use of org.apache.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator 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));
}
use of org.apache.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator 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;
}
use of org.apache.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator in project asterixdb by apache.
the class OptimizableOperatorSubTree method getPrimaryKeyVars.
/**
* Get primary key variables from the given data-source.
*/
public void getPrimaryKeyVars(Mutable<ILogicalOperator> dataSourceRefToFetch, List<LogicalVariable> target) throws AlgebricksException {
Mutable<ILogicalOperator> dataSourceRefToFetchKey = (dataSourceRefToFetch == null) ? dataSourceRef : dataSourceRefToFetch;
switch(dataSourceType) {
case DATASOURCE_SCAN:
DataSourceScanOperator dataSourceScan = (DataSourceScanOperator) getDataSourceRef().getValue();
int numPrimaryKeys = dataset.getPrimaryKeys().size();
for (int i = 0; i < numPrimaryKeys; i++) {
target.add(dataSourceScan.getVariables().get(i));
}
break;
case PRIMARY_INDEX_LOOKUP:
AbstractUnnestMapOperator unnestMapOp = (AbstractUnnestMapOperator) dataSourceRefToFetchKey.getValue();
List<LogicalVariable> primaryKeys = null;
primaryKeys = AccessMethodUtils.getPrimaryKeyVarsFromPrimaryUnnestMap(dataset, unnestMapOp);
target.addAll(primaryKeys);
break;
case EXTERNAL_SCAN:
break;
case NO_DATASOURCE:
default:
throw CompilationException.create(ErrorCode.SUBTREE_HAS_NO_DATA_SOURCE);
}
}
use of org.apache.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator in project asterixdb by apache.
the class IntroduceLSMComponentFilterRule method findMacthedExprFieldName.
private boolean findMacthedExprFieldName(IOptimizableFuncExpr optFuncExpr, AbstractLogicalOperator op, Dataset dataset, ARecordType recType, List<Index> datasetIndexes, IOptimizationContext context) throws AlgebricksException {
AbstractLogicalOperator descendantOp = (AbstractLogicalOperator) op.getInputs().get(0).getValue();
while (descendantOp != null) {
if (descendantOp.getOperatorTag() == LogicalOperatorTag.ASSIGN) {
AssignOperator assignOp = (AssignOperator) descendantOp;
List<LogicalVariable> varList = assignOp.getVariables();
for (int varIndex = 0; varIndex < varList.size(); varIndex++) {
LogicalVariable var = varList.get(varIndex);
int funcVarIndex = optFuncExpr.findLogicalVar(var);
if (funcVarIndex == -1) {
continue;
}
List<String> fieldName = getFieldNameFromSubAssignTree(optFuncExpr, descendantOp, varIndex, recType).second;
if (fieldName == null) {
return false;
}
optFuncExpr.setFieldName(funcVarIndex, fieldName);
return true;
}
} else if (descendantOp.getOperatorTag() == LogicalOperatorTag.DATASOURCESCAN) {
DataSourceScanOperator scanOp = (DataSourceScanOperator) descendantOp;
List<LogicalVariable> varList = scanOp.getVariables();
for (int varIndex = 0; varIndex < varList.size(); varIndex++) {
LogicalVariable var = varList.get(varIndex);
int funcVarIndex = optFuncExpr.findLogicalVar(var);
if (funcVarIndex == -1) {
continue;
}
// The variable value is one of the partitioning fields.
List<String> fieldName = dataset.getPrimaryKeys().get(varIndex);
if (fieldName == null) {
return false;
}
optFuncExpr.setFieldName(funcVarIndex, fieldName);
return true;
}
} else if (descendantOp.getOperatorTag() == LogicalOperatorTag.UNNEST_MAP) {
UnnestMapOperator unnestMapOp = (UnnestMapOperator) descendantOp;
List<LogicalVariable> varList = unnestMapOp.getVariables();
for (int varIndex = 0; varIndex < varList.size(); varIndex++) {
LogicalVariable var = varList.get(varIndex);
int funcVarIndex = optFuncExpr.findLogicalVar(var);
if (funcVarIndex == -1) {
continue;
}
String indexName;
Index index = null;
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());
indexName = jobGenParams.indexName;
for (Index idx : datasetIndexes) {
if (idx.getIndexName().compareTo(indexName) == 0) {
index = idx;
break;
}
}
}
IAType metaItemType = ((MetadataProvider) context.getMetadataProvider()).findType(dataset.getMetaItemTypeDataverseName(), dataset.getMetaItemTypeName());
ARecordType metaRecType = (ARecordType) metaItemType;
int numSecondaryKeys = KeyFieldTypeUtil.getNumSecondaryKeys(index, recType, metaRecType);
List<String> fieldName;
if (varIndex >= numSecondaryKeys) {
fieldName = dataset.getPrimaryKeys().get(varIndex - numSecondaryKeys);
} else {
fieldName = index.getKeyFieldNames().get(varIndex);
}
if (fieldName == null) {
return false;
}
optFuncExpr.setFieldName(funcVarIndex, fieldName);
return true;
}
}
if (descendantOp.getInputs().isEmpty()) {
break;
}
descendantOp = (AbstractLogicalOperator) descendantOp.getInputs().get(0).getValue();
}
return false;
}
use of org.apache.hyracks.algebricks.core.algebra.operators.logical.DataSourceScanOperator 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;
}
Aggregations