Examples with StructField org.apache.hadoop.hive.serde2.objectinspector.StructField used on opensource projects

Example 86 with StructField

use of org.apache.hadoop.hive.serde2.objectinspector.StructField in project hive by apache.

the class VectorExtractRow method init.

/*
   * Initialize using an StructObjectInspector and a column projection list.
   */
public void init(StructObjectInspector structObjectInspector, List<Integer> projectedColumns) throws HiveException {
    List<? extends StructField> fields = structObjectInspector.getAllStructFieldRefs();
    final int count = fields.size();
    allocateArrays(count);
    for (int i = 0; i < count; i++) {
        int projectionColumnNum = projectedColumns.get(i);
        StructField field = fields.get(i);
        ObjectInspector fieldInspector = field.getFieldObjectInspector();
        TypeInfo typeInfo = TypeInfoUtils.getTypeInfoFromTypeString(fieldInspector.getTypeName());
        initEntry(i, projectionColumnNum, typeInfo);
    }
}

Example 87 with StructField

use of org.apache.hadoop.hive.serde2.objectinspector.StructField in project hive by apache.

the class VectorExtractRow method extractRowColumn.

public Object extractRowColumn(ColumnVector colVector, TypeInfo typeInfo, ObjectInspector objectInspector, int batchIndex) {
    if (colVector == null) {
        // may ask for them..
        return null;
    }
    final int adjustedIndex = (colVector.isRepeating ? 0 : batchIndex);
    if (!colVector.noNulls && colVector.isNull[adjustedIndex]) {
        return null;
    }
    final Category category = typeInfo.getCategory();
    switch(category) {
        case PRIMITIVE:
            {
                final PrimitiveTypeInfo primitiveTypeInfo = (PrimitiveTypeInfo) typeInfo;
                final PrimitiveCategory primitiveCategory = primitiveTypeInfo.getPrimitiveCategory();
                final Writable primitiveWritable = VectorizedBatchUtil.getPrimitiveWritable(primitiveCategory);
                switch(primitiveCategory) {
                    case VOID:
                        return null;
                    case BOOLEAN:
                        ((BooleanWritable) primitiveWritable).set(((LongColumnVector) colVector).vector[adjustedIndex] == 0 ? false : true);
                        return primitiveWritable;
                    case BYTE:
                        ((ByteWritable) primitiveWritable).set((byte) ((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case SHORT:
                        ((ShortWritable) primitiveWritable).set((short) ((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case INT:
                        ((IntWritable) primitiveWritable).set((int) ((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case LONG:
                        ((LongWritable) primitiveWritable).set(((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case TIMESTAMP:
                        ((TimestampWritable) primitiveWritable).set(((TimestampColumnVector) colVector).asScratchTimestamp(adjustedIndex));
                        return primitiveWritable;
                    case DATE:
                        ((DateWritable) primitiveWritable).set((int) ((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case FLOAT:
                        ((FloatWritable) primitiveWritable).set((float) ((DoubleColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case DOUBLE:
                        ((DoubleWritable) primitiveWritable).set(((DoubleColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case BINARY:
                        {
                            final BytesColumnVector bytesColVector = ((BytesColumnVector) colVector);
                            final byte[] bytes = bytesColVector.vector[adjustedIndex];
                            final int start = bytesColVector.start[adjustedIndex];
                            final int length = bytesColVector.length[adjustedIndex];
                            if (bytesColVector.isRepeating) {
                                if (!bytesColVector.isNull[0] && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            } else {
                                if ((bytesColVector.noNulls || !bytesColVector.isNull[batchIndex]) && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            }
                            BytesWritable bytesWritable = (BytesWritable) primitiveWritable;
                            bytesWritable.set(bytes, start, length);
                            return primitiveWritable;
                        }
                    case STRING:
                        {
                            final BytesColumnVector bytesColVector = ((BytesColumnVector) colVector);
                            final byte[] bytes = bytesColVector.vector[adjustedIndex];
                            final int start = bytesColVector.start[adjustedIndex];
                            final int length = bytesColVector.length[adjustedIndex];
                            if (bytesColVector.isRepeating) {
                                if (!bytesColVector.isNull[0] && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            } else {
                                if ((bytesColVector.noNulls || !bytesColVector.isNull[batchIndex]) && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            }
                            // Use org.apache.hadoop.io.Text as our helper to go from byte[] to String.
                            ((Text) primitiveWritable).set(bytes, start, length);
                            return primitiveWritable;
                        }
                    case VARCHAR:
                        {
                            final BytesColumnVector bytesColVector = ((BytesColumnVector) colVector);
                            final byte[] bytes = bytesColVector.vector[adjustedIndex];
                            final int start = bytesColVector.start[adjustedIndex];
                            final int length = bytesColVector.length[adjustedIndex];
                            if (bytesColVector.isRepeating) {
                                if (!bytesColVector.isNull[0] && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            } else {
                                if ((bytesColVector.noNulls || !bytesColVector.isNull[batchIndex]) && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            }
                            final int adjustedLength = StringExpr.truncate(bytes, start, length, ((VarcharTypeInfo) primitiveTypeInfo).getLength());
                            final HiveVarcharWritable hiveVarcharWritable = (HiveVarcharWritable) primitiveWritable;
                            hiveVarcharWritable.set(new String(bytes, start, adjustedLength, Charsets.UTF_8), -1);
                            return primitiveWritable;
                        }
                    case CHAR:
                        {
                            final BytesColumnVector bytesColVector = ((BytesColumnVector) colVector);
                            final byte[] bytes = bytesColVector.vector[adjustedIndex];
                            final int start = bytesColVector.start[adjustedIndex];
                            final int length = bytesColVector.length[adjustedIndex];
                            if (bytesColVector.isRepeating) {
                                if (!bytesColVector.isNull[0] && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            } else {
                                if ((bytesColVector.noNulls || !bytesColVector.isNull[batchIndex]) && bytes == null) {
                                    nullBytesReadError(primitiveCategory, batchIndex);
                                }
                            }
                            final int adjustedLength = StringExpr.rightTrimAndTruncate(bytes, start, length, ((CharTypeInfo) primitiveTypeInfo).getLength());
                            final HiveCharWritable hiveCharWritable = (HiveCharWritable) primitiveWritable;
                            hiveCharWritable.set(new String(bytes, start, adjustedLength, Charsets.UTF_8), ((CharTypeInfo) primitiveTypeInfo).getLength());
                            return primitiveWritable;
                        }
                    case DECIMAL:
                        if (colVector instanceof Decimal64ColumnVector) {
                            Decimal64ColumnVector dec32ColVector = (Decimal64ColumnVector) colVector;
                            ((HiveDecimalWritable) primitiveWritable).deserialize64(dec32ColVector.vector[adjustedIndex], dec32ColVector.scale);
                        } else {
                            // The HiveDecimalWritable set method will quickly copy the deserialized decimal writable fields.
                            ((HiveDecimalWritable) primitiveWritable).set(((DecimalColumnVector) colVector).vector[adjustedIndex]);
                        }
                        return primitiveWritable;
                    case INTERVAL_YEAR_MONTH:
                        ((HiveIntervalYearMonthWritable) primitiveWritable).set((int) ((LongColumnVector) colVector).vector[adjustedIndex]);
