Examples with StructTypeInfo org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo used on opensource projects

Example 76 with StructTypeInfo

use of org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo in project hive by apache.

the class BinarySortableSerDe method deserialize.

static Object deserialize(InputByteBuffer buffer, TypeInfo type, boolean invert, byte nullMarker, byte notNullMarker, Object reuse) throws IOException {
    // Is this field a null?
    byte isNull = buffer.read(invert);
    if (isNull == nullMarker) {
        return null;
    }
    assert (isNull == notNullMarker);
    switch(type.getCategory()) {
        case PRIMITIVE:
            {
                PrimitiveTypeInfo ptype = (PrimitiveTypeInfo) type;
                switch(ptype.getPrimitiveCategory()) {
                    case VOID:
                        {
                            return null;
                        }
                    case BOOLEAN:
                        {
                            BooleanWritable r = reuse == null ? new BooleanWritable() : (BooleanWritable) reuse;
                            byte b = buffer.read(invert);
                            assert (b == 1 || b == 2);
                            r.set(b == 2);
                            return r;
                        }
                    case BYTE:
                        {
                            ByteWritable r = reuse == null ? new ByteWritable() : (ByteWritable) reuse;
                            r.set((byte) (buffer.read(invert) ^ 0x80));
                            return r;
                        }
                    case SHORT:
                        {
                            ShortWritable r = reuse == null ? new ShortWritable() : (ShortWritable) reuse;
                            int v = buffer.read(invert) ^ 0x80;
                            v = (v << 8) + (buffer.read(invert) & 0xff);
                            r.set((short) v);
                            return r;
                        }
                    case INT:
                        {
                            IntWritable r = reuse == null ? new IntWritable() : (IntWritable) reuse;
                            r.set(deserializeInt(buffer, invert));
                            return r;
                        }
                    case LONG:
                        {
                            LongWritable r = reuse == null ? new LongWritable() : (LongWritable) reuse;
                            r.set(deserializeLong(buffer, invert));
                            return r;
                        }
                    case FLOAT:
                        {
                            FloatWritable r = reuse == null ? new FloatWritable() : (FloatWritable) reuse;
                            int v = 0;
                            for (int i = 0; i < 4; i++) {
                                v = (v << 8) + (buffer.read(invert) & 0xff);
                            }
                            if ((v & (1 << 31)) == 0) {
                                // negative number, flip all bits
                                v = ~v;
                            } else {
                                // positive number, flip the first bit
                                v = v ^ (1 << 31);
                            }
                            r.set(Float.intBitsToFloat(v));
                            return r;
                        }
                    case DOUBLE:
                        {
                            DoubleWritable r = reuse == null ? new DoubleWritable() : (DoubleWritable) reuse;
                            long v = 0;
                            for (int i = 0; i < 8; i++) {
                                v = (v << 8) + (buffer.read(invert) & 0xff);
                            }
                            if ((v & (1L << 63)) == 0) {
                                // negative number, flip all bits
                                v = ~v;
                            } else {
                                // positive number, flip the first bit
                                v = v ^ (1L << 63);
                            }
                            r.set(Double.longBitsToDouble(v));
                            return r;
                        }
                    case STRING:
                        {
                            Text r = reuse == null ? new Text() : (Text) reuse;
                            return deserializeText(buffer, invert, r);
                        }
                    case CHAR:
                        {
                            HiveCharWritable r = reuse == null ? new HiveCharWritable() : (HiveCharWritable) reuse;
                            // Use internal text member to read value
                            deserializeText(buffer, invert, r.getTextValue());
                            r.enforceMaxLength(getCharacterMaxLength(type));
                            return r;
                        }
                    case VARCHAR:
                        {
                            HiveVarcharWritable r = reuse == null ? new HiveVarcharWritable() : (HiveVarcharWritable) reuse;
                            // Use HiveVarchar's internal Text member to read the value.
                            deserializeText(buffer, invert, r.getTextValue());
                            // If we cache helper data for deserialization we could avoid having
                            // to call getVarcharMaxLength() on every deserialize call.
                            r.enforceMaxLength(getCharacterMaxLength(type));
                            return r;
                        }
                    case BINARY:
                        {
                            BytesWritable bw = new BytesWritable();
                            // Get the actual length first
                            int start = buffer.tell();
                            int length = 0;
                            do {
                                byte b = buffer.read(invert);
                                if (b == 0) {
                                    // end of string
                                    break;
                                }
                                if (b == 1) {
                                    // the last char is an escape char. read the actual char
                                    buffer.read(invert);
                                }
                                length++;
                            } while (true);
                            if (length == buffer.tell() - start) {
                                // No escaping happened, so we are already done.
                                bw.set(buffer.getData(), start, length);
                            } else {
                                // Escaping happened, we need to copy byte-by-byte.
                                // 1. Set the length first.
                                bw.set(buffer.getData(), start, length);
                                // 2. Reset the pointer.
                                buffer.seek(start);
                                // 3. Copy the data.
                                byte[] rdata = bw.getBytes();
                                for (int i = 0; i < length; i++) {
                                    byte b = buffer.read(invert);
                                    if (b == 1) {
                                        // The last char is an escape char, read the actual char.
                                        // The serialization format escape \0 to \1, and \1 to \2,
                                        // to make sure the string is null-terminated.
                                        b = (byte) (buffer.read(invert) - 1);
                                    }
                                    rdata[i] = b;
                                }
                                // 4. Read the null terminator.
                                byte b = buffer.read(invert);
                                assert (b == 0);
                            }
                            return bw;
                        }
                    case DATE:
                        {
                            DateWritableV2 d = reuse == null ? new DateWritableV2() : (DateWritableV2) reuse;
                            d.set(deserializeInt(buffer, invert));
                            return d;
                        }
                    case TIMESTAMP:
                        TimestampWritableV2 t = (reuse == null ? new TimestampWritableV2() : (TimestampWritableV2) reuse);
                        byte[] bytes = new byte[TimestampWritableV2.BINARY_SORTABLE_LENGTH];
                        for (int i = 0; i < bytes.length; i++) {
                            bytes[i] = buffer.read(invert);
                        }
                        t.setBinarySortable(bytes, 0);
                        return t;
                    case TIMESTAMPLOCALTZ:
                        TimestampLocalTZWritable tstz = (reuse == null ? new TimestampLocalTZWritable() : (TimestampLocalTZWritable) reuse);
                        byte[] data = new byte[TimestampLocalTZWritable.BINARY_SORTABLE_LENGTH];
                        for (int i = 0; i < data.length; i++) {
                            data[i] = buffer.read(invert);
                        }
                        // Across MR process boundary tz is normalized and stored in type
                        // and is not carried in data for each row.
                        tstz.fromBinarySortable(data, 0, ((TimestampLocalTZTypeInfo) type).timeZone());
                        return tstz;
                    case INTERVAL_YEAR_MONTH:
                        {
                            HiveIntervalYearMonthWritable i = reuse == null ? new HiveIntervalYearMonthWritable() : (HiveIntervalYearMonthWritable) reuse;
                            i.set(deserializeInt(buffer, invert));
                            return i;
                        }
                    case INTERVAL_DAY_TIME:
                        {
                            HiveIntervalDayTimeWritable i = reuse == null ? new HiveIntervalDayTimeWritable() : (HiveIntervalDayTimeWritable) reuse;
                            long totalSecs = deserializeLong(buffer, invert);
                            int nanos = deserializeInt(buffer, invert);
                            i.set(totalSecs, nanos);
                            return i;
                        }
                    case DECIMAL:
                        {
                            // See serialization of decimal for explanation (below)
                            HiveDecimalWritable bdw = (reuse == null ? new HiveDecimalWritable() : (HiveDecimalWritable) reuse);
                            int b = buffer.read(invert) - 1;
                            assert (b == 1 || b == -1 || b == 0);
                            boolean positive = b != -1;
                            int factor = buffer.read(invert) ^ 0x80;
                            for (int i = 0; i < 3; i++) {
                                factor = (factor << 8) + (buffer.read(invert) & 0xff);
                            }
                            if (!positive) {
                                factor = -factor;
                            }
                            int start = buffer.tell();
                            int length = 0;
                            do {
                                b = buffer.read(positive ? invert : !invert);
                                assert (b != 1);
                                if (b == 0) {
                                    // end of digits
                                    break;
                                }
                                length++;
                            } while (true);
                            final byte[] decimalBuffer = new byte[length];
                            buffer.seek(start);
                            for (int i = 0; i < length; ++i) {
                                decimalBuffer[i] = buffer.read(positive ? invert : !invert);
                            }
                            // read the null byte again
                            buffer.read(positive ? invert : !invert);
                            String digits = new String(decimalBuffer, 0, length, decimalCharSet);
                            BigInteger bi = new BigInteger(digits);
                            HiveDecimal bd = HiveDecimal.create(bi).scaleByPowerOfTen(factor - length);
                            if (!positive) {
                                bd = bd.negate();
                            }
                            bdw.set(bd);
                            return bdw;
                        }
                    default:
                        {
                            throw new RuntimeException("Unrecognized type: " + ptype.getPrimitiveCategory());
                        }
                }
            }
        case LIST:
            {
                ListTypeInfo ltype = (ListTypeInfo) type;
                TypeInfo etype = ltype.getListElementTypeInfo();
                // Create the list if needed
                ArrayList<Object> r = reuse == null ? new ArrayList<Object>() : (ArrayList<Object>) reuse;
                // Read the list
                int size = 0;
                while (true) {
                    int more = buffer.read(invert);
                    if (more == 0) {
                        // \0 to terminate
                        break;
                    }
                    // \1 followed by each element
                    assert (more == 1);
                    if (size == r.size()) {
                        r.add(null);
                    }
                    r.set(size, deserialize(buffer, etype, invert, