Examples with FieldReference org.apache.drill.common.expression.FieldReference used on opensource projects

Example 36 with FieldReference

use of org.apache.drill.common.expression.FieldReference in project drill by apache.

the class AggPrelBase method createKeysAndExprs.

protected void createKeysAndExprs() {
    final List<String> childFields = getInput().getRowType().getFieldNames();
    final List<String> fields = getRowType().getFieldNames();
    for (int group : BitSets.toIter(groupSet)) {
        FieldReference fr = FieldReference.getWithQuotedRef(childFields.get(group));
        keys.add(new NamedExpression(fr, fr));
    }
    for (Ord<AggregateCall> aggCall : Ord.zip(aggCalls)) {
        int aggExprOrdinal = groupSet.cardinality() + aggCall.i;
        FieldReference ref = FieldReference.getWithQuotedRef(fields.get(aggExprOrdinal));
        LogicalExpression expr = toDrill(aggCall.e, childFields);
        NamedExpression ne = new NamedExpression(expr, ref);
        aggExprs.add(ne);
        if (getOperatorPhase() == OperatorPhase.PHASE_1of2) {
            if (aggCall.e.getAggregation().getName().equals("COUNT")) {
                // If we are doing a COUNT aggregate in Phase1of2, then in Phase2of2 we should SUM the COUNTs,
                SqlAggFunction sumAggFun = new SqlSumCountAggFunction(aggCall.e.getType());
                AggregateCall newAggCall = AggregateCall.create(sumAggFun, aggCall.e.isDistinct(), aggCall.e.isApproximate(), Collections.singletonList(aggExprOrdinal), aggCall.e.filterArg, aggCall.e.getType(), aggCall.e.getName());
                phase2AggCallList.add(newAggCall);
            } else {
                AggregateCall newAggCall = AggregateCall.create(aggCall.e.getAggregation(), aggCall.e.isDistinct(), aggCall.e.isApproximate(), Collections.singletonList(aggExprOrdinal), aggCall.e.filterArg, aggCall.e.getType(), aggCall.e.getName());
                phase2AggCallList.add(newAggCall);
            }
        }
    }
}

Example 37 with FieldReference

use of org.apache.drill.common.expression.FieldReference in project drill by apache.

the class PrelUtil method getOrdering.

public static List<Ordering> getOrdering(RelCollation collation, RelDataType rowType) {
    List<Ordering> orderExpr = new ArrayList<>();
    List<String> childFields = rowType.getFieldNames();
    Function<RelFieldCollation, String> fieldNameProvider;
    if (collation instanceof MetadataAggPrule.NamedRelCollation) {
        fieldNameProvider = fieldCollation -> {
            MetadataAggPrule.NamedRelCollation namedCollation = (MetadataAggPrule.NamedRelCollation) collation;
            return namedCollation.getName(fieldCollation.getFieldIndex());
        };
    } else {
        fieldNameProvider = fieldCollation -> childFields.get(fieldCollation.getFieldIndex());
    }
    collation.getFieldCollations().forEach(fieldCollation -> {
        FieldReference fieldReference = new FieldReference(fieldNameProvider.apply(fieldCollation), ExpressionPosition.UNKNOWN);
        orderExpr.add(new Ordering(fieldCollation.getDirection(), fieldReference, fieldCollation.nullDirection));
    });
    return orderExpr;
}

Example 38 with FieldReference

use of org.apache.drill.common.expression.FieldReference in project drill by apache.

the class ProjectRecordBatch method setupNewSchema.

@Override
protected boolean setupNewSchema() throws SchemaChangeException {
    if (allocationVectors != null) {
        for (final ValueVector v : allocationVectors) {
            v.clear();
        }
    }
    this.allocationVectors = Lists.newArrayList();
    if (complexWriters != null) {
        container.clear();
    } else {
        container.zeroVectors();
    }
    final List<NamedExpression> exprs = getExpressionList();
    final ErrorCollector collector = new ErrorCollectorImpl();
    final List<TransferPair> transfers = Lists.newArrayList();
    final ClassGenerator<Projector> cg = CodeGenerator.getRoot(Projector.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 boolean isAnyWildcard = isAnyWildcard(exprs);
    final ClassifierResult result = new ClassifierResult();
    final boolean classify = isClassificationNeeded(exprs);
    for (int i = 0; i < exprs.size(); i++) {
        final NamedExpression namedExpression = exprs.get(i);
        result.clear();
        if (classify && namedExpression.getExpr() instanceof SchemaPath) {
            classifyExpr(namedExpression, incoming, result);
            if (result.isStar) {
                // The value indicates which wildcard we are processing now
                final Integer value = result.prefixMap.get(result.prefix);
                if (value != null && value.intValue() == 1) {
                    int k = 0;
                    for (final VectorWrapper<?> wrapper : incoming) {
                        final ValueVector vvIn = wrapper.getValueVector();
                        if (k > result.outputNames.size() - 1) {
                            assert false;
                        }
                        // get the renamed column names
                        final String name = result.outputNames.get(k++);
                        if (name == EMPTY_STRING) {
                            continue;
                        }
                        if (isImplicitFileColumn(vvIn)) {
                            continue;
                        }
                        final FieldReference ref = new FieldReference(name);
                        final ValueVector vvOut = container.addOrGet(MaterializedField.create(ref.getAsNamePart().getName(), vvIn.getField().getType()), callBack);
                        final TransferPair tp = vvIn.makeTransferPair(vvOut);
                        transfers.add(tp);
                    }
                } else if (value != null && value.intValue() > 1) {
                    // subsequent wildcards should do a copy of incoming valuevectors
                    int k = 0;
                    for (final VectorWrapper<?> wrapper : incoming) {
                        final ValueVector vvIn = wrapper.getValueVector();
                        final SchemaPath originalPath = SchemaPath.getSimplePath(vvIn.getField().getPath());
                        if (k > result.outputNames.size() - 1) {
                            assert false;
                        }
                        // get the renamed column names
                        final String name = result.outputNames.get(k++);
                        if (name == EMPTY_STRING) {
                            continue;
                        }
                        if (isImplicitFileColumn(vvIn)) {
                            continue;
                        }
                        final LogicalExpression expr = ExpressionTreeMaterializer.materialize(originalPath, incoming, collector, context.getFunctionRegistry());
                        if (collector.hasErrors()) {
                            throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema.  Errors:\n %s.", collector.toErrorString()));
                        }
                        final MaterializedField outputField = MaterializedField.create(name, expr.getMajorType());
                        final ValueVector vv = container.addOrGet(outputField, callBack);
                        allocationVectors.add(vv);
                        final TypedFieldId fid = container.getValueVectorId(SchemaPath.getSimplePath(outputField.getPath()));
                        final ValueVectorWriteExpression write = new ValueVectorWriteExpression(fid, expr, true);
                        final HoldingContainer hc = cg.addExpr(write, ClassGenerator.BlkCreateMode.TRUE_IF_BOUND);
                    }
                }
                continue;
            }
        } else {
            // For the columns which do not needed to be classified,
            // it is still necessary to ensure the output column name is unique
            result.outputNames = Lists.newArrayList();
            final String outputName = getRef(namedExpression).getRootSegment().getPath();
            addToResultMaps(outputName, result, true);
        }
        String outputName = getRef(namedExpression).getRootSegment().getPath();
        if (result != null && result.outputNames != null && result.outputNames.size() > 0) {
            boolean isMatched = false;
            for (int j = 0; j < result.outputNames.size(); j++) {
                if (!result.outputNames.get(j).equals(EMPTY_STRING)) {
                    outputName = result.outputNames.get(j);
                    isMatched = true;
                    break;
                }
            }
            if (!isMatched) {
                continue;
            }
        }
        final LogicalExpression expr = ExpressionTreeMaterializer.materialize(namedExpression.getExpr(), incoming, collector, context.getFunctionRegistry(), true, unionTypeEnabled);
        final MaterializedField outputField = MaterializedField.create(outputName, expr.getMajorType());
        if (collector.hasErrors()) {
            throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema.  Errors:\n %s.", collector.toErrorString()));
        }
        // add value vector to transfer if direct reference and this is allowed, otherwise, add to evaluation stack.
        if (expr instanceof ValueVectorReadExpression && incoming.getSchema().getSelectionVectorMode() == SelectionVectorMode.NONE && !((ValueVectorReadExpression) expr).hasReadPath() && !isAnyWildcard && !transferFieldIds.contains(((ValueVectorReadExpression) expr).getFieldId().getFieldIds()[0])) {
            final ValueVectorReadExpression vectorRead = (ValueVectorReadExpression) expr;
            final TypedFieldId id = vectorRead.getFieldId();
            final ValueVector vvIn = incoming.getValueAccessorById(id.getIntermediateClass(), id.getFieldIds()).getValueVector();
            Preconditions.checkNotNull(incoming);
            final FieldReference ref = getRef(namedExpression);
            final ValueVector vvOut = container.addOrGet(MaterializedField.create(ref.getAsUnescapedPath(), vectorRead.getMajorType()), callBack);
            final TransferPair tp = vvIn.makeTransferPair(vvOut);
            transfers.add(tp);
            transferFieldIds.add(vectorRead.getFieldId().getFieldIds()[0]);
        } else 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();
            } else {
                complexWriters.clear();
            }
            // 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, ClassGenerator.BlkCreateMode.TRUE_IF_BOUND);
            if (complexFieldReferencesList == null) {
                complexFieldReferencesList = Lists.newArrayList();
            }
            // save the field reference for later for getting schema when input is empty
            complexFieldReferencesList.add(namedExpression.getRef());
        } else {
            // need to do evaluation.
            final ValueVector vector = container.addOrGet(outputField, callBack);
            allocationVectors.add(vector);
            final TypedFieldId fid = container.getValueVectorId(SchemaPath.getSimplePath(outputField.getPath()));
            final boolean useSetSafe = !(vector instanceof FixedWidthVector);
            final ValueVectorWriteExpression write = new ValueVectorWriteExpression(fid, expr, useSetSafe);
            final HoldingContainer hc = cg.addExpr(write, ClassGenerator.BlkCreateMode.TRUE_IF_BOUND);
            // We cannot do multiple transfers from the same vector. However we still need to instantiate the output vector.
            if (expr instanceof ValueVectorReadExpression) {
                final ValueVectorReadExpression vectorRead = (ValueVectorReadExpression) expr;
                if (!vectorRead.hasReadPath()) {
                    final TypedFieldId id = vectorRead.getFieldId();
                    final ValueVector vvIn = incoming.getValueAccessorById(id.getIntermediateClass(), id.getFieldIds()).getValueVector();
                    vvIn.makeTransferPair(vector);
                }
            }
            logger.debug("Added eval for project expression.");
        }
    }
    try {
        CodeGenerator<Projector> codeGen = cg.getCodeGenerator();
        codeGen.plainJavaCapable(true);
        // Uncomment out this line to debug the generated code.
        //      codeGen.saveCodeForDebugging(true);
        this.projector = context.getImplementationClass(codeGen);
        projector.setup(context, incoming, this, transfers);
    } catch (ClassTransformationException | IOException e) {
        throw new SchemaChangeException("Failure while attempting to load generated class", e);
    }
    if (container.isSchemaChanged()) {
        container.buildSchema(SelectionVectorMode.NONE);
        return true;
    } else {
        return false;
    }
}

Example 39 with FieldReference

use of org.apache.drill.common.expression.FieldReference in project drill by apache.

the class ProjectRecordBatch method getExpressionList.

private List<NamedExpression> getExpressionList() {
    if (popConfig.getExprs() != null) {
        return popConfig.getExprs();
    }
    final List<NamedExpression> exprs = Lists.newArrayList();
    for (final MaterializedField field : incoming.getSchema()) {
        if (Types.isComplex(field.getType()) || Types.isRepeated(field.getType())) {
            final LogicalExpression convertToJson = FunctionCallFactory.createConvert(ConvertExpression.CONVERT_TO, "JSON", SchemaPath.getSimplePath(field.getPath()), ExpressionPosition.UNKNOWN);
            final String castFuncName = CastFunctions.getCastFunc(MinorType.VARCHAR);
            final List<LogicalExpression> castArgs = Lists.newArrayList();
            //input_expr
            castArgs.add(convertToJson);
            // implicitly casting to varchar, since we don't know actual source length, cast to undefined length, which will preserve source length
            castArgs.add(new ValueExpressions.LongExpression(Types.MAX_VARCHAR_LENGTH, null));
            final FunctionCall castCall = new FunctionCall(castFuncName, castArgs, ExpressionPosition.UNKNOWN);
            exprs.add(new NamedExpression(castCall, new FieldReference(field.getPath())));
        } else {
            exprs.add(new NamedExpression(SchemaPath.getSimplePath(field.getPath()), new FieldReference(field.getPath())));
        }
    }
    return exprs;
}

Example 40 with FieldReference

use of org.apache.drill.common.expression.FieldReference in project drill by apache.

the class HashAggTemplate method setup.

@Override
public void setup(HashAggregate hashAggrConfig, HashTableConfig htConfig, FragmentContext context, OperatorStats stats, BufferAllocator allocator, RecordBatch incoming, HashAggBatch outgoing, LogicalExpression[] valueExprs, List<TypedFieldId> valueFieldIds, TypedFieldId[] groupByOutFieldIds, VectorContainer outContainer) throws SchemaChangeException, ClassTransformationException, IOException {
    if (valueExprs == null || valueFieldIds == null) {
        throw new IllegalArgumentException("Invalid aggr value exprs or workspace variables.");
    }
    if (valueFieldIds.size() < valueExprs.length) {
        throw new IllegalArgumentException("Wrong number of workspace variables.");
    }
    //    this.context = context;
    this.stats = stats;
    this.allocator = allocator;
    this.incoming = incoming;
    //    this.schema = incoming.getSchema();
    this.outgoing = outgoing;
    this.outContainer = outContainer;
    // TODO:  This functionality will be added later.
    if (hashAggrConfig.getGroupByExprs().size() == 0) {
        throw new IllegalArgumentException("Currently, hash aggregation is only applicable if there are group-by " + "expressions.");
    }
    this.htIdxHolder = new IndexPointer();
    this.outStartIdxHolder = new IndexPointer();
    this.outNumRecordsHolder = new IndexPointer();
    materializedValueFields = new MaterializedField[valueFieldIds.size()];
    if (valueFieldIds.size() > 0) {
        int i = 0;
        FieldReference ref = new FieldReference("dummy", ExpressionPosition.UNKNOWN, valueFieldIds.get(0).getIntermediateType());
        for (TypedFieldId id : valueFieldIds) {
            materializedValueFields[i++] = MaterializedField.create(ref.getAsNamePart().getName(), id.getIntermediateType());
        }
    }
    ChainedHashTable ht = new ChainedHashTable(htConfig, context, allocator, incoming, null, /* no incoming probe */
    outgoing);
    this.htable = ht.createAndSetupHashTable(groupByOutFieldIds);
    numGroupByOutFields = groupByOutFieldIds.length;
    batchHolders = new ArrayList<BatchHolder>();
    // First BatchHolder is created when the first put request is received.
    doSetup(incoming);
}