Search in sources :

Example 1 with TokenizeOperator

use of org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator in project asterixdb by apache.

the class IntroduceSecondaryIndexInsertDeleteRule method rewritePost.

@Override
public boolean rewritePost(Mutable<ILogicalOperator> opRef, IOptimizationContext context) throws AlgebricksException {
    AbstractLogicalOperator op0 = (AbstractLogicalOperator) opRef.getValue();
    if (op0.getOperatorTag() != LogicalOperatorTag.DELEGATE_OPERATOR && op0.getOperatorTag() != LogicalOperatorTag.SINK) {
        return false;
    }
    if (op0.getOperatorTag() == LogicalOperatorTag.DELEGATE_OPERATOR) {
        DelegateOperator eOp = (DelegateOperator) op0;
        if (!(eOp.getDelegate() instanceof CommitOperator)) {
            return false;
        }
    }
    AbstractLogicalOperator op1 = (AbstractLogicalOperator) op0.getInputs().get(0).getValue();
    if (op1.getOperatorTag() != LogicalOperatorTag.INSERT_DELETE_UPSERT) {
        return false;
    }
    /** find the record variable */
    InsertDeleteUpsertOperator primaryIndexModificationOp = (InsertDeleteUpsertOperator) op0.getInputs().get(0).getValue();
    boolean isBulkload = primaryIndexModificationOp.isBulkload();
    ILogicalExpression newRecordExpr = primaryIndexModificationOp.getPayloadExpression().getValue();
    List<Mutable<ILogicalExpression>> newMetaExprs = primaryIndexModificationOp.getAdditionalNonFilteringExpressions();
    LogicalVariable newRecordVar;
    LogicalVariable newMetaVar = null;
    /**
         * inputOp is the assign operator which extracts primary keys from the input
         * variables (record or meta)
         */
    AbstractLogicalOperator inputOp = (AbstractLogicalOperator) primaryIndexModificationOp.getInputs().get(0).getValue();
    newRecordVar = getRecordVar(context, inputOp, newRecordExpr, 0);
    if (newMetaExprs != null && !newMetaExprs.isEmpty()) {
        if (newMetaExprs.size() > 1) {
            throw new AlgebricksException("Number of meta records can't be more than 1. Number of meta records found = " + newMetaExprs.size());
        }
        newMetaVar = getRecordVar(context, inputOp, newMetaExprs.get(0).getValue(), 1);
    }
    /*
         * At this point, we have the record variable and the insert/delete/upsert operator
         * Note: We have two operators:
         * 1. An InsertDeleteOperator (primary)
         * 2. An IndexInsertDeleteOperator (secondary)
         * The current primaryIndexModificationOp is of the first type
         */
    DataSource datasetSource = (DataSource) primaryIndexModificationOp.getDataSource();
    MetadataProvider mp = (MetadataProvider) context.getMetadataProvider();
    String dataverseName = datasetSource.getId().getDataverseName();
    String datasetName = datasetSource.getId().getDatasourceName();
    Dataset dataset = mp.findDataset(dataverseName, datasetName);
    if (dataset == null) {
        throw new AlgebricksException("Unknown dataset " + datasetName + " in dataverse " + dataverseName);
    }
    if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
        return false;
    }
    // Create operators for secondary index insert / delete.
    String itemTypeName = dataset.getItemTypeName();
    IAType itemType = mp.findType(dataset.getItemTypeDataverseName(), itemTypeName);
    if (itemType.getTypeTag() != ATypeTag.OBJECT) {
        throw new AlgebricksException("Only record types can be indexed.");
    }
    ARecordType recType = (ARecordType) itemType;
    // meta type
    ARecordType metaType = null;
    if (dataset.hasMetaPart()) {
        metaType = (ARecordType) mp.findType(dataset.getMetaItemTypeDataverseName(), dataset.getMetaItemTypeName());
    }
    List<Index> indexes = mp.getDatasetIndexes(dataset.getDataverseName(), dataset.getDatasetName());
    // Set the top operator pointer to the primary IndexInsertDeleteOperator
    ILogicalOperator currentTop = primaryIndexModificationOp;
    boolean hasSecondaryIndex = false;
    // Put an n-gram or a keyword index in the later stage of index-update,
    // since TokenizeOperator needs to be involved.
