Examples with DatasetDataSource org.apache.asterix.metadata.declared.DatasetDataSource used on opensource projects

Example 1 with DatasetDataSource

use of org.apache.asterix.metadata.declared.DatasetDataSource in project asterixdb by apache.

the class LangExpressionToPlanTranslator method translate.

public ILogicalPlan translate(Query expr, String outputDatasetName, ICompiledDmlStatement stmt, ILogicalOperator baseOp) throws AlgebricksException {
    MutableObject<ILogicalOperator> base = new MutableObject<>(new EmptyTupleSourceOperator());
    if (baseOp != null) {
        base = new MutableObject<>(baseOp);
    }
    Pair<ILogicalOperator, LogicalVariable> p = expr.accept(this, base);
    ArrayList<Mutable<ILogicalOperator>> globalPlanRoots = new ArrayList<>();
    ILogicalOperator topOp = p.first;
    List<LogicalVariable> liveVars = new ArrayList<>();
    VariableUtilities.getLiveVariables(topOp, liveVars);
    LogicalVariable unnestVar = liveVars.get(0);
    LogicalVariable resVar = unnestVar;
    if (outputDatasetName == null) {
        FileSplit outputFileSplit = metadataProvider.getOutputFile();
        if (outputFileSplit == null) {
            outputFileSplit = getDefaultOutputFileLocation(metadataProvider.getApplicationContext());
        }
        metadataProvider.setOutputFile(outputFileSplit);
        List<Mutable<ILogicalExpression>> writeExprList = new ArrayList<>(1);
        writeExprList.add(new MutableObject<>(new VariableReferenceExpression(resVar)));
        ResultSetSinkId rssId = new ResultSetSinkId(metadataProvider.getResultSetId());
        ResultSetDataSink sink = new ResultSetDataSink(rssId, null);
        DistributeResultOperator newTop = new DistributeResultOperator(writeExprList, sink);
        newTop.getInputs().add(new MutableObject<>(topOp));
        topOp = newTop;
        // Retrieve the Output RecordType (if any) and store it on
        // the DistributeResultOperator
        IAType outputRecordType = metadataProvider.findOutputRecordType();
        if (outputRecordType != null) {
            topOp.getAnnotations().put("output-record-type", outputRecordType);
        }
    } else {
        /**
             * add the collection-to-sequence right before the project,
             * because dataset only accept non-collection records
             */
        LogicalVariable seqVar = context.newVar();
        /**
             * This assign adds a marker function collection-to-sequence: if the input is a singleton collection, unnest
             * it; otherwise do nothing.
             */
        AssignOperator assignCollectionToSequence = new AssignOperator(seqVar, new MutableObject<>(new ScalarFunctionCallExpression(FunctionUtil.getFunctionInfo(BuiltinFunctions.COLLECTION_TO_SEQUENCE), new MutableObject<>(new VariableReferenceExpression(resVar)))));
        assignCollectionToSequence.getInputs().add(new MutableObject<>(topOp.getInputs().get(0).getValue()));
        topOp.getInputs().get(0).setValue(assignCollectionToSequence);
        ProjectOperator projectOperator = (ProjectOperator) topOp;
        projectOperator.getVariables().set(0, seqVar);
        resVar = seqVar;
        DatasetDataSource targetDatasource = validateDatasetInfo(metadataProvider, stmt.getDataverseName(), stmt.getDatasetName());
        List<Integer> keySourceIndicator = ((InternalDatasetDetails) targetDatasource.getDataset().getDatasetDetails()).getKeySourceIndicator();
        ArrayList<LogicalVariable> vars = new ArrayList<>();
        ArrayList<Mutable<ILogicalExpression>> exprs = new ArrayList<>();
        List<Mutable<ILogicalExpression>> varRefsForLoading = new ArrayList<>();
        List<List<String>> partitionKeys = targetDatasource.getDataset().getPrimaryKeys();
        int numOfPrimaryKeys = partitionKeys.size();
        for (int i = 0; i < numOfPrimaryKeys; i++) {
            if (keySourceIndicator == null || keySourceIndicator.get(i).intValue() == 0) {
                // record part
                PlanTranslationUtil.prepareVarAndExpression(partitionKeys.get(i), resVar, vars, exprs, varRefsForLoading, context);
            } else {
                // meta part
                PlanTranslationUtil.prepareMetaKeyAccessExpression(partitionKeys.get(i), unnestVar, exprs, vars, varRefsForLoading, context);
            }
        }
        AssignOperator assign = new AssignOperator(vars, exprs);
        List<String> additionalFilteringField = DatasetUtil.getFilterField(targetDatasource.getDataset());
        List<LogicalVariable> additionalFilteringVars;
        List<Mutable<ILogicalExpression>> additionalFilteringAssignExpressions;
        List<Mutable<ILogicalExpression>> additionalFilteringExpressions = null;
        AssignOperator additionalFilteringAssign = null;
        if (additionalFilteringField != null) {
            additionalFilteringVars = new ArrayList<>();
            additionalFilteringAssignExpressions = new ArrayList<>();
            additionalFilteringExpressions = new ArrayList<>();
            PlanTranslationUtil.prepareVarAndExpression(additionalFilteringField, resVar, additionalFilteringVars, additionalFilteringAssignExpressions, additionalFilteringExpressions, context);
            additionalFilteringAssign = new AssignOperator(additionalFilteringVars, additionalFilteringAssignExpressions);
            additionalFilteringAssign.getInputs().add(new MutableObject<>(topOp));
            assign.getInputs().add(new MutableObject<>(additionalFilteringAssign));
        } else {
            assign.getInputs().add(new MutableObject<>(topOp));
        }
        Mutable<ILogicalExpression> varRef = new MutableObject<>(new VariableReferenceExpression(resVar));
        ILogicalOperator leafOperator;
        switch(stmt.getKind()) {
            case Statement.Kind.INSERT:
                leafOperator = translateInsert(targetDatasource, varRef, varRefsForLoading, additionalFilteringExpressions, assign, stmt);
                break;
            case Statement.Kind.UPSERT:
                leafOperator = translateUpsert(targetDatasource, varRef, varRefsForLoading, additionalFilteringExpressions, assign, additionalFilteringField, unnestVar, topOp, exprs, resVar, additionalFilteringAssign, stmt);
                break;
            case Statement.Kind.DELETE:
                leafOperator = translateDelete(targetDatasource, varRef, varRefsForLoading, additionalFilteringExpressions, assign);
                break;
            case Statement.Kind.CONNECT_FEED:
                leafOperator = translateConnectFeed(targetDatasource, varRef, varRefsForLoading, additionalFilteringExpressions, assign);
                break;
            case Statement.Kind.SUBSCRIBE_FEED:
                leafOperator = translateSubscribeFeed((CompiledSubscribeFeedStatement) stmt, targetDatasource, unnestVar, topOp, exprs, resVar, varRefsForLoading, varRef, assign, additionalFilteringField, additionalFilteringAssign, additionalFilteringExpressions);
                break;
            default:
                throw new AlgebricksException("Unsupported statement kind " + stmt.getKind());
        }
        topOp = leafOperator;
    }
    globalPlanRoots.add(new MutableObject<>(topOp));
    ILogicalPlan plan = new ALogicalPlanImpl(globalPlanRoots);
    eliminateSharedOperatorReferenceForPlan(plan);
    return plan;
}

