Examples with TypedFieldId org.apache.drill.exec.record.TypedFieldId used on opensource projects

Example 6 with TypedFieldId

use of org.apache.drill.exec.record.TypedFieldId in project drill by apache.

the class FlattenRecordBatch method setFlattenVector.

@SuppressWarnings("resource")
private void setFlattenVector() {
    final TypedFieldId typedFieldId = incoming.getValueVectorId(popConfig.getColumn());
    final MaterializedField field = incoming.getSchema().getColumn(typedFieldId.getFieldIds()[0]);
    final RepeatedValueVector vector;
    final ValueVector inVV = incoming.getValueAccessorById(field.getValueClass(), typedFieldId.getFieldIds()).getValueVector();
    if (!(inVV instanceof RepeatedValueVector)) {
        if (incoming.getRecordCount() != 0) {
            throw UserException.unsupportedError().message("Flatten does not support inputs of non-list values.").build(logger);
        }
        //when incoming recordCount is 0, don't throw exception since the type being seen here is not solid
        logger.error("setFlattenVector cast failed and recordcount is 0, create empty vector anyway.");
        vector = new RepeatedMapVector(field, oContext.getAllocator(), null);
    } else {
        vector = RepeatedValueVector.class.cast(inVV);
    }
    flattener.setFlattenField(vector);
}

Example 7 with TypedFieldId

use of org.apache.drill.exec.record.TypedFieldId in project drill by apache.

the class FlattenRecordBatch method getFlattenFieldTransferPair.

/**
   * The data layout is the same for the actual data within a repeated field, as it is in a scalar vector for
   * the same sql type. For example, a repeated int vector has a vector of offsets into a regular int vector to
   * represent the lists. As the data layout for the actual values in the same in the repeated vector as in the
   * scalar vector of the same type, we can avoid making individual copies for the column being flattened, and just
   * use vector copies between the inner vector of the repeated field to the resulting scalar vector from the flatten
   * operation. This is completed after we determine how many records will fit (as we will hit either a batch end, or
   * the end of one of the other vectors while we are copying the data of the other vectors alongside each new flattened
   * value coming out of the repeated field.)
   */
@SuppressWarnings("resource")
private TransferPair getFlattenFieldTransferPair(FieldReference reference) {
    final TypedFieldId fieldId = incoming.getValueVectorId(popConfig.getColumn());
    final Class<?> vectorClass = incoming.getSchema().getColumn(fieldId.getFieldIds()[0]).getValueClass();
    final ValueVector flattenField = incoming.getValueAccessorById(vectorClass, fieldId.getFieldIds()).getValueVector();
    TransferPair tp = null;
    if (flattenField instanceof RepeatedMapVector) {
        tp = ((RepeatedMapVector) flattenField).getTransferPairToSingleMap(reference.getAsNamePart().getName(), oContext.getAllocator());
    } else if (!(flattenField instanceof RepeatedValueVector)) {
        if (incoming.getRecordCount() != 0) {
            throw UserException.unsupportedError().message("Flatten does not support inputs of non-list values.").build(logger);
        }
        logger.error("Cannot cast {} to RepeatedValueVector", flattenField);
        //when incoming recordCount is 0, don't throw exception since the type being seen here is not solid
        final ValueVector vv = new RepeatedMapVector(flattenField.getField(), oContext.getAllocator(), null);
        tp = RepeatedValueVector.class.cast(vv).getTransferPair(reference.getAsNamePart().getName(), oContext.getAllocator());
    } else {
        final ValueVector vvIn = RepeatedValueVector.class.cast(flattenField).getDataVector();
        // vvIn may be null because of fast schema return for repeated list vectors
        if (vvIn != null) {
            tp = vvIn.getTransferPair(reference.getAsNamePart().getName(), oContext.getAllocator());
        }
    }
    return tp;
}

Example 8 with TypedFieldId

use of org.apache.drill.exec.record.TypedFieldId in project drill by apache.

the class FlattenRecordBatch method setupNewSchema.