    Collections.sort(indexes, (o1, o2) -> o1.getIndexType().ordinal() - o2.getIndexType().ordinal());
    // At this point, we have the data type info, and the indexes info as well
    int secondaryIndexTotalCnt = indexes.size() - 1;
    if (secondaryIndexTotalCnt > 0) {
        op0.getInputs().clear();
    } else {
        return false;
    }
    // Initialize inputs to the SINK operator Op0 (The SINK) is now without input
    // Prepare filtering field information (This is the filter created using the "filter with" key word in the
    // create dataset ddl)
    List<String> filteringFields = ((InternalDatasetDetails) dataset.getDatasetDetails()).getFilterField();
    List<LogicalVariable> filteringVars;
    List<Mutable<ILogicalExpression>> filteringExpressions = null;
    if (filteringFields != null) {
        // The filter field var already exists. we can simply get it from the insert op
        filteringVars = new ArrayList<>();
        filteringExpressions = new ArrayList<>();
        for (Mutable<ILogicalExpression> filteringExpression : primaryIndexModificationOp.getAdditionalFilteringExpressions()) {
            filteringExpression.getValue().getUsedVariables(filteringVars);
            for (LogicalVariable var : filteringVars) {
                filteringExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(var)));
            }
        }
    }
    // Replicate Operator is applied only when doing the bulk-load.
    ReplicateOperator replicateOp = null;
    if (secondaryIndexTotalCnt > 1 && primaryIndexModificationOp.isBulkload()) {
        // Split the logical plan into "each secondary index update branch"
        // to replicate each <PK,OBJECT> pair.
        replicateOp = new ReplicateOperator(secondaryIndexTotalCnt);
        replicateOp.getInputs().add(new MutableObject<ILogicalOperator>(currentTop));
        replicateOp.setExecutionMode(ExecutionMode.PARTITIONED);
        context.computeAndSetTypeEnvironmentForOperator(replicateOp);
        currentTop = replicateOp;
    }
    /*
         * The two maps are used to store variables to which [casted] field access is assigned.
         * One for the beforeOp record and the other for the new record.
         * There are two uses for these maps:
         * 1. used for shared fields in indexes with overlapping keys.
         * 2. used for setting variables of secondary keys for each secondary index operator.
         */
    Map<IndexFieldId, LogicalVariable> fieldVarsForBeforeOperation = new HashMap<>();
    Map<IndexFieldId, LogicalVariable> fieldVarsForNewRecord = new HashMap<>();
    /*
         * if the index is enforcing field types (For open indexes), We add a cast
         * operator to ensure type safety
         */
    try {
        if (primaryIndexModificationOp.getOperation() == Kind.INSERT || primaryIndexModificationOp.getOperation() == Kind.UPSERT || /* Actually, delete should not be here but it is now until issue
                     * https://issues.apache.org/jira/browse/ASTERIXDB-1507
                     * is solved
                     */
        primaryIndexModificationOp.getOperation() == Kind.DELETE) {
            injectFieldAccessesForIndexes(context, dataset, indexes, fieldVarsForNewRecord, recType, metaType, newRecordVar, newMetaVar, primaryIndexModificationOp, false);
            if (replicateOp != null) {
                context.computeAndSetTypeEnvironmentForOperator(replicateOp);
            }
        }
        if (primaryIndexModificationOp.getOperation() == Kind.UPSERT) /* Actually, delete should be here but it is not until issue
             * https://issues.apache.org/jira/browse/ASTERIXDB-1507
             * is solved
             */
        {
            List<LogicalVariable> beforeOpMetaVars = primaryIndexModificationOp.getBeforeOpAdditionalNonFilteringVars();
            LogicalVariable beforeOpMetaVar = beforeOpMetaVars == null ? null : beforeOpMetaVars.get(0);
            currentTop = injectFieldAccessesForIndexes(context, dataset, indexes, fieldVarsForBeforeOperation, recType, metaType, primaryIndexModificationOp.getBeforeOpRecordVar(), beforeOpMetaVar, currentTop, true);