Example 2 with DatasetDataSource

use of org.apache.asterix.metadata.declared.DatasetDataSource in project asterixdb by apache.

the class LangExpressionToPlanTranslator method validateDatasetInfo.

private DatasetDataSource validateDatasetInfo(MetadataProvider metadataProvider, String dataverseName, String datasetName) throws AlgebricksException {
    Dataset dataset = metadataProvider.findDataset(dataverseName, datasetName);
    if (dataset == null) {
        throw new AlgebricksException("Cannot find dataset " + datasetName + " in dataverse " + dataverseName);
    }
    if (dataset.getDatasetType() == DatasetType.EXTERNAL) {
        throw new AlgebricksException("Cannot write output to an external dataset.");
    }
    DataSourceId sourceId = new DataSourceId(dataverseName, datasetName);
    IAType itemType = metadataProvider.findType(dataset.getItemTypeDataverseName(), dataset.getItemTypeName());
    IAType metaItemType = metadataProvider.findType(dataset.getMetaItemTypeDataverseName(), dataset.getMetaItemTypeName());
    INodeDomain domain = metadataProvider.findNodeDomain(dataset.getNodeGroupName());
    return new DatasetDataSource(sourceId, dataset, itemType, metaItemType, DataSource.Type.INTERNAL_DATASET, dataset.getDatasetDetails(), domain);
}

Example 3 with DatasetDataSource

use of org.apache.asterix.metadata.declared.DatasetDataSource in project asterixdb by apache.

the class LangExpressionToPlanTranslator method translateLoad.

@Override
public ILogicalPlan translateLoad(ICompiledDmlStatement stmt) throws AlgebricksException {
    CompiledLoadFromFileStatement clffs = (CompiledLoadFromFileStatement) stmt;
    Dataset dataset = metadataProvider.findDataset(clffs.getDataverseName(), clffs.getDatasetName());
    if (dataset == null) {
        // This would never happen since we check for this in AqlTranslator
        throw new AlgebricksException("Unable to load dataset " + clffs.getDatasetName() + " since it does not exist");
    }
    IAType itemType = metadataProvider.findType(dataset.getItemTypeDataverseName(), dataset.getItemTypeName());
    IAType metaItemType = metadataProvider.findType(dataset.getMetaItemTypeDataverseName(), dataset.getMetaItemTypeName());
    DatasetDataSource targetDatasource = validateDatasetInfo(metadataProvider, stmt.getDataverseName(), stmt.getDatasetName());
    List<List<String>> partitionKeys = targetDatasource.getDataset().getPrimaryKeys();
    if (dataset.hasMetaPart()) {
        throw new AlgebricksException(dataset.getDatasetName() + ": load dataset is not supported on Datasets with Meta records");
    }
    LoadableDataSource lds;
    try {
        lds = new LoadableDataSource(dataset, itemType, metaItemType, clffs.getAdapter(), clffs.getProperties());
    } catch (IOException e) {
        throw new AlgebricksException(e);
    }
    // etsOp is a dummy input operator used to keep the compiler happy. it
    // could be removed but would result in
    // the need to fix many rewrite rules that assume that datasourcescan
    // operators always have input.
    ILogicalOperator etsOp = new EmptyTupleSourceOperator();
    // Add a logical variable for the record.
    List<LogicalVariable> payloadVars = new ArrayList<>();
    payloadVars.add(context.newVar());
    // Create a scan operator and make the empty tuple source its input
    DataSourceScanOperator dssOp = new DataSourceScanOperator(payloadVars, lds);
    dssOp.getInputs().add(new MutableObject<>(etsOp));
    ILogicalExpression payloadExpr = new VariableReferenceExpression(payloadVars.get(0));
    Mutable<ILogicalExpression> payloadRef = new MutableObject<>(payloadExpr);
    // Creating the assign to extract the PK out of the record
    ArrayList<LogicalVariable> pkVars = new ArrayList<>();
    ArrayList<Mutable<ILogicalExpression>> pkExprs = new ArrayList<>();
    List<Mutable<ILogicalExpression>> varRefsForLoading = new ArrayList<>();
    LogicalVariable payloadVar = payloadVars.get(0);
    for (List<String> keyFieldName : partitionKeys) {
        PlanTranslationUtil.prepareVarAndExpression(keyFieldName, payloadVar, pkVars, pkExprs, varRefsForLoading, context);
    }
    AssignOperator assign = new AssignOperator(pkVars, pkExprs);
    assign.getInputs().add(new MutableObject<>(dssOp));
    // If the input is pre-sorted, we set the ordering property explicitly in the assign
    if (clffs.alreadySorted()) {
        List<OrderColumn> orderColumns = new ArrayList<>();
        for (int i = 0; i < pkVars.size(); ++i) {
            orderColumns.add(new OrderColumn(pkVars.get(i), OrderKind.ASC));
        }
        assign.setExplicitOrderingProperty(new LocalOrderProperty(orderColumns));
    }
    List<String> additionalFilteringField = DatasetUtil.getFilterField(targetDatasource.getDataset());
    List<LogicalVariable> additionalFilteringVars;
    List<Mutable<ILogicalExpression>> additionalFilteringAssignExpressions;
    List<Mutable<ILogicalExpression>> additionalFilteringExpressions = null;
    AssignOperator additionalFilteringAssign = null;
    if (additionalFilteringField != null) {
        additionalFilteringVars = new ArrayList<>();
        additionalFilteringAssignExpressions = new ArrayList<>();
        additionalFilteringExpressions = new ArrayList<>();
        PlanTranslationUtil.prepareVarAndExpression(additionalFilteringField, payloadVar, additionalFilteringVars, additionalFilteringAssignExpressions, additionalFilteringExpressions, context);
        additionalFilteringAssign = new AssignOperator(additionalFilteringVars, additionalFilteringAssignExpressions);
    }
    InsertDeleteUpsertOperator insertOp = new InsertDeleteUpsertOperator(targetDatasource, payloadRef, varRefsForLoading, InsertDeleteUpsertOperator.Kind.INSERT, true);
    insertOp.setAdditionalFilteringExpressions(additionalFilteringExpressions);
    if (additionalFilteringAssign != null) {
        additionalFilteringAssign.getInputs().add(new MutableObject<>(assign));
        insertOp.getInputs().add(new MutableObject<>(additionalFilteringAssign));
    } else {
        insertOp.getInputs().add(new MutableObject<>(assign));
    }
    ILogicalOperator leafOperator = new SinkOperator();
    leafOperator.getInputs().add(new MutableObject<>(insertOp));
    return new ALogicalPlanImpl(new MutableObject<>(leafOperator));
}

Example 4 with DatasetDataSource

use of org.apache.asterix.metadata.declared.DatasetDataSource in project asterixdb by apache.

the class IntroduceMaterializationForInsertWithSelfScanRule method checkIfInsertAndScanDatasetsSame.

