logFlusher.terminate();
try {
- logFlusher.join();
- } catch (InterruptedException e) {
- throw new IllegalStateException(e);
+ futureLogFlusher.get();
+ } catch (ExecutionException | InterruptedException e) {
+ if (LOGGER.isLoggable(Level.INFO)) {
+ LOGGER.info("---------- warning(begin): LogFlusher thread is terminated abnormally --------");
+ e.printStackTrace();
+ LOGGER.info("---------- warning(end) : LogFlusher thread is terminated abnormally --------");
+ }
+ }
+ if (LOGGER.isLoggable(Level.INFO)) {
+ LOGGER.info("LogFlusher thread is terminated.");
}
}
private void deleteAllLogFiles() {
+ if (appendChannel != null) {
+ try {
+ appendChannel.close();
+ } catch (IOException e) {
+ throw new IllegalStateException("Failed to close a fileChannel of a log file");
+ }
+ }
List<Long> logFileIds = getLogFileIds();
for (Long id : logFileIds) {
File file = new File(getLogFilePath(id));
@@ -364,43 +386,69 @@
}
}
-class LogFlusher extends Thread {
- private final static LogPage POISON_PILL = new LogPage(null, ILogRecord.COMMIT_LOG_SIZE, null);
+class LogFlusher implements Callable<Boolean> {
+ private final static LogPage POISON_PILL = new LogPage(null, ILogRecord.JOB_COMMIT_LOG_SIZE, null);
private final LogManager logMgr;//for debugging
private final LinkedBlockingQueue<LogPage> emptyQ;
private final LinkedBlockingQueue<LogPage> flushQ;
private LogPage flushPage;
+ private final AtomicBoolean isStarted;
+ private final AtomicBoolean terminateFlag;
public LogFlusher(LogManager logMgr, LinkedBlockingQueue<LogPage> emptyQ, LinkedBlockingQueue<LogPage> flushQ) {
this.logMgr = logMgr;
this.emptyQ = emptyQ;
this.flushQ = flushQ;
flushPage = null;
+ isStarted = new AtomicBoolean(false);
+ terminateFlag = new AtomicBoolean(false);
+
}
public void terminate() {
+ //make sure the LogFlusher thread started before terminating it.
+ synchronized (isStarted) {
+ while(!isStarted.get()) {
+ try {
+ isStarted.wait();
+ } catch (InterruptedException e) {
+ //ignore
+ }
+ }
+ }
+
+ terminateFlag.set(true);
if (flushPage != null) {
synchronized (flushPage) {
flushPage.isStop(true);
flushPage.notify();
}
}
+ //[Notice]
+ //The return value doesn't need to be checked
+ //since terminateFlag will trigger termination if the flushQ is full.
flushQ.offer(POISON_PILL);
}
@Override
- public void run() {
+ public Boolean call() {
+ synchronized(isStarted) {
+ isStarted.set(true);
+ isStarted.notify();
+ }
while (true) {
flushPage = null;
try {
flushPage = flushQ.take();
- if (flushPage == POISON_PILL) {
- break;
+ if (flushPage == POISON_PILL || terminateFlag.get()) {
+ return true;
}
- flushPage.flush();
} catch (InterruptedException e) {
- //ignore
+ if (flushPage == null) {
+ continue;
+ }
}
+ flushPage.flush();
emptyQ.offer(flushPage);
}
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogPage.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogPage.java
index 45c3e65..edfec69 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogPage.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogPage.java
@@ -58,7 +58,7 @@
appendOffset = 0;
flushOffset = 0;
isLastPage = false;
- syncCommitQ = new LinkedBlockingQueue<ILogRecord>(logPageSize / ILogRecord.COMMIT_LOG_SIZE);
+ syncCommitQ = new LinkedBlockingQueue<ILogRecord>(logPageSize / ILogRecord.JOB_COMMIT_LOG_SIZE);
}
////////////////////////////////////
@@ -144,7 +144,7 @@
}
this.wait();
} catch (InterruptedException e) {
- throw new IllegalStateException(e);
+ continue;
}
}
endOffset = appendOffset;
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogReader.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogReader.java
index 173088c..9dc966c 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogReader.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogReader.java
@@ -121,7 +121,7 @@
readBuffer.limit(logPageSize);
try {
fileChannel.position(readLSN % logFileSize);
- size = fileChannel.read(readBuffer, logPageSize);
+ size = fileChannel.read(readBuffer);
} catch (IOException e) {
throw new ACIDException(e);
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogRecord.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogRecord.java
index 380e524..4b0e1f2 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogRecord.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/logging/LogRecord.java
@@ -30,13 +30,18 @@
/*
* == LogRecordFormat ==
* ---------------------------
- * [Header1] (13 bytes) : for all log types
+ * [Header1] (5 bytes) : for all log types
* LogType(1)
* JobId(4)
+ * ---------------------------
+ * [Header2] (16 bytes + PKValueSize) : for entity_commit and update log types
* DatasetId(4) //stored in dataset_dataset in Metadata Node
* PKHashValue(4)
+ * PKFieldCnt(4)
+ * PKValueSize(4)
+ * PKValue(PKValueSize)
* ---------------------------
- * [Header2] (21 bytes) : only for update log type
+ * [Header3] (21 bytes) : only for update log type
* PrevLSN(8)
* ResourceId(8) //stored in .metadata of the corresponding index in NC node
* ResourceType(1)
@@ -45,18 +50,21 @@
* [Body] (Variable size) : only for update log type
* FieldCnt(4)
* NewOp(1)
- * NewValueLength(4)
- * NewValue(NewValueLength)
+ * NewValueSize(4)
+ * NewValue(NewValueSize)
* OldOp(1)
- * OldValueLength(4)
- * OldValue(OldValueLength)
+ * OldValueSize(4)
+ * OldValue(OldValueSize)
* ---------------------------
* [Tail] (8 bytes) : for all log types
* Checksum(8)
* ---------------------------
* = LogSize =
- * 1) JOB_COMMIT and ENTITY_COMMIT: 21 bytes
- * 2) UPDATE: 56 + old and new value size (13 + 21 + 14 + old and newValueSize + 8)
+ * 1) JOB_COMMIT_LOG_SIZE: 13 bytes (5 + 8)
+ * 2) ENTITY_COMMIT: 29 + PKSize (5 + 16 + PKSize + 8)
+ * --> ENTITY_COMMIT_LOG_BASE_SIZE = 29
+ * 3) UPDATE: 64 + PKSize + New/OldValueSize (5 + 16 + PKSize + 21 + 14 + New/OldValueSize + 8)
+ * --> UPDATE_LOG_BASE_SIZE = 64
*/
public class LogRecord implements ILogRecord {
@@ -65,6 +73,9 @@
private int jobId;
private int datasetId;
private int PKHashValue;
+ private int PKFieldCnt;
+ private int PKValueSize;
+ private ITupleReference PKValue;
private long prevLSN;
private long resourceId;
private byte resourceType;
@@ -84,13 +95,18 @@
private long LSN;
private final AtomicBoolean isFlushed;
private final SimpleTupleWriter tupleWriter;
- private final SimpleTupleReference newTuple;
+ private final SimpleTupleReference readPKValue;
+ private final SimpleTupleReference readNewValue;
+ private final SimpleTupleReference readOldValue;
private final CRC32 checksumGen;
+ private int[] PKFields;
public LogRecord() {
isFlushed = new AtomicBoolean(false);
tupleWriter = new SimpleTupleWriter();
- newTuple = (SimpleTupleReference) tupleWriter.createTupleReference();
+ readPKValue = (SimpleTupleReference) tupleWriter.createTupleReference();
+ readNewValue = (SimpleTupleReference) tupleWriter.createTupleReference();
+ readOldValue = (SimpleTupleReference) tupleWriter.createTupleReference();
checksumGen = new CRC32();
}
@@ -99,8 +115,16 @@
int beginOffset = buffer.position();
buffer.put(logType);
buffer.putInt(jobId);
- buffer.putInt(datasetId);
- buffer.putInt(PKHashValue);
+ if (logType != LogType.JOB_COMMIT) {
+ buffer.putInt(datasetId);
+ buffer.putInt(PKHashValue);
+ buffer.putInt(PKFieldCnt);
+ if (PKValueSize <= 0) {
+ throw new IllegalStateException("Primary Key Size is less than or equal to 0");
+ }
+ buffer.putInt(PKValueSize);
+ writePKValue(buffer);
+ }
if (logType == LogType.UPDATE) {
buffer.putLong(prevLSN);
buffer.putLong(resourceId);
@@ -124,8 +148,16 @@
buffer.putLong(checksum);
}
+ private void writePKValue(ByteBuffer buffer) {
+ int i;
+ for (i = 0; i < PKFieldCnt; i++) {
+ buffer.put(PKValue.getFieldData(0), PKValue.getFieldStart(PKFields[i]), PKValue.getFieldLength(PKFields[i]));
+ }
+ }
+
private void writeTuple(ByteBuffer buffer, ITupleReference tuple, int size) {
tupleWriter.writeTuple(tuple, buffer.array(), buffer.position());
+ //writeTuple() doesn't change the position of the buffer.
buffer.position(buffer.position() + size);
}
@@ -141,8 +173,19 @@
try {
logType = buffer.get();
jobId = buffer.getInt();
- datasetId = buffer.getInt();
- PKHashValue = buffer.getInt();
+ if (logType == LogType.JOB_COMMIT) {
+ datasetId = -1;
+ PKHashValue = -1;
+ } else {
+ datasetId = buffer.getInt();
+ PKHashValue = buffer.getInt();
+ PKFieldCnt = buffer.getInt();
+ PKValueSize = buffer.getInt();
+ if (PKValueSize <= 0) {
+ throw new IllegalStateException("Primary Key Size is less than or equal to 0");
+ }
+ PKValue = readPKValue(buffer);
+ }
if (logType == LogType.UPDATE) {
prevLSN = buffer.getLong();
resourceId = buffer.getLong();
@@ -151,18 +194,18 @@
fieldCnt = buffer.getInt();
newOp = buffer.get();
newValueSize = buffer.getInt();
- newValue = readTuple(buffer, newValueSize);
+ newValue = readTuple(buffer, readNewValue, fieldCnt, newValueSize);
if (resourceType == ResourceType.LSM_BTREE) {
oldOp = buffer.get();
if (oldOp != (byte) (IndexOperation.NOOP.ordinal())) {
oldValueSize = buffer.getInt();
if (oldValueSize > 0) {
- oldValue = readTuple(buffer, oldValueSize);
+ oldValue = readTuple(buffer, readOldValue, fieldCnt, oldValueSize);
}
}
}
} else {
- logSize = COMMIT_LOG_SIZE;
+ computeAndSetLogSize();
}
checksum = buffer.getLong();
if (checksum != generateChecksum(buffer, beginOffset, logSize - CHECKSUM_SIZE)) {
@@ -174,27 +217,54 @@
}
return true;
}
+
+ private ITupleReference readPKValue(ByteBuffer buffer) {
+ return readTuple(buffer, readPKValue, PKFieldCnt, PKValueSize);
+ }
- private ITupleReference readTuple(ByteBuffer buffer, int size) {
- newTuple.setFieldCount(fieldCnt);
- newTuple.resetByTupleOffset(buffer, buffer.position());
- buffer.position(buffer.position() + size);
- return newTuple;
+ private ITupleReference readTuple(ByteBuffer srcBuffer, SimpleTupleReference destTuple, int fieldCnt, int size) {
+ destTuple.setFieldCount(fieldCnt);
+ destTuple.resetByTupleOffset(srcBuffer, srcBuffer.position());
+ srcBuffer.position(srcBuffer.position() + size);
+ return destTuple;
}
@Override
- public void formCommitLogRecord(ITransactionContext txnCtx, byte logType, int jobId, int datasetId, int PKHashValue) {
+ public void formJobCommitLogRecord(ITransactionContext txnCtx) {
this.txnCtx = txnCtx;
- this.logType = logType;
- this.jobId = jobId;
+ this.logType = LogType.JOB_COMMIT;
+ this.jobId = txnCtx.getJobId().getId();
+ this.datasetId = -1;
+ this.PKHashValue = -1;
+ computeAndSetLogSize();
+ }
+
+ @Override
+ public void formEntityCommitLogRecord(ITransactionContext txnCtx, int datasetId, int PKHashValue,
+ ITupleReference PKValue, int[] PKFields) {
+ this.txnCtx = txnCtx;
+ this.logType = LogType.ENTITY_COMMIT;
+ this.jobId = txnCtx.getJobId().getId();
this.datasetId = datasetId;
this.PKHashValue = PKHashValue;
- this.logSize = COMMIT_LOG_SIZE;
+ this.PKFieldCnt = PKFields.length;
+ this.PKValue = PKValue;
+ this.PKFields = PKFields;
+ computeAndSetPKValueSize();
+ computeAndSetLogSize();
}
@Override
- public void setUpdateLogSize() {
- logSize = UPDATE_LOG_BASE_SIZE + newValueSize + oldValueSize;
+ public void computeAndSetPKValueSize() {
+ int i;
+ PKValueSize = 0;
+ for (i = 0; i < PKFieldCnt; i++) {
+ PKValueSize += PKValue.getFieldLength(PKFields[i]);
+ }
+ }
+
+ private void setUpdateLogSize() {
+ logSize = UPDATE_LOG_BASE_SIZE + PKValueSize + newValueSize + oldValueSize;
if (resourceType != ResourceType.LSM_BTREE) {
logSize -= 5; //oldOp(byte: 1) + oldValueLength(int: 4)
} else {
@@ -205,18 +275,39 @@
}
@Override
+ public void computeAndSetLogSize() {
+ switch (logType) {
+ case LogType.UPDATE:
+ setUpdateLogSize();
+ break;
+ case LogType.JOB_COMMIT:
+ logSize = JOB_COMMIT_LOG_SIZE;
+ break;
+ case LogType.ENTITY_COMMIT:
+ logSize = ENTITY_COMMIT_LOG_BASE_SIZE + PKValueSize;
+ break;
+ default:
+ throw new IllegalStateException("Unsupported Log Type");
+ }
+ }
+
+ @Override
public String getLogRecordForDisplay() {
StringBuilder builder = new StringBuilder();
builder.append(" LSN : ").append(LSN);
builder.append(" LogType : ").append(LogType.toString(logType));
+ builder.append(" LogSize : ").append(logSize);
builder.append(" JobId : ").append(jobId);
- builder.append(" DatasetId : ").append(datasetId);
- builder.append(" PKHashValue : ").append(PKHashValue);
+ if (logType != LogType.JOB_COMMIT) {
+ builder.append(" DatasetId : ").append(datasetId);
+ builder.append(" PKHashValue : ").append(PKHashValue);
+ builder.append(" PKFieldCnt : ").append(PKFieldCnt);
+ builder.append(" PKSize: ").append(PKValueSize);
+ }
if (logType == LogType.UPDATE) {
builder.append(" PrevLSN : ").append(prevLSN);
builder.append(" ResourceId : ").append(resourceId);
builder.append(" ResourceType : ").append(resourceType);
- builder.append(" LogSize : ").append(logSize);
}
return builder.toString();
}
@@ -405,4 +496,25 @@
public void setLSN(long LSN) {
this.LSN = LSN;
}
+
+ @Override
+ public int getPKValueSize() {
+ return PKValueSize;
+ }
+
+ @Override
+ public ITupleReference getPKValue() {
+ return PKValue;
+ }
+
+ @Override
+ public void setPKFields(int[] primaryKeyFields) {
+ PKFields = primaryKeyFields;
+ PKFieldCnt = PKFields.length;
+ }
+
+ @Override
+ public void setPKValue(ITupleReference PKValue) {
+ this.PKValue = PKValue;
+ }
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/recovery/RecoveryManager.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/recovery/RecoveryManager.java
index 81a73d5..2ad3055 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/recovery/RecoveryManager.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/recovery/RecoveryManager.java
@@ -27,6 +27,7 @@
import java.util.Arrays;
import java.util.Collections;
import java.util.HashMap;
+import java.util.HashSet;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.List;
@@ -52,6 +53,7 @@
import edu.uci.ics.asterix.transaction.management.service.transaction.TransactionSubsystem;
import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
import edu.uci.ics.hyracks.api.lifecycle.ILifeCycleComponent;
+import edu.uci.ics.hyracks.dataflow.common.data.accessors.ITupleReference;
import edu.uci.ics.hyracks.storage.am.common.api.IIndex;
import edu.uci.ics.hyracks.storage.am.common.api.IIndexLifecycleManager;
import edu.uci.ics.hyracks.storage.am.common.impls.NoOpOperationCallback;
@@ -133,8 +135,10 @@
public void startRecovery(boolean synchronous) throws IOException, ACIDException {
int updateLogCount = 0;
- int commitLogCount = 0;
+ int entityCommitLogCount = 0;
+ int jobCommitLogCount = 0;
int redoCount = 0;
+ int jobId = -1;
state = SystemState.RECOVERING;
@@ -142,9 +146,12 @@
LOGGER.info("[RecoveryMgr] starting recovery ...");
}
- //winnerTxnTable is used to add pairs, <committed TxnId, the most recent commit Lsn of the TxnId>
- Map<TxnId, Long> winnerTxnTable = new HashMap<TxnId, Long>();
- TxnId tempKeyTxnId = new TxnId(-1, -1, -1);
+ Set<Integer> winnerJobSet = new HashSet<Integer>();
+ Map<Integer, Set<TxnId>> jobId2WinnerEntitiesMap = new HashMap<Integer, Set<TxnId>>();
+ //winnerEntity is used to add pairs, <committed TxnId, the most recent commit Lsn of the TxnId>
+ Set<TxnId> winnerEntitySet = null;
+ TxnId tempKeyTxnId = new TxnId(-1, -1, -1, null, -1, false);
+ TxnId winnerEntity = null;
//#. read checkpoint file and set lowWaterMark where anaylsis and redo start
CheckpointObject checkpointObject = readCheckpoint();
@@ -157,8 +164,6 @@
//-------------------------------------------------------------------------
// [ analysis phase ]
// - collect all committed Lsn
- // - if there are duplicate commits for the same TxnId,
- // keep only the mostRecentCommitLsn among the duplicates.
//-------------------------------------------------------------------------
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info("[RecoveryMgr] in analysis phase");
@@ -176,23 +181,34 @@
if (logRecord.getJobId() > maxJobId) {
maxJobId = logRecord.getJobId();
}
- TxnId commitTxnId = null;
switch (logRecord.getLogType()) {
case LogType.UPDATE:
if (IS_DEBUG_MODE) {
updateLogCount++;
}
break;
-
case LogType.JOB_COMMIT:
- case LogType.ENTITY_COMMIT:
- commitTxnId = new TxnId(logRecord.getJobId(), logRecord.getDatasetId(), logRecord.getPKHashValue());
- winnerTxnTable.put(commitTxnId, logRecord.getLSN());
+ winnerJobSet.add(Integer.valueOf(logRecord.getJobId()));
+ jobId2WinnerEntitiesMap.remove(Integer.valueOf(logRecord.getJobId()));
if (IS_DEBUG_MODE) {
- commitLogCount++;
+ jobCommitLogCount++;
}
break;
-
+ case LogType.ENTITY_COMMIT:
+ jobId = logRecord.getJobId();
+ winnerEntity = new TxnId(jobId, logRecord.getDatasetId(), logRecord.getPKHashValue(),
+ logRecord.getPKValue(), logRecord.getPKValueSize(), true);
+ if (!jobId2WinnerEntitiesMap.containsKey(Integer.valueOf(jobId))) {
+ winnerEntitySet = new HashSet<TxnId>();
+ jobId2WinnerEntitiesMap.put(Integer.valueOf(jobId), winnerEntitySet);
+ } else {
+ winnerEntitySet = jobId2WinnerEntitiesMap.get(Integer.valueOf(jobId));
+ }
+ winnerEntitySet.add(winnerEntity);
+ if (IS_DEBUG_MODE) {
+ entityCommitLogCount++;
+ }
+ break;
default:
throw new ACIDException("Unsupported LogType: " + logRecord.getLogType());
}
@@ -203,55 +219,48 @@
// [ redo phase ]
// - redo if
// 1) The TxnId is committed && --> guarantee durability
- // 2) lsn < commitLog's Lsn && --> deal with a case of pkHashValue collision
- // 3) lsn > maxDiskLastLsn of the index --> guarantee idempotance
+ // 2) lsn > maxDiskLastLsn of the index --> guarantee idempotence
//-------------------------------------------------------------------------
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info("[RecoveryMgr] in redo phase");
}
- //#. set log reader to the lowWaterMarkLsn again.
- logReader.initializeScan(lowWaterMarkLsn);
-
long resourceId;
long maxDiskLastLsn;
- long lsn = -1;
- long commitLsn = -1;
+ long LSN = -1;
ILSMIndex index = null;
LocalResource localResource = null;
ILocalResourceMetadata localResourceMetadata = null;
- Map<Long, Long> resourceId2MaxLsnMap = new HashMap<Long, Long>();
- TxnId jobLevelTxnId = new TxnId(-1, -1, -1);
- boolean foundWinnerTxn = false;
+ Map<Long, Long> resourceId2MaxLSNMap = new HashMap<Long, Long>();
+ boolean foundWinner = false;
//#. get indexLifeCycleManager
IAsterixAppRuntimeContextProvider appRuntimeContext = txnSubsystem.getAsterixAppRuntimeContextProvider();
IIndexLifecycleManager indexLifecycleManager = appRuntimeContext.getIndexLifecycleManager();
ILocalResourceRepository localResourceRepository = appRuntimeContext.getLocalResourceRepository();
+ //#. set log reader to the lowWaterMarkLsn again.
+ logReader.initializeScan(lowWaterMarkLsn);
logRecord = logReader.next();
while (logRecord != null) {
- lsn = logRecord.getLSN();
- foundWinnerTxn = false;
if (LogManager.IS_DEBUG_MODE) {
System.out.println(logRecord.getLogRecordForDisplay());
}
+ LSN = logRecord.getLSN();
+ jobId = logRecord.getJobId();
+ foundWinner = false;
switch (logRecord.getLogType()) {
case LogType.UPDATE:
- tempKeyTxnId.setTxnId(logRecord.getJobId(), logRecord.getDatasetId(), logRecord.getPKHashValue());
- jobLevelTxnId.setTxnId(logRecord.getJobId(), -1, -1);
- if (winnerTxnTable.containsKey(tempKeyTxnId)) {
- commitLsn = winnerTxnTable.get(tempKeyTxnId);
- if (lsn < commitLsn) {
- foundWinnerTxn = true;
- }
- } else if (winnerTxnTable.containsKey(jobLevelTxnId)) {
- commitLsn = winnerTxnTable.get(jobLevelTxnId);
- if (lsn < commitLsn) {
- foundWinnerTxn = true;
+ if (winnerJobSet.contains(Integer.valueOf(jobId))) {
+ foundWinner = true;
+ } else if (jobId2WinnerEntitiesMap.containsKey(Integer.valueOf(jobId))) {
+ winnerEntitySet = jobId2WinnerEntitiesMap.get(Integer.valueOf(jobId));
+ tempKeyTxnId.setTxnId(jobId, logRecord.getDatasetId(), logRecord.getPKHashValue(),
+ logRecord.getPKValue(), logRecord.getPKValueSize());
+ if (winnerEntitySet.contains(tempKeyTxnId)) {
+ foundWinner = true;
}
}
-
- if (foundWinnerTxn) {
+ if (foundWinner) {
resourceId = logRecord.getResourceId();
localResource = localResourceRepository.getResourceById(resourceId);
@@ -294,12 +303,12 @@
maxDiskLastLsn = abstractLSMIOCallback.getComponentLSN(index.getImmutableComponents());
//#. set resourceId and maxDiskLastLSN to the map
- resourceId2MaxLsnMap.put(resourceId, maxDiskLastLsn);
+ resourceId2MaxLSNMap.put(Long.valueOf(resourceId), Long.valueOf(maxDiskLastLsn));
} else {
- maxDiskLastLsn = resourceId2MaxLsnMap.get(resourceId);
+ maxDiskLastLsn = resourceId2MaxLSNMap.get(Long.valueOf(resourceId));
}
- if (lsn > maxDiskLastLsn) {
+ if (LSN > maxDiskLastLsn) {
redo(logRecord);
if (IS_DEBUG_MODE) {
redoCount++;
@@ -316,12 +325,11 @@
default:
throw new ACIDException("Unsupported LogType: " + logRecord.getLogType());
}
-
logRecord = logReader.next();
}
//close all indexes
- Set<Long> resourceIdList = resourceId2MaxLsnMap.keySet();
+ Set<Long> resourceIdList = resourceId2MaxLSNMap.keySet();
for (long r : resourceIdList) {
indexLifecycleManager.close(r);
}
@@ -332,8 +340,8 @@
LOGGER.info("[RecoveryMgr] recovery is completed.");
}
if (IS_DEBUG_MODE) {
- System.out.println("[RecoveryMgr] Count: Update/Commit/Redo = " + updateLogCount + "/" + commitLogCount
- + "/" + redoCount);
+ System.out.println("[RecoveryMgr] Count: Update/EntityCommit/JobCommit/Redo = " + updateLogCount + "/"
+ + entityCommitLogCount + "/" + jobCommitLogCount + "/" + redoCount);
}
}
@@ -533,15 +541,18 @@
@Override
public void rollbackTransaction(ITransactionContext txnContext) throws ACIDException {
Map<TxnId, List<Long>> loserTxnTable = new HashMap<TxnId, List<Long>>();
- TxnId tempKeyTxnId = new TxnId(-1, -1, -1);
+ TxnId tempKeyTxnId = new TxnId(-1, -1, -1, null, -1, false);
int updateLogCount = 0;
- int commitLogCount = 0;
+ int entityCommitLogCount = 0;
+ int jobId = -1;
+ int abortedJobId = txnContext.getJobId().getId();
+ long currentLSN = -1;
+ TxnId loserEntity = null;
- // Obtain the first log record written by the Job
+ // Obtain the first/last log record LSNs written by the Job
long firstLSN = txnContext.getFirstLSN();
long lastLSN = txnContext.getLastLSN();
- //TODO: make sure that the lastLsn is not updated anymore by another thread belonging to the same job.
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info(" rollbacking transaction log records from " + firstLSN + " to " + lastLSN);
}
@@ -559,62 +570,62 @@
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info(" collecting loser transaction's LSNs from " + firstLSN + " to " + lastLSN);
}
- boolean reachedLastLog = false;
List<Long> undoLSNSet = null;
ILogReader logReader = logMgr.getLogReader(false);
logReader.initializeScan(firstLSN);
- ILogRecord logRecord = logReader.next();
- while (logRecord != null) {
- if (IS_DEBUG_MODE) {
- System.out.println(logRecord.getLogRecordForDisplay());
+ ILogRecord logRecord = null;
+ while (currentLSN < lastLSN) {
+ logRecord = logReader.next();
+ if (logRecord == null) {
+ break;
+ } else {
+ if (IS_DEBUG_MODE) {
+ System.out.println(logRecord.getLogRecordForDisplay());
+ }
+ currentLSN = logRecord.getLSN();
}
-
- tempKeyTxnId.setTxnId(logRecord.getJobId(), logRecord.getDatasetId(), logRecord.getPKHashValue());
+ jobId = logRecord.getJobId();
+ if (jobId != abortedJobId) {
+ continue;
+ }
+ tempKeyTxnId.setTxnId(jobId, logRecord.getDatasetId(), logRecord.getPKHashValue(), logRecord.getPKValue(),
+ logRecord.getPKValueSize());
switch (logRecord.getLogType()) {
case LogType.UPDATE:
undoLSNSet = loserTxnTable.get(tempKeyTxnId);
if (undoLSNSet == null) {
- TxnId txnId = new TxnId(logRecord.getJobId(), logRecord.getDatasetId(),
- logRecord.getPKHashValue());
+ loserEntity = new TxnId(jobId, logRecord.getDatasetId(), logRecord.getPKHashValue(),
+ logRecord.getPKValue(), logRecord.getPKValueSize(), true);
undoLSNSet = new LinkedList<Long>();
- loserTxnTable.put(txnId, undoLSNSet);
+ loserTxnTable.put(loserEntity, undoLSNSet);
}
- undoLSNSet.add(logRecord.getLSN());
+ undoLSNSet.add(Long.valueOf(currentLSN));
if (IS_DEBUG_MODE) {
updateLogCount++;
- System.out.println("" + Thread.currentThread().getId() + "======> update[" + logRecord.getLSN()
- + "]:" + tempKeyTxnId);
+ System.out.println("" + Thread.currentThread().getId() + "======> update[" + currentLSN + "]:"
+ + tempKeyTxnId);
}
break;
case LogType.JOB_COMMIT:
+ throw new ACIDException("Unexpected LogType(" + logRecord.getLogType() + ") during abort.");
+
case LogType.ENTITY_COMMIT:
- undoLSNSet = loserTxnTable.get(tempKeyTxnId);
- if (undoLSNSet != null) {
- loserTxnTable.remove(tempKeyTxnId);
- }
+ loserTxnTable.remove(tempKeyTxnId);
if (IS_DEBUG_MODE) {
- commitLogCount++;
- System.out.println("" + Thread.currentThread().getId() + "======> commit[" + logRecord.getLSN()
- + "]" + tempKeyTxnId);
+ entityCommitLogCount++;
+ System.out.println("" + Thread.currentThread().getId() + "======> entity_commit[" + currentLSN + "]"
+ + tempKeyTxnId);
}
break;
default:
throw new ACIDException("Unsupported LogType: " + logRecord.getLogType());
}
- if (logRecord.getLSN() == lastLSN) {
- reachedLastLog = true;
- break;
- } else if (logRecord.getLSN() > lastLSN) {
- throw new IllegalStateException("LastLSN mismatch");
- }
- logRecord = logReader.next();
}
-
- if (!reachedLastLog) {
- throw new ACIDException("LastLSN mismatch: " + lastLSN + " vs " + logRecord.getLSN()
- + " during Rollback a transaction( " + txnContext.getJobId() + ")");
+ if (currentLSN != lastLSN) {
+ throw new ACIDException("LastLSN mismatch: lastLSN(" + lastLSN + ") vs currentLSN(" + currentLSN
+ + ") during abort( " + txnContext.getJobId() + ")");
}
//undo loserTxn's effect
@@ -622,7 +633,6 @@
LOGGER.info(" undoing loser transaction's effect");
}
- TxnId txnId = null;
Iterator<Entry<TxnId, List<Long>>> iter = loserTxnTable.entrySet().iterator();
int undoCount = 0;
while (iter.hasNext()) {
@@ -630,16 +640,15 @@
//Sort the lsns in order to undo in one pass.
Map.Entry<TxnId, List<Long>> loserTxn = (Map.Entry<TxnId, List<Long>>) iter.next();
- txnId = loserTxn.getKey();
-
undoLSNSet = loserTxn.getValue();
for (long undoLSN : undoLSNSet) {
- // here, all the log records are UPDATE type. So, we don't need to check the type again.
-
+ //here, all the log records are UPDATE type. So, we don't need to check the type again.
//read the corresponding log record to be undone.
logRecord = logReader.read(undoLSN);
- assert logRecord != null;
+ if (logRecord == null) {
+ throw new ACIDException("IllegalState exception during abort( " + txnContext.getJobId() + ")");
+ }
if (IS_DEBUG_MODE) {
System.out.println(logRecord.getLogRecordForDisplay());
}
@@ -656,8 +665,8 @@
LOGGER.info(" undone loser transaction's effect");
}
if (IS_DEBUG_MODE) {
- System.out.println("UpdateLogCount/CommitLogCount/UndoCount:" + updateLogCount + "/" + commitLogCount + "/"
- + undoCount);
+ System.out.println("UpdateLogCount/CommitLogCount/UndoCount:" + updateLogCount + "/" + entityCommitLogCount
+ + "/" + undoCount);
}
}
@@ -720,36 +729,53 @@
}
class TxnId {
+ public boolean isByteArrayPKValue;
public int jobId;
public int datasetId;
- public int pkHashVal;
+ public int pkHashValue;
+ public int pkSize;
+ public byte[] byteArrayPKValue;
+ public ITupleReference tupleReferencePKValue;
- public TxnId(int jobId, int datasetId, int pkHashVal) {
+ public TxnId(int jobId, int datasetId, int pkHashValue, ITupleReference pkValue, int pkSize,
+ boolean isByteArrayPKValue) {
this.jobId = jobId;
this.datasetId = datasetId;
- this.pkHashVal = pkHashVal;
+ this.pkHashValue = pkHashValue;
+ this.pkSize = pkSize;
+ this.isByteArrayPKValue = isByteArrayPKValue;
+ if (isByteArrayPKValue) {
+ this.byteArrayPKValue = new byte[pkSize];
+ readPKValueIntoByteArray(pkValue, pkSize, byteArrayPKValue);
+ } else {
+ this.tupleReferencePKValue = pkValue;
+ }
}
- public void setTxnId(int jobId, int datasetId, int pkHashVal) {
+ private void readPKValueIntoByteArray(ITupleReference pkValue, int pkSize, byte[] byteArrayPKValue) {
+ int readOffset = pkValue.getFieldStart(0);
+ byte[] readBuffer = pkValue.getFieldData(0);
+ for (int i = 0; i < pkSize; i++) {
+ byteArrayPKValue[i] = readBuffer[readOffset + i];
+ }
+ }
+
+ public void setTxnId(int jobId, int datasetId, int pkHashValue, ITupleReference pkValue, int pkSize) {
this.jobId = jobId;
this.datasetId = datasetId;
- this.pkHashVal = pkHashVal;
- }
-
- public void setTxnId(TxnId txnId) {
- this.jobId = txnId.jobId;
- this.datasetId = txnId.datasetId;
- this.pkHashVal = txnId.pkHashVal;
+ this.pkHashValue = pkHashValue;
+ this.tupleReferencePKValue = pkValue;
+ isByteArrayPKValue = false;
}
@Override
public String toString() {
- return "[" + jobId + "," + datasetId + "," + pkHashVal + "]";
+ return "[" + jobId + "," + datasetId + "," + pkHashValue + "," + pkSize + "]";
}
@Override
public int hashCode() {
- return pkHashVal;
+ return pkHashValue;
}
@Override
@@ -761,7 +787,52 @@
return false;
}
TxnId txnId = (TxnId) o;
+ return (txnId.pkHashValue == pkHashValue && txnId.datasetId == datasetId && txnId.jobId == jobId
+ && pkSize == txnId.pkSize && isEqualTo(txnId));
+ }
- return (txnId.pkHashVal == pkHashVal && txnId.datasetId == datasetId && txnId.jobId == jobId);
+ private boolean isEqualTo(TxnId txnId) {
+ if (isByteArrayPKValue && txnId.isByteArrayPKValue) {
+ return isEqual(byteArrayPKValue, txnId.byteArrayPKValue, pkSize);
+ } else if (isByteArrayPKValue && (!txnId.isByteArrayPKValue)) {
+ return isEqual(byteArrayPKValue, txnId.tupleReferencePKValue, pkSize);
+ } else if ((!isByteArrayPKValue) && txnId.isByteArrayPKValue) {
+ return isEqual(txnId.byteArrayPKValue, tupleReferencePKValue, pkSize);
+ } else {
+ return isEqual(tupleReferencePKValue, txnId.tupleReferencePKValue, pkSize);
+ }
+ }
+
+ private boolean isEqual(byte[] a, byte[] b, int size) {
+ for (int i = 0; i < size; i++) {
+ if (a[i] != b[i]) {
+ return false;
+ }
+ }
+ return true;
+ }
+
+ private boolean isEqual(byte[] a, ITupleReference b, int size) {
+ int readOffset = b.getFieldStart(0);
+ byte[] readBuffer = b.getFieldData(0);
+ for (int i = 0; i < size; i++) {
+ if (a[i] != readBuffer[readOffset + i]) {
+ return false;
+ }
+ }
+ return true;
+ }
+
+ private boolean isEqual(ITupleReference a, ITupleReference b, int size) {
+ int aOffset = a.getFieldStart(0);
+ byte[] aBuffer = a.getFieldData(0);
+ int bOffset = b.getFieldStart(0);
+ byte[] bBuffer = b.getFieldData(0);
+ for (int i = 0; i < size; i++) {
+ if (aBuffer[aOffset + i] != bBuffer[bOffset + i]) {
+ return false;
+ }
+ }
+ return true;
}
}
diff --git a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/transaction/TransactionManager.java b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/transaction/TransactionManager.java
index 01bce83..01b38c2 100644
--- a/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/transaction/TransactionManager.java
+++ b/asterix-transactions/src/main/java/edu/uci/ics/asterix/transaction/management/service/transaction/TransactionManager.java
@@ -29,7 +29,6 @@
import edu.uci.ics.asterix.common.transactions.ITransactionManager;
import edu.uci.ics.asterix.common.transactions.JobId;
import edu.uci.ics.asterix.transaction.management.service.logging.LogRecord;
-import edu.uci.ics.asterix.transaction.management.service.logging.LogType;
import edu.uci.ics.hyracks.api.lifecycle.ILifeCycleComponent;
/**
@@ -60,7 +59,8 @@
if (LOGGER.isLoggable(Level.SEVERE)) {
LOGGER.severe(msg);
}
- throw new Error(msg, ae);
+ ae.printStackTrace();
+ throw new ACIDException(msg, ae);
} finally {
txnSubsystem.getLockManager().releaseLocks(txnCtx);
transactionContextRepository.remove(txnCtx.getJobId());
@@ -90,20 +90,11 @@
@Override
public void commitTransaction(ITransactionContext txnCtx, DatasetId datasetId, int PKHashVal) throws ACIDException {
- //There is either job-level commit or entity-level commit.
- //The job-level commit will have -1 value both for datasetId and PKHashVal.
-
- //for entity-level commit
- if (PKHashVal != -1) {
- txnSubsystem.getLockManager().unlock(datasetId, PKHashVal, txnCtx, true);
- return;
- }
-
- //for job-level commit
+ //Only job-level commits call this method.
try {
if (txnCtx.isWriteTxn()) {
LogRecord logRecord = ((TransactionContext) txnCtx).getLogRecord();
- logRecord.formCommitLogRecord(txnCtx, LogType.JOB_COMMIT, txnCtx.getJobId().getId(), -1, -1);
+ logRecord.formJobCommitLogRecord(txnCtx);
txnSubsystem.getLogManager().log(logRecord);
}
} catch (Exception ae) {