        }
    } catch (AsterixException e) {
        throw new AlgebricksException(e);
    }
    // At first, op1 is the index insert op insertOp
    for (Index index : indexes) {
        if (!index.isSecondaryIndex()) {
            continue;
        }
        hasSecondaryIndex = true;
        // Get the secondary fields names and types
        List<List<String>> secondaryKeyFields = index.getKeyFieldNames();
        List<LogicalVariable> secondaryKeyVars = new ArrayList<>();
        List<Mutable<ILogicalExpression>> secondaryExpressions = new ArrayList<>();
        List<Mutable<ILogicalExpression>> beforeOpSecondaryExpressions = new ArrayList<>();
        ILogicalOperator replicateOutput;
        for (int i = 0; i < secondaryKeyFields.size(); i++) {
            IndexFieldId indexFieldId = new IndexFieldId(index.getKeyFieldSourceIndicators().get(i), secondaryKeyFields.get(i));
            LogicalVariable skVar = fieldVarsForNewRecord.get(indexFieldId);
            secondaryKeyVars.add(skVar);
            secondaryExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(skVar)));
            if (primaryIndexModificationOp.getOperation() == Kind.UPSERT) {
                beforeOpSecondaryExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(fieldVarsForBeforeOperation.get(indexFieldId))));
            }
        }
        IndexInsertDeleteUpsertOperator indexUpdate;
        if (index.getIndexType() != IndexType.RTREE) {
            // Create an expression per key
            Mutable<ILogicalExpression> filterExpression = (primaryIndexModificationOp.getOperation() == Kind.UPSERT) ? null : createFilterExpression(secondaryKeyVars, context.getOutputTypeEnvironment(currentTop), index.isEnforcingKeyFileds());
            DataSourceIndex dataSourceIndex = new DataSourceIndex(index, dataverseName, datasetName, mp);
            // and index type is keyword or n-gram.
            if (index.getIndexType() != IndexType.BTREE && primaryIndexModificationOp.isBulkload()) {
                // Note: Bulk load case, we don't need to take care of it for upsert operation
                // Check whether the index is length-partitioned or not.
                // If partitioned, [input variables to TokenizeOperator,
                // token, number of token] pairs will be generated and
                // fed into the IndexInsertDeleteOperator.
                // If not, [input variables, token] pairs will be generated
                // and fed into the IndexInsertDeleteOperator.
                // Input variables are passed since TokenizeOperator is not an
                // filtering operator.
                boolean isPartitioned = index.getIndexType() == IndexType.LENGTH_PARTITIONED_WORD_INVIX || index.getIndexType() == IndexType.LENGTH_PARTITIONED_NGRAM_INVIX;
                // Create a new logical variable - token
                List<LogicalVariable> tokenizeKeyVars = new ArrayList<>();
                List<Mutable<ILogicalExpression>> tokenizeKeyExprs = new ArrayList<>();
                LogicalVariable tokenVar = context.newVar();
                tokenizeKeyVars.add(tokenVar);
                tokenizeKeyExprs.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(tokenVar)));
                // Check the field type of the secondary key.
                IAType secondaryKeyType;
                Pair<IAType, Boolean> keyPairType = Index.getNonNullableOpenFieldType(index.getKeyFieldTypes().get(0), secondaryKeyFields.get(0), recType);
                secondaryKeyType = keyPairType.first;
                List<Object> varTypes = new ArrayList<>();
                varTypes.add(NonTaggedFormatUtil.getTokenType(secondaryKeyType));
                // The type is short, and this does not contain type info.
                if (isPartitioned) {
                    LogicalVariable lengthVar = context.newVar();
                    tokenizeKeyVars.add(lengthVar);
                    tokenizeKeyExprs.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(lengthVar)));
                    varTypes.add(BuiltinType.SHORTWITHOUTTYPEINFO);
                }
                // TokenizeOperator to tokenize [SK, PK] pairs
                TokenizeOperator tokenUpdate = new TokenizeOperator(dataSourceIndex, primaryIndexModificationOp.getPrimaryKeyExpressions(), secondaryExpressions, tokenizeKeyVars, filterExpression, primaryIndexModificationOp.getOperation(), primaryIndexModificationOp.isBulkload(), isPartitioned, varTypes);
                tokenUpdate.getInputs().add(new MutableObject<ILogicalOperator>(currentTop));
                context.computeAndSetTypeEnvironmentForOperator(tokenUpdate);
                replicateOutput = tokenUpdate;
                indexUpdate = new IndexInsertDeleteUpsertOperator(dataSourceIndex, primaryIndexModificationOp.getPrimaryKeyExpressions(), tokenizeKeyExprs, filterExpression, primaryIndexModificationOp.getOperation(), primaryIndexModificationOp.isBulkload(), primaryIndexModificationOp.getAdditionalNonFilteringExpressions() == null ? 0 : primaryIndexModificationOp.getAdditionalNonFilteringExpressions().size());
                indexUpdate.setAdditionalFilteringExpressions(filteringExpressions);
                indexUpdate.getInputs().add(new MutableObject<ILogicalOperator>(tokenUpdate));
            } else {
                // When TokenizeOperator is not needed
                indexUpdate = new IndexInsertDeleteUpsertOperator(dataSourceIndex, primaryIndexModificationOp.getPrimaryKeyExpressions(), secondaryExpressions, filterExpression, primaryIndexModificationOp.getOperation(), primaryIndexModificationOp.isBulkload(), primaryIndexModificationOp.getAdditionalNonFilteringExpressions() == null ? 0 : primaryIndexModificationOp.getAdditionalNonFilteringExpressions().size());
                indexUpdate.setAdditionalFilteringExpressions(filteringExpressions);
                replicateOutput = indexUpdate;
                // We add the necessary expressions for upsert
                if (primaryIndexModificationOp.getOperation() == Kind.UPSERT) {
                    indexUpdate.setBeforeOpSecondaryKeyExprs(beforeOpSecondaryExpressions);
                    if (filteringFields != null) {
                        indexUpdate.setBeforeOpAdditionalFilteringExpression(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(primaryIndexModificationOp.getBeforeOpFilterVar())));
                    }
                }
                indexUpdate.getInputs().add(new MutableObject<ILogicalOperator>(currentTop));
            }
        } else {
            // Get type, dimensions and number of keys
            Pair<IAType, Boolean> keyPairType = Index.getNonNullableOpenFieldType(index.getKeyFieldTypes().get(0), secondaryKeyFields.get(0), recType);
            IAType spatialType = keyPairType.first;
            boolean isPointMBR = spatialType.getTypeTag() == ATypeTag.POINT || spatialType.getTypeTag() == ATypeTag.POINT3D;
            int dimension = NonTaggedFormatUtil.getNumDimensions(spatialType.getTypeTag());
            int numKeys = (isPointMBR && isBulkload) ? dimension : dimension * 2;
            // Get variables and expressions
            List<LogicalVariable> keyVarList = new ArrayList<>();
            List<Mutable<ILogicalExpression>> keyExprList = new ArrayList<>();
            for (int i = 0; i < numKeys; i++) {
                LogicalVariable keyVar = context.newVar();
                keyVarList.add(keyVar);
                AbstractFunctionCallExpression createMBR = new ScalarFunctionCallExpression(FunctionUtil.getFunctionInfo(BuiltinFunctions.CREATE_MBR));
                createMBR.getArguments().add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(secondaryKeyVars.get(0))));
                createMBR.getArguments().add(new MutableObject<ILogicalExpression>(new ConstantExpression(new AsterixConstantValue(new AInt32(dimension)))));
                createMBR.getArguments().add(new MutableObject<ILogicalExpression>(new ConstantExpression(new AsterixConstantValue(new AInt32(i)))));
                keyExprList.add(new MutableObject<ILogicalExpression>(createMBR));
            }
            secondaryExpressions.clear();
            for (LogicalVariable secondaryKeyVar : keyVarList) {
                secondaryExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(secondaryKeyVar)));
            }
            if (isPointMBR && isBulkload) {
                //createFieldPermutationForBulkLoadOp(int) for more details.
                for (LogicalVariable secondaryKeyVar : keyVarList) {
                    secondaryExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(secondaryKeyVar)));
                }
            }
            AssignOperator assignCoordinates = new AssignOperator(keyVarList, keyExprList);
            assignCoordinates.getInputs().add(new MutableObject<ILogicalOperator>(currentTop));
            context.computeAndSetTypeEnvironmentForOperator(assignCoordinates);
            replicateOutput = assignCoordinates;
            Mutable<ILogicalExpression> filterExpression = null;
            AssignOperator originalAssignCoordinates = null;
            // We do something similar for beforeOp key if the operation is an upsert
            if (primaryIndexModificationOp.getOperation() == Kind.UPSERT) {
                List<LogicalVariable> originalKeyVarList = new ArrayList<>();
                List<Mutable<ILogicalExpression>> originalKeyExprList = new ArrayList<>();
                // we don't do any filtering since nulls are expected here and there
                for (int i = 0; i < numKeys; i++) {
                    LogicalVariable keyVar = context.newVar();
                    originalKeyVarList.add(keyVar);
                    AbstractFunctionCallExpression createMBR = new ScalarFunctionCallExpression(FunctionUtil.getFunctionInfo(BuiltinFunctions.CREATE_MBR));
                    createMBR.getArguments().add(beforeOpSecondaryExpressions.get(0));
                    createMBR.getArguments().add(new MutableObject<ILogicalExpression>(new ConstantExpression(new AsterixConstantValue(new AInt32(dimension)))));
                    createMBR.getArguments().add(new MutableObject<ILogicalExpression>(new ConstantExpression(new AsterixConstantValue(new AInt32(i)))));
                    originalKeyExprList.add(new MutableObject<ILogicalExpression>(createMBR));
                }
                beforeOpSecondaryExpressions.clear();
                for (LogicalVariable secondaryKeyVar : originalKeyVarList) {
                    beforeOpSecondaryExpressions.add(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(secondaryKeyVar)));
                }
                originalAssignCoordinates = new AssignOperator(originalKeyVarList, originalKeyExprList);
                originalAssignCoordinates.getInputs().add(new MutableObject<ILogicalOperator>(assignCoordinates));
                context.computeAndSetTypeEnvironmentForOperator(originalAssignCoordinates);
            } else {
                // We must enforce the filter if the originating spatial type is
                // nullable.
                boolean forceFilter = keyPairType.second;
                filterExpression = createFilterExpression(keyVarList, context.getOutputTypeEnvironment(assignCoordinates), forceFilter);
            }
            DataSourceIndex dataSourceIndex = new DataSourceIndex(index, dataverseName, datasetName, mp);
            indexUpdate = new IndexInsertDeleteUpsertOperator(dataSourceIndex, primaryIndexModificationOp.getPrimaryKeyExpressions(), secondaryExpressions, filterExpression, primaryIndexModificationOp.getOperation(), primaryIndexModificationOp.isBulkload(), primaryIndexModificationOp.getAdditionalNonFilteringExpressions() == null ? 0 : primaryIndexModificationOp.getAdditionalNonFilteringExpressions().size());
            indexUpdate.setAdditionalFilteringExpressions(filteringExpressions);
            if (primaryIndexModificationOp.getOperation() == Kind.UPSERT) {
                // set before op secondary key expressions
                if (filteringFields != null) {
                    indexUpdate.setBeforeOpAdditionalFilteringExpression(new MutableObject<ILogicalExpression>(new VariableReferenceExpression(primaryIndexModificationOp.getBeforeOpFilterVar())));
                }
                // set filtering expressions
                indexUpdate.setBeforeOpSecondaryKeyExprs(beforeOpSecondaryExpressions);
                // assign --> assign beforeOp values --> secondary index upsert
                indexUpdate.getInputs().add(new MutableObject<ILogicalOperator>(originalAssignCoordinates));
            } else {
                indexUpdate.getInputs().add(new MutableObject<ILogicalOperator>(assignCoordinates));
            }
        }
        context.computeAndSetTypeEnvironmentForOperator(indexUpdate);
        if (!primaryIndexModificationOp.isBulkload() || secondaryIndexTotalCnt == 1) {
            currentTop = indexUpdate;
        } else {
            replicateOp.getOutputs().add(new MutableObject<>(replicateOutput));
        }
        if (primaryIndexModificationOp.isBulkload()) {
            // For bulk load, we connect all fanned out insert operator to a single SINK operator
            op0.getInputs().add(new MutableObject<ILogicalOperator>(indexUpdate));
        }
    }
    if (!hasSecondaryIndex) {
        return false;
    }
    if (!primaryIndexModificationOp.isBulkload()) {
        // If this is an upsert, we need to
        // Remove the current input to the SINK operator (It is actually already removed above)
        op0.getInputs().clear();
        // Connect the last index update to the SINK
        op0.getInputs().add(new MutableObject<ILogicalOperator>(currentTop));
    }
    return true;
}
Also used : HashMap(java.util.HashMap) ConstantExpression(org.apache.hyracks.algebricks.core.algebra.expressions.ConstantExpression) ArrayList(java.util.ArrayList) Index(org.apache.asterix.metadata.entities.Index) DataSourceIndex(org.apache.asterix.metadata.declared.DataSourceIndex) AString(org.apache.asterix.om.base.AString) AsterixException(org.apache.asterix.common.exceptions.AsterixException) TokenizeOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator) AsterixConstantValue(org.apache.asterix.om.constants.AsterixConstantValue) DelegateOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.DelegateOperator) List(java.util.List) AOrderedList(org.apache.asterix.om.base.AOrderedList) ArrayList(java.util.ArrayList) CommitOperator(org.apache.asterix.algebra.operators.CommitOperator) ScalarFunctionCallExpression(org.apache.hyracks.algebricks.core.algebra.expressions.ScalarFunctionCallExpression) ReplicateOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.ReplicateOperator) AbstractFunctionCallExpression(org.apache.hyracks.algebricks.core.algebra.expressions.AbstractFunctionCallExpression) AlgebricksException(org.apache.hyracks.algebricks.common.exceptions.AlgebricksException) DataSource(org.apache.asterix.metadata.declared.DataSource) MetadataProvider(org.apache.asterix.metadata.declared.MetadataProvider) VariableReferenceExpression(org.apache.hyracks.algebricks.core.algebra.expressions.VariableReferenceExpression) MutableObject(org.apache.commons.lang3.mutable.MutableObject) IAObject(org.apache.asterix.om.base.IAObject) IAType(org.apache.asterix.om.types.IAType) LogicalVariable(org.apache.hyracks.algebricks.core.algebra.base.LogicalVariable) AbstractLogicalOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator) Dataset(org.apache.asterix.metadata.entities.Dataset) ILogicalOperator(org.apache.hyracks.algebricks.core.algebra.base.ILogicalOperator) InternalDatasetDetails(org.apache.asterix.metadata.entities.InternalDatasetDetails) DataSourceIndex(org.apache.asterix.metadata.declared.DataSourceIndex) IndexInsertDeleteUpsertOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.IndexInsertDeleteUpsertOperator) AssignOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.AssignOperator) AInt32(org.apache.asterix.om.base.AInt32) Mutable(org.apache.commons.lang3.mutable.Mutable) ILogicalExpression(org.apache.hyracks.algebricks.core.algebra.base.ILogicalExpression) IndexInsertDeleteUpsertOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.IndexInsertDeleteUpsertOperator) InsertDeleteUpsertOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.InsertDeleteUpsertOperator) ARecordType(org.apache.asterix.om.types.ARecordType)

Example 2 with TokenizeOperator

use of org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator in project asterixdb by apache.

the class LogicalOperatorDeepCopyWithNewVariablesVisitor method visitTokenizeOperator.

@Override
public ILogicalOperator visitTokenizeOperator(TokenizeOperator op, ILogicalOperator arg) throws AlgebricksException {
    TokenizeOperator opCopy = new TokenizeOperator(op.getDataSourceIndex(), exprDeepCopyVisitor.deepCopyExpressionReferenceList(op.getPrimaryKeyExpressions()), exprDeepCopyVisitor.deepCopyExpressionReferenceList(op.getSecondaryKeyExpressions()), this.deepCopyVariableList(op.getTokenizeVars()), exprDeepCopyVisitor.deepCopyExpressionReference(op.getFilterExpression()), op.getOperation(), op.isBulkload(), op.isPartitioned(), op.getTokenizeVarTypes());
    deepCopyInputsAnnotationsAndExecutionMode(op, arg, opCopy);
    return opCopy;
}
Also used : TokenizeOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator)

Example 3 with TokenizeOperator

use of org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator in project asterixdb by apache.

the class OperatorDeepCopyVisitor method visitTokenizeOperator.

@Override
public ILogicalOperator visitTokenizeOperator(TokenizeOperator op, Void arg) throws AlgebricksException {
    List<Mutable<ILogicalExpression>> newPrimaryKeyExpressions = new ArrayList<>();
    deepCopyExpressionRefs(newPrimaryKeyExpressions, op.getPrimaryKeyExpressions());
    List<Mutable<ILogicalExpression>> newSecondaryKeyExpressions = new ArrayList<>();
    deepCopyExpressionRefs(newSecondaryKeyExpressions, op.getSecondaryKeyExpressions());
    List<LogicalVariable> newTokenizeVars = new ArrayList<>();
    deepCopyVars(newTokenizeVars, op.getTokenizeVars());
    Mutable<ILogicalExpression> newFilterExpression = new MutableObject<>(((AbstractLogicalExpression) op.getFilterExpression()).cloneExpression());
    List<Object> newTokenizeVarTypes = new ArrayList<>();
    deepCopyObjects(newTokenizeVarTypes, op.getTokenizeVarTypes());
    TokenizeOperator tokenizeOp = new TokenizeOperator(op.getDataSourceIndex(), newPrimaryKeyExpressions, newSecondaryKeyExpressions, newTokenizeVars, newFilterExpression, op.getOperation(), op.isBulkload(), op.isPartitioned(), newTokenizeVarTypes);
    return tokenizeOp;
}
Also used : LogicalVariable(org.apache.hyracks.algebricks.core.algebra.base.LogicalVariable) Mutable(org.apache.commons.lang3.mutable.Mutable) ILogicalExpression(org.apache.hyracks.algebricks.core.algebra.base.ILogicalExpression) TokenizeOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator) ArrayList(java.util.ArrayList) MutableObject(org.apache.commons.lang3.mutable.MutableObject) MutableObject(org.apache.commons.lang3.mutable.MutableObject)

Example 4 with TokenizeOperator

use of org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator in project asterixdb by apache.

the class IsomorphismOperatorVisitor method visitTokenizeOperator.

@Override
public Boolean visitTokenizeOperator(TokenizeOperator op, ILogicalOperator arg) throws AlgebricksException {
    AbstractLogicalOperator aop = (AbstractLogicalOperator) arg;
    if (aop.getOperatorTag() != LogicalOperatorTag.TOKENIZE) {
        return Boolean.FALSE;
    }
    TokenizeOperator tokenizeOpArg = (TokenizeOperator) copyAndSubstituteVar(op, arg);
    boolean isomorphic = VariableUtilities.varListEqualUnordered(op.getSchema(), tokenizeOpArg.getSchema());
    if (!op.getDataSourceIndex().equals(tokenizeOpArg.getDataSourceIndex())) {
        isomorphic = false;
    }
    return isomorphic;
}
Also used : AbstractLogicalOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator) TokenizeOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator)

Example 5 with TokenizeOperator

use of org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator in project asterixdb by apache.

the class TokenizePOperator method contributeRuntimeOperator.

@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema propagatedSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
    TokenizeOperator tokenizeOp = (TokenizeOperator) op;
    if (tokenizeOp.getOperation() != Kind.INSERT || !tokenizeOp.isBulkload()) {
        throw new AlgebricksException("Tokenize Operator only works when bulk-loading data.");
    }
    IMetadataProvider mp = context.getMetadataProvider();
    IVariableTypeEnvironment typeEnv = context.getTypeEnvironment(op);
    JobSpecification spec = builder.getJobSpec();
    RecordDescriptor inputDesc = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op.getInputs().get(0).getValue()), inputSchemas[0], context);
    Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> runtimeAndConstraints = mp.getTokenizerRuntime(dataSourceIndex, propagatedSchema, inputSchemas, typeEnv, primaryKeys, secondaryKeys, null, inputDesc, context, spec, true);
    builder.contributeHyracksOperator(tokenizeOp, runtimeAndConstraints.first);
    builder.contributeAlgebricksPartitionConstraint(runtimeAndConstraints.first, runtimeAndConstraints.second);
    ILogicalOperator src = tokenizeOp.getInputs().get(0).getValue();
    builder.contributeGraphEdge(src, 0, tokenizeOp, 0);
}
Also used : TokenizeOperator(org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator) RecordDescriptor(org.apache.hyracks.api.dataflow.value.RecordDescriptor) IOperatorDescriptor(org.apache.hyracks.api.dataflow.IOperatorDescriptor) ILogicalOperator(org.apache.hyracks.algebricks.core.algebra.base.ILogicalOperator) AlgebricksException(org.apache.hyracks.algebricks.common.exceptions.AlgebricksException) AlgebricksPartitionConstraint(org.apache.hyracks.algebricks.common.constraints.AlgebricksPartitionConstraint) JobSpecification(org.apache.hyracks.api.job.JobSpecification) IMetadataProvider(org.apache.hyracks.algebricks.core.algebra.metadata.IMetadataProvider) IVariableTypeEnvironment(org.apache.hyracks.algebricks.core.algebra.expressions.IVariableTypeEnvironment)

Aggregations

TokenizeOperator (org.apache.hyracks.algebricks.core.algebra.operators.logical.TokenizeOperator)5 ArrayList (java.util.ArrayList)2 Mutable (org.apache.commons.lang3.mutable.Mutable)2 MutableObject (org.apache.commons.lang3.mutable.MutableObject)2 AlgebricksException (org.apache.hyracks.algebricks.common.exceptions.AlgebricksException)2 ILogicalExpression (org.apache.hyracks.algebricks.core.algebra.base.ILogicalExpression)2 ILogicalOperator (org.apache.hyracks.algebricks.core.algebra.base.ILogicalOperator)2 LogicalVariable (org.apache.hyracks.algebricks.core.algebra.base.LogicalVariable)2 AbstractLogicalOperator (org.apache.hyracks.algebricks.core.algebra.operators.logical.AbstractLogicalOperator)2 HashMap (java.util.HashMap)1 List (java.util.List)1 CommitOperator (org.apache.asterix.algebra.operators.CommitOperator)1 AsterixException (org.apache.asterix.common.exceptions.AsterixException)1 DataSource (org.apache.asterix.metadata.declared.DataSource)1 DataSourceIndex (org.apache.asterix.metadata.declared.DataSourceIndex)1 MetadataProvider (org.apache.asterix.metadata.declared.MetadataProvider)1 Dataset (org.apache.asterix.metadata.entities.Dataset)1 Index (org.apache.asterix.metadata.entities.Index)1 InternalDatasetDetails (org.apache.asterix.metadata.entities.InternalDatasetDetails)1 AInt32 (org.apache.asterix.om.base.AInt32)1