private boolean checkIfInsertAndScanDatasetsSame(AbstractLogicalOperator op, String insertDatasetName) {
    boolean sameDataset = false;
    for (int i = 0; i < op.getInputs().size(); ++i) {
        AbstractLogicalOperator descendantOp = (AbstractLogicalOperator) op.getInputs().get(i).getValue();
        if (descendantOp.getOperatorTag() == LogicalOperatorTag.UNNEST_MAP) {
            UnnestMapOperator unnestMapOp = (UnnestMapOperator) descendantOp;
            ILogicalExpression unnestExpr = unnestMapOp.getExpressionRef().getValue();
            if (unnestExpr.getExpressionTag() == LogicalExpressionTag.FUNCTION_CALL) {
                AbstractFunctionCallExpression f = (AbstractFunctionCallExpression) unnestExpr;
                FunctionIdentifier fid = f.getFunctionIdentifier();
                if (!fid.equals(BuiltinFunctions.INDEX_SEARCH)) {
                    throw new IllegalStateException();
                }
                AccessMethodJobGenParams jobGenParams = new AccessMethodJobGenParams();
                jobGenParams.readFromFuncArgs(f.getArguments());
                boolean isPrimaryIndex = jobGenParams.isPrimaryIndex();
                String indexName = jobGenParams.getIndexName();
                if (isPrimaryIndex && indexName.compareTo(insertDatasetName) == 0) {
                    return true;
                }
            }
        } else if (descendantOp.getOperatorTag() == LogicalOperatorTag.DATASOURCESCAN) {
            DataSourceScanOperator dataSourceScanOp = (DataSourceScanOperator) descendantOp;
            DataSource ds = (DataSource) dataSourceScanOp.getDataSource();
            if ((ds.getDatasourceType() == Type.INTERNAL_DATASET || ds.getDatasourceType() == Type.EXTERNAL_DATASET) && ((DatasetDataSource) ds).getDataset().getDatasetName().compareTo(insertDatasetName) == 0) {
                return true;
            }
        }
        sameDataset = checkIfInsertAndScanDatasetsSame(descendantOp, insertDatasetName);
        if (sameDataset) {
            break;
        }
    }
    return sameDataset;
}

Example 5 with DatasetDataSource

use of org.apache.asterix.metadata.declared.DatasetDataSource in project asterixdb by apache.

the class IntroduceLSMComponentFilterRule method getDataset.

private Dataset getDataset(AbstractLogicalOperator op, IOptimizationContext context) throws AlgebricksException {
    AbstractLogicalOperator descendantOp = (AbstractLogicalOperator) op.getInputs().get(0).getValue();
    while (descendantOp != null) {
        if (descendantOp.getOperatorTag() == LogicalOperatorTag.DATASOURCESCAN) {
            DataSourceScanOperator dataSourceScanOp = (DataSourceScanOperator) descendantOp;
            DataSource ds = (DataSource) dataSourceScanOp.getDataSource();
            if (ds.getDatasourceType() != DataSource.Type.INTERNAL_DATASET) {
                return null;
            }
            return ((DatasetDataSource) ds).getDataset();
        } else if (descendantOp.getOperatorTag() == LogicalOperatorTag.UNNEST_MAP) {
            UnnestMapOperator unnestMapOp = (UnnestMapOperator) descendantOp;
            ILogicalExpression unnestExpr = unnestMapOp.getExpressionRef().getValue();
            if (unnestExpr.getExpressionTag() == LogicalExpressionTag.FUNCTION_CALL) {
                AbstractFunctionCallExpression f = (AbstractFunctionCallExpression) unnestExpr;
                FunctionIdentifier fid = f.getFunctionIdentifier();
                String dataverseName;
                String datasetName;
                if (BuiltinFunctions.EXTERNAL_LOOKUP.equals(fid)) {
                    dataverseName = AccessMethodUtils.getStringConstant(f.getArguments().get(0));
                    datasetName = AccessMethodUtils.getStringConstant(f.getArguments().get(1));
                } else if (fid.equals(BuiltinFunctions.INDEX_SEARCH)) {
                    AccessMethodJobGenParams jobGenParams = new AccessMethodJobGenParams();
                    jobGenParams.readFromFuncArgs(f.getArguments());
                    dataverseName = jobGenParams.dataverseName;
                    datasetName = jobGenParams.datasetName;
                } else {
                    throw new AlgebricksException("Unexpected function for Unnest Map: " + fid);
                }
                return ((MetadataProvider) context.getMetadataProvider()).findDataset(dataverseName, datasetName);
            }
        }
        if (descendantOp.getInputs().isEmpty()) {
            break;
        }
        descendantOp = (AbstractLogicalOperator) descendantOp.getInputs().get(0).getValue();
    }
    return null;
}