                        return primitiveWritable;
                    case INTERVAL_DAY_TIME:
                        ((HiveIntervalDayTimeWritable) primitiveWritable).set(((IntervalDayTimeColumnVector) colVector).asScratchIntervalDayTime(adjustedIndex));
                        return primitiveWritable;
                    default:
                        throw new RuntimeException("Primitive category " + primitiveCategory.name() + " not supported");
                }
            }
        case LIST:
            {
                final ListColumnVector listColumnVector = (ListColumnVector) colVector;
                final ListTypeInfo listTypeInfo = (ListTypeInfo) typeInfo;
                final ListObjectInspector listObjectInspector = (ListObjectInspector) objectInspector;
                final int offset = (int) listColumnVector.offsets[adjustedIndex];
                final int size = (int) listColumnVector.lengths[adjustedIndex];
                final List list = new ArrayList();
                for (int i = 0; i < size; i++) {
                    list.add(extractRowColumn(listColumnVector.child, listTypeInfo.getListElementTypeInfo(), listObjectInspector.getListElementObjectInspector(), offset + i));
                }
                return list;
            }
        case MAP:
            {
                final MapColumnVector mapColumnVector = (MapColumnVector) colVector;
                final MapTypeInfo mapTypeInfo = (MapTypeInfo) typeInfo;
                final MapObjectInspector mapObjectInspector = (MapObjectInspector) objectInspector;
                final int offset = (int) mapColumnVector.offsets[adjustedIndex];
                final int size = (int) mapColumnVector.lengths[adjustedIndex];
                final Map map = new HashMap();
                for (int i = 0; i < size; i++) {
                    final Object key = extractRowColumn(mapColumnVector.keys, mapTypeInfo.getMapKeyTypeInfo(), mapObjectInspector.getMapKeyObjectInspector(), offset + i);
                    final Object value = extractRowColumn(mapColumnVector.values, mapTypeInfo.getMapValueTypeInfo(), mapObjectInspector.getMapValueObjectInspector(), offset + i);
                    map.put(key, value);
                }
                return map;
            }
        case STRUCT:
            {
                final StructColumnVector structColumnVector = (StructColumnVector) colVector;
                final StructTypeInfo structTypeInfo = (StructTypeInfo) typeInfo;
                final StandardStructObjectInspector structInspector = (StandardStructObjectInspector) objectInspector;
                final List<TypeInfo> fieldTypeInfos = structTypeInfo.getAllStructFieldTypeInfos();
                final int size = fieldTypeInfos.size();
                final List<? extends StructField> structFields = structInspector.getAllStructFieldRefs();
                final Object struct = structInspector.create();
                for (int i = 0; i < size; i++) {
                    final StructField structField = structFields.get(i);
                    final TypeInfo fieldTypeInfo = fieldTypeInfos.get(i);
                    final Object value = extractRowColumn(structColumnVector.fields[i], fieldTypeInfo, structField.getFieldObjectInspector(), adjustedIndex);
                    structInspector.setStructFieldData(struct, structField, value);
                }
                return struct;
            }
        case UNION:
            {
                final UnionTypeInfo unionTypeInfo = (UnionTypeInfo) typeInfo;
                final List<TypeInfo> objectTypeInfos = unionTypeInfo.getAllUnionObjectTypeInfos();
                final UnionObjectInspector unionInspector = (UnionObjectInspector) objectInspector;
                final List<ObjectInspector> unionInspectors = unionInspector.getObjectInspectors();
                final UnionColumnVector unionColumnVector = (UnionColumnVector) colVector;
                final byte tag = (byte) unionColumnVector.tags[adjustedIndex];
                final Object object = extractRowColumn(unionColumnVector.fields[tag], objectTypeInfos.get(tag), unionInspectors.get(tag), adjustedIndex);
                final StandardUnion standardUnion = new StandardUnion();
                standardUnion.setTag(tag);
                standardUnion.setObject(object);
                return standardUnion;
            }
        default:
            throw new RuntimeException("Category " + category.name() + " not supported");
    }
}

Example 88 with StructField

use of org.apache.hadoop.hive.serde2.objectinspector.StructField in project hive by apache.

the class VectorAssignRow method init.

/*
   * Initialize using an StructObjectInspector.
   * No projection -- the column range 0 .. fields.size()-1
   */
public void init(StructObjectInspector structObjectInspector) throws HiveException {
    final List<? extends StructField> fields = structObjectInspector.getAllStructFieldRefs();
    final int count = fields.size();
    allocateArrays(count);
    for (int i = 0; i < count; i++) {
        final StructField field = fields.get(i);
        final ObjectInspector fieldInspector = field.getFieldObjectInspector();
        final TypeInfo typeInfo = TypeInfoUtils.getTypeInfoFromTypeString(fieldInspector.getTypeName());
        initTargetEntry(i, i, typeInfo);
    }
}

Example 89 with StructField

use of org.apache.hadoop.hive.serde2.objectinspector.StructField in project hive by apache.

the class Utilities method constructVectorizedReduceRowOI.

/**
 * Create row key and value object inspectors for reduce vectorization.
 * The row object inspector used by ReduceWork needs to be a **standard**
 * struct object inspector, not just any struct object inspector.
 * @param keyInspector
 * @param valueInspector
 * @return OI
 * @throws HiveException
 */
public static StandardStructObjectInspector constructVectorizedReduceRowOI(StructObjectInspector keyInspector, StructObjectInspector valueInspector) throws HiveException {
    ArrayList<String> colNames = new ArrayList<String>();
    ArrayList<ObjectInspector> ois = new ArrayList<ObjectInspector>();
    List<? extends StructField> fields = keyInspector.getAllStructFieldRefs();
    for (StructField field : fields) {
        colNames.add(Utilities.ReduceField.KEY.toString() + '.' + field.getFieldName());
        ois.add(field.getFieldObjectInspector());
    }
    fields = valueInspector.getAllStructFieldRefs();
    for (StructField field : fields) {
        colNames.add(Utilities.ReduceField.VALUE.toString() + '.' + field.getFieldName());
        ois.add(field.getFieldObjectInspector());
    }
    StandardStructObjectInspector rowObjectInspector = ObjectInspectorFactory.getStandardStructObjectInspector(colNames, ois);
    return rowObjectInspector;
}

Example 90 with StructField

use of org.apache.hadoop.hive.serde2.objectinspector.StructField in project hive by apache.

the class GroupByOperator method initializeOp.

@Override
protected void initializeOp(Configuration hconf) throws HiveException {
    super.initializeOp(hconf);
    numRowsInput = 0;
    numRowsHashTbl = 0;
    heartbeatInterval = HiveConf.getIntVar(hconf, HiveConf.ConfVars.HIVESENDHEARTBEAT);
    countAfterReport = 0;
    ObjectInspector rowInspector = inputObjInspectors[0];
    // init keyFields
    int numKeys = conf.getKeys().size();
    keyFields = new ExprNodeEvaluator[numKeys];
    keyObjectInspectors = new ObjectInspector[numKeys];
    currentKeyObjectInspectors = new ObjectInspector[numKeys];
    for (int i = 0; i < numKeys; i++) {
        keyFields[i] = ExprNodeEvaluatorFactory.get(conf.getKeys().get(i), hconf);
        keyObjectInspectors[i] = keyFields[i].initialize(rowInspector);
        currentKeyObjectInspectors[i] = ObjectInspectorUtils.getStandardObjectInspector(keyObjectInspectors[i], ObjectInspectorCopyOption.WRITABLE);
    }
    // Initialize the constants for the grouping sets, so that they can be re-used for
    // each row
    groupingSetsPresent = conf.isGroupingSetsPresent();
    if (groupingSetsPresent) {
        groupingSets = conf.getListGroupingSets();
        groupingSetsPosition = conf.getGroupingSetPosition();
        newKeysGroupingSets = new LongWritable[groupingSets.size()];
        groupingSetsBitSet = new FastBitSet[groupingSets.size()];
        int pos = 0;
        for (Long groupingSet : groupingSets) {
            // Create the mapping corresponding to the grouping set
            newKeysGroupingSets[pos] = new LongWritable(groupingSet);
            groupingSetsBitSet[pos] = groupingSet2BitSet(groupingSet, groupingSetsPosition);
            pos++;
        }
    }
    // initialize unionExpr for reduce-side
    // reduce KEY has union field as the last field if there are distinct
    // aggregates in group-by.
    List<? extends StructField> sfs = ((StructObjectInspector) rowInspector).getAllStructFieldRefs();
    if (sfs.size() > 0) {
        StructField keyField = sfs.get(0);
        if (keyField.getFieldName().toUpperCase().equals(Utilities.ReduceField.KEY.name())) {
            ObjectInspector keyObjInspector = keyField.getFieldObjectInspector();
            if (keyObjInspector instanceof StructObjectInspector) {
                List<? extends StructField> keysfs = ((StructObjectInspector) keyObjInspector).getAllStructFieldRefs();
                if (keysfs.size() > 0) {
                    // the last field is the union field, if any
                    StructField sf = keysfs.get(keysfs.size() - 1);
                    if (sf.getFieldObjectInspector().getCategory().equals(ObjectInspector.Category.UNION)) {
                        unionExprEval = ExprNodeEvaluatorFactory.get(new ExprNodeColumnDesc(TypeInfoUtils.getTypeInfoFromObjectInspector(sf.getFieldObjectInspector()), keyField.getFieldName() + "." + sf.getFieldName(), null, false), hconf);
                        unionExprEval.initialize(rowInspector);
                    }
                }
            }
        }
    }
    // init aggregationParameterFields
    ArrayList<AggregationDesc> aggrs = conf.getAggregators();
    aggregationParameterFields = new ExprNodeEvaluator[aggrs.size()][];
    aggregationParameterObjectInspectors = new ObjectInspector[aggrs.size()][];
    aggregationParameterStandardObjectInspectors = new ObjectInspector[aggrs.size()][];
    aggregationParameterObjects = new Object[aggrs.size()][];
    aggregationIsDistinct = new boolean[aggrs.size()];
    for (int i = 0; i < aggrs.size(); i++) {
        AggregationDesc aggr = aggrs.get(i);
        ArrayList<ExprNodeDesc> parameters = aggr.getParameters();
        aggregationParameterFields[i] = new ExprNodeEvaluator[parameters.size()];
        aggregationParameterObjectInspectors[i] = new ObjectInspector[parameters.size()];
        aggregationParameterStandardObjectInspectors[i] = new ObjectInspector[parameters.size()];
        aggregationParameterObjects[i] = new Object[parameters.size()];
        for (int j = 0; j < parameters.size(); j++) {
            aggregationParameterFields[i][j] = ExprNodeEvaluatorFactory.get(parameters.get(j), hconf);
            aggregationParameterObjectInspectors[i][j] = aggregationParameterFields[i][j].initialize(rowInspector);
            if (unionExprEval != null) {
                String[] names = parameters.get(j).getExprString().split("\\.");
                // parameters of the form : KEY.colx:t.coly
                if (Utilities.ReduceField.KEY.name().equals(names[0]) && names.length > 2) {
                    String name = names[names.length - 2];
                    int tag = Integer.parseInt(name.split("\\:")[1]);
                    if (aggr.getDistinct()) {
                        // is distinct
                        Set<Integer> set = distinctKeyAggrs.get(tag);
                        if (null == set) {
                            set = new HashSet<Integer>();
                            distinctKeyAggrs.put(tag, set);
                        }
                        if (!set.contains(i)) {
                            set.add(i);
                        }
                    } else {
                        Set<Integer> set = nonDistinctKeyAggrs.get(tag);
                        if (null == set) {
                            set = new HashSet<Integer>();
                            nonDistinctKeyAggrs.put(tag, set);
                        }
                        if (!set.contains(i)) {
                            set.add(i);
                        }
                    }
                } else {
                    // will be KEY._COLx or VALUE._COLx
                    if (!nonDistinctAggrs.contains(i)) {
                        nonDistinctAggrs.add(i);
                    }
                }
            } else {
                if (aggr.getDistinct()) {
                    aggregationIsDistinct[i] = true;
                }
            }
            aggregationParameterStandardObjectInspectors[i][j] = ObjectInspectorUtils.getStandardObjectInspector(aggregationParameterObjectInspectors[i][j], ObjectInspectorCopyOption.WRITABLE);
            aggregationParameterObjects[i][j] = null;
        }
        if (parameters.size() == 0) {
            // for ex: count(*)
            if (!nonDistinctAggrs.contains(i)) {
                nonDistinctAggrs.add(i);
            }
        }
    }
    // init aggregationClasses
    aggregationEvaluators = new GenericUDAFEvaluator[conf.getAggregators().size()];
    for (int i = 0; i < aggregationEvaluators.length; i++) {
        AggregationDesc agg = conf.getAggregators().get(i);
        aggregationEvaluators[i] = agg.getGenericUDAFEvaluator();
    }
    MapredContext context = MapredContext.get();
    if (context != null) {
        for (GenericUDAFEvaluator genericUDAFEvaluator : aggregationEvaluators) {
            context.setup(genericUDAFEvaluator);
        }
    }
    // grouping id should be pruned, which is the last of key columns
    // see ColumnPrunerGroupByProc
    outputKeyLength = conf.pruneGroupingSetId() ? keyFields.length - 1 : keyFields.length;
    // init objectInspectors
    ObjectInspector[] objectInspectors = new ObjectInspector[outputKeyLength + aggregationEvaluators.length];
    for (int i = 0; i < outputKeyLength; i++) {
        objectInspectors[i] = currentKeyObjectInspectors[i];
    }
    for (int i = 0; i < aggregationEvaluators.length; i++) {
        objectInspectors[outputKeyLength + i] = aggregationEvaluators[i].init(conf.getAggregators().get(i).getMode(), aggregationParameterObjectInspectors[i]);
    }
    aggregationsParametersLastInvoke = new Object[conf.getAggregators().size()][];
    if ((conf.getMode() != GroupByDesc.Mode.HASH || conf.getBucketGroup()) && (!groupingSetsPresent)) {
        aggregations = newAggregations();
        hashAggr = false;
    } else {
        hashAggregations = new HashMap<KeyWrapper, AggregationBuffer[]>(256);
        aggregations = newAggregations();
        hashAggr = true;
        keyPositionsSize = new ArrayList<Integer>();
        aggrPositions = new List[aggregations.length];
        groupbyMapAggrInterval = HiveConf.getIntVar(hconf, HiveConf.ConfVars.HIVEGROUPBYMAPINTERVAL);
        // compare every groupbyMapAggrInterval rows
        numRowsCompareHashAggr = groupbyMapAggrInterval;
        minReductionHashAggr = HiveConf.getFloatVar(hconf, HiveConf.ConfVars.HIVEMAPAGGRHASHMINREDUCTION);
    }
    List<String> fieldNames = new ArrayList<String>(conf.getOutputColumnNames());
    outputObjInspector = ObjectInspectorFactory.getStandardStructObjectInspector(fieldNames, Arrays.asList(objectInspectors));
    KeyWrapperFactory keyWrapperFactory = new KeyWrapperFactory(keyFields, keyObjectInspectors, currentKeyObjectInspectors);
    newKeys = keyWrapperFactory.getKeyWrapper();
    isTez = HiveConf.getVar(hconf, HiveConf.ConfVars.HIVE_EXECUTION_ENGINE).equals("tez");
    isLlap = LlapDaemonInfo.INSTANCE.isLlap();
    numExecutors = isLlap ? LlapDaemonInfo.INSTANCE.getNumExecutors() : 1;
    firstRow = true;
    // is not known, estimate that based on the number of entries
    if (hashAggr) {
        computeMaxEntriesHashAggr();
    }
    memoryMXBean = ManagementFactory.getMemoryMXBean();
    maxMemory = isTez ? getConf().getMaxMemoryAvailable() : memoryMXBean.getHeapMemoryUsage().getMax();
    memoryThreshold = this.getConf().getMemoryThreshold();
    LOG.info("isTez: {} isLlap: {} numExecutors: {} maxMemory: {}", isTez, isLlap, numExecutors, maxMemory);
}