@Override
protected boolean setupNewSchema() throws SchemaChangeException {
    this.allocationVectors = Lists.newArrayList();
    container.clear();
    final List<NamedExpression> exprs = getExpressionList();
    final ErrorCollector collector = new ErrorCollectorImpl();
    final List<TransferPair> transfers = Lists.newArrayList();
    final ClassGenerator<Flattener> cg = CodeGenerator.getRoot(Flattener.TEMPLATE_DEFINITION, context.getFunctionRegistry(), context.getOptions());
    cg.getCodeGenerator().plainJavaCapable(true);
    // Uncomment out this line to debug the generated code.
    //    cg.getCodeGenerator().saveCodeForDebugging(true);
    final IntHashSet transferFieldIds = new IntHashSet();
    final NamedExpression flattenExpr = new NamedExpression(popConfig.getColumn(), new FieldReference(popConfig.getColumn()));
    final ValueVectorReadExpression vectorRead = (ValueVectorReadExpression) ExpressionTreeMaterializer.materialize(flattenExpr.getExpr(), incoming, collector, context.getFunctionRegistry(), true);
    final FieldReference fieldReference = flattenExpr.getRef();
    final TransferPair transferPair = getFlattenFieldTransferPair(fieldReference);
    if (transferPair != null) {
        final ValueVector flattenVector = transferPair.getTo();
        // checks that list has only default ValueVector and replaces resulting ValueVector to INT typed ValueVector
        if (exprs.size() == 0 && flattenVector.getField().getType().equals(Types.LATE_BIND_TYPE)) {
            final MaterializedField outputField = MaterializedField.create(fieldReference.getAsNamePart().getName(), Types.OPTIONAL_INT);
            final ValueVector vector = TypeHelper.getNewVector(outputField, oContext.getAllocator());
            container.add(vector);
        } else {
            transfers.add(transferPair);
            container.add(flattenVector);
            transferFieldIds.add(vectorRead.getFieldId().getFieldIds()[0]);
        }
    }
    logger.debug("Added transfer for project expression.");
    ClassifierResult result = new ClassifierResult();
    for (int i = 0; i < exprs.size(); i++) {
        final NamedExpression namedExpression = exprs.get(i);
        result.clear();
        String outputName = getRef(namedExpression).getRootSegment().getPath();
        if (result != null && result.outputNames != null && result.outputNames.size() > 0) {
            for (int j = 0; j < result.outputNames.size(); j++) {
                if (!result.outputNames.get(j).equals(EMPTY_STRING)) {
                    outputName = result.outputNames.get(j);
                    break;
                }
            }
        }
        final LogicalExpression expr = ExpressionTreeMaterializer.materialize(namedExpression.getExpr(), incoming, collector, context.getFunctionRegistry(), true);
        if (collector.hasErrors()) {
            throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema.  Errors:\n %s.", collector.toErrorString()));
        }
        if (expr instanceof DrillFuncHolderExpr && ((DrillFuncHolderExpr) expr).getHolder().isComplexWriterFuncHolder()) {
            // Lazy initialization of the list of complex writers, if not done yet.
            if (complexWriters == null) {
                complexWriters = Lists.newArrayList();
            }
            // The reference name will be passed to ComplexWriter, used as the name of the output vector from the writer.
            ((DrillFuncHolderExpr) expr).getFieldReference(namedExpression.getRef());
            cg.addExpr(expr);
        } else {
            // need to do evaluation.
            final MaterializedField outputField;
            if (expr instanceof ValueVectorReadExpression) {
                final TypedFieldId id = ValueVectorReadExpression.class.cast(expr).getFieldId();
                @SuppressWarnings("resource") final ValueVector incomingVector = incoming.getValueAccessorById(id.getIntermediateClass(), id.getFieldIds()).getValueVector();
                // when the first batch will be empty.
                if (incomingVector != null) {
                    outputField = incomingVector.getField().clone();
                } else {
                    outputField = MaterializedField.create(outputName, expr.getMajorType());
                }
            } else {
                outputField = MaterializedField.create(outputName, expr.getMajorType());
            }
            @SuppressWarnings("resource") final ValueVector vector = TypeHelper.getNewVector(outputField, oContext.getAllocator());
            allocationVectors.add(vector);
            TypedFieldId fid = container.add(vector);
            ValueVectorWriteExpression write = new ValueVectorWriteExpression(fid, expr, true);
            cg.addExpr(write);
            logger.debug("Added eval for project expression.");
        }
    }
    cg.rotateBlock();
    cg.getEvalBlock()._return(JExpr.TRUE);
    container.buildSchema(SelectionVectorMode.NONE);
    try {
        this.flattener = context.getImplementationClass(cg.getCodeGenerator());
        flattener.setup(context, incoming, this, transfers);
    } catch (ClassTransformationException | IOException e) {
        throw new SchemaChangeException("Failure while attempting to load generated class", e);
    }
    return true;
}

Example 9 with TypedFieldId

use of org.apache.drill.exec.record.TypedFieldId in project drill by apache.

the class HashJoinBatch method setupHashJoinProbe.

public HashJoinProbe setupHashJoinProbe() throws ClassTransformationException, IOException {
    final CodeGenerator<HashJoinProbe> cg = CodeGenerator.get(HashJoinProbe.TEMPLATE_DEFINITION, context.getFunctionRegistry(), context.getOptions());
    cg.plainJavaCapable(true);
    // Uncomment out this line to debug the generated code.
    //    cg.saveCodeForDebugging(true);
    final ClassGenerator<HashJoinProbe> g = cg.getRoot();
    // Generate the code to project build side records
    g.setMappingSet(projectBuildMapping);
    int fieldId = 0;
    final JExpression buildIndex = JExpr.direct("buildIndex");
    final JExpression outIndex = JExpr.direct("outIndex");
    g.rotateBlock();
    if (rightSchema != null) {
        for (final MaterializedField field : rightSchema) {
            final MajorType inputType = field.getType();
            final MajorType outputType;
            // not nullable so we must exclude them from the check below (see DRILL-2197).
            if ((joinType == JoinRelType.LEFT || joinType == JoinRelType.FULL) && inputType.getMode() == DataMode.REQUIRED && inputType.getMinorType() != TypeProtos.MinorType.MAP) {
                outputType = Types.overrideMode(inputType, DataMode.OPTIONAL);
            } else {
                outputType = inputType;
            }
            // make sure to project field with children for children to show up in the schema
            final MaterializedField projected = field.withType(outputType);
            // Add the vector to our output container
            container.addOrGet(projected);
            final JVar inVV = g.declareVectorValueSetupAndMember("buildBatch", new TypedFieldId(field.getType(), true, fieldId));
            final JVar outVV = g.declareVectorValueSetupAndMember("outgoing", new TypedFieldId(outputType, false, fieldId));
            g.getEvalBlock().add(outVV.invoke("copyFromSafe").arg(buildIndex.band(JExpr.lit((int) Character.MAX_VALUE))).arg(outIndex).arg(inVV.component(buildIndex.shrz(JExpr.lit(16)))));
            g.rotateBlock();
            fieldId++;
        }
    }
    // Generate the code to project probe side records
    g.setMappingSet(projectProbeMapping);
    int outputFieldId = fieldId;
    fieldId = 0;
    final JExpression probeIndex = JExpr.direct("probeIndex");
    if (leftUpstream == IterOutcome.OK || leftUpstream == IterOutcome.OK_NEW_SCHEMA) {
        for (final VectorWrapper<?> vv : left) {
            final MajorType inputType = vv.getField().getType();
            final MajorType outputType;
            // not nullable so we must exclude them from the check below (see DRILL-2771, DRILL-2197).
            if ((joinType == JoinRelType.RIGHT || joinType == JoinRelType.FULL) && inputType.getMode() == DataMode.REQUIRED && inputType.getMinorType() != TypeProtos.MinorType.MAP) {
                outputType = Types.overrideMode(inputType, DataMode.OPTIONAL);
            } else {
                outputType = inputType;
            }
            final ValueVector v = container.addOrGet(MaterializedField.create(vv.getField().getPath(), outputType));
            if (v instanceof AbstractContainerVector) {
                vv.getValueVector().makeTransferPair(v);
                v.clear();
            }
            final JVar inVV = g.declareVectorValueSetupAndMember("probeBatch", new TypedFieldId(inputType, false, fieldId));
            final JVar outVV = g.declareVectorValueSetupAndMember("outgoing", new TypedFieldId(outputType, false, outputFieldId));
            g.getEvalBlock().add(outVV.invoke("copyFromSafe").arg(probeIndex).arg(outIndex).arg(inVV));
            g.rotateBlock();
            fieldId++;
            outputFieldId++;
        }
    }
    final HashJoinProbe hj = context.getImplementationClass(cg);
    return hj;
}

Example 10 with TypedFieldId

use of org.apache.drill.exec.record.TypedFieldId in project drill by apache.

the class MergeJoinBatch method generateNewWorker.

private JoinWorker generateNewWorker() throws ClassTransformationException, IOException, SchemaChangeException {
    final ClassGenerator<JoinWorker> cg = CodeGenerator.getRoot(JoinWorker.TEMPLATE_DEFINITION, context.getFunctionRegistry(), context.getOptions());
    cg.getCodeGenerator().plainJavaCapable(true);
    // Uncomment out this line to debug the generated code.
    //    cg.getCodeGenerator().saveCodeForDebugging(true);
    final ErrorCollector collector = new ErrorCollectorImpl();
    // Generate members and initialization code
    /////////////////////////////////////////
    // declare and assign JoinStatus member
    cg.setMappingSet(setupMapping);
    JClass joinStatusClass = cg.getModel().ref(JoinStatus.class);
    JVar joinStatus = cg.clazz.field(JMod.NONE, joinStatusClass, "status");
    cg.getSetupBlock().assign(JExpr._this().ref(joinStatus), JExpr.direct("status"));
    // declare and assign outgoing VectorContainer member
    JClass vectorContainerClass = cg.getModel().ref(VectorContainer.class);
    JVar outgoingVectorContainer = cg.clazz.field(JMod.NONE, vectorContainerClass, "outgoing");
    cg.getSetupBlock().assign(JExpr._this().ref(outgoingVectorContainer), JExpr.direct("outgoing"));
    // declare and assign incoming left RecordBatch member
    JClass recordBatchClass = cg.getModel().ref(RecordIterator.class);
    JVar incomingLeftRecordBatch = cg.clazz.field(JMod.NONE, recordBatchClass, "incomingLeft");
    cg.getSetupBlock().assign(JExpr._this().ref(incomingLeftRecordBatch), joinStatus.ref("left"));
    // declare and assign incoming right RecordBatch member
    JVar incomingRightRecordBatch = cg.clazz.field(JMod.NONE, recordBatchClass, "incomingRight");
    cg.getSetupBlock().assign(JExpr._this().ref(incomingRightRecordBatch), joinStatus.ref("right"));
    // declare 'incoming' member so VVReadExpr generated code can point to the left or right batch
    JVar incomingRecordBatch = cg.clazz.field(JMod.NONE, recordBatchClass, "incoming");
    /*
     * Materialize expressions on both sides of the join condition. Check if both the sides
     * have the same return type, if not then inject casts so that comparison function will work as
     * expected
     */
    LogicalExpression[] leftExpr = new LogicalExpression[conditions.size()];
    LogicalExpression[] rightExpr = new LogicalExpression[conditions.size()];
    IterOutcome lastLeftStatus = status.getLeftStatus();
    IterOutcome lastRightStatus = status.getRightStatus();
    for (int i = 0; i < conditions.size(); i++) {
        JoinCondition condition = conditions.get(i);
        leftExpr[i] = materializeExpression(condition.getLeft(), lastLeftStatus, leftIterator, collector);
        rightExpr[i] = materializeExpression(condition.getRight(), lastRightStatus, rightIterator, collector);
    }
    // call to throw an exception. In this case we can safely skip adding the casts
    if (lastRightStatus != IterOutcome.NONE) {
        JoinUtils.addLeastRestrictiveCasts(leftExpr, leftIterator, rightExpr, rightIterator, context);
    }
    //generate doCompare() method
    /////////////////////////////////////////
    generateDoCompare(cg, incomingRecordBatch, leftExpr, incomingLeftRecordBatch, rightExpr, incomingRightRecordBatch, collector);
    // generate copyLeft()
    //////////////////////
    cg.setMappingSet(copyLeftMapping);
    int vectorId = 0;
    if (worker == null || !status.left.finished()) {
        for (VectorWrapper<?> vw : leftIterator) {
            MajorType inputType = vw.getField().getType();
            MajorType outputType;
            if (joinType == JoinRelType.RIGHT && inputType.getMode() == DataMode.REQUIRED) {
                outputType = Types.overrideMode(inputType, DataMode.OPTIONAL);
            } else {
                outputType = inputType;
            }
            // TODO (DRILL-4011): Factor out CopyUtil and use it here.
            JVar vvIn = cg.declareVectorValueSetupAndMember("incomingLeft", new TypedFieldId(inputType, vectorId));
            JVar vvOut = cg.declareVectorValueSetupAndMember("outgoing", new TypedFieldId(outputType, vectorId));
            // todo: check result of copyFromSafe and grow allocation
            cg.getEvalBlock().add(vvOut.invoke("copyFromSafe").arg(copyLeftMapping.getValueReadIndex()).arg(copyLeftMapping.getValueWriteIndex()).arg(vvIn));
            cg.rotateBlock();
            ++vectorId;
        }
    }
    // generate copyRight()
    ///////////////////////
    cg.setMappingSet(copyRightMappping);
    int rightVectorBase = vectorId;
    if (status.getRightStatus() != IterOutcome.NONE && (worker == null || !status.right.finished())) {
        for (VectorWrapper<?> vw : rightIterator) {
            MajorType inputType = vw.getField().getType();
            MajorType outputType;
            if (joinType == JoinRelType.LEFT && inputType.getMode() == DataMode.REQUIRED) {
                outputType = Types.overrideMode(inputType, DataMode.OPTIONAL);
            } else {
                outputType = inputType;
            }
            // TODO (DRILL-4011): Factor out CopyUtil and use it here.
            JVar vvIn = cg.declareVectorValueSetupAndMember("incomingRight", new TypedFieldId(inputType, vectorId - rightVectorBase));
            JVar vvOut = cg.declareVectorValueSetupAndMember("outgoing", new TypedFieldId(outputType, vectorId));
            // todo: check result of copyFromSafe and grow allocation
            cg.getEvalBlock().add(vvOut.invoke("copyFromSafe").arg(copyRightMappping.getValueReadIndex()).arg(copyRightMappping.getValueWriteIndex()).arg(vvIn));
            cg.rotateBlock();
            ++vectorId;
        }
    }
    JoinWorker w = context.getImplementationClass(cg);
    w.setupJoin(context, status, this.container);
    return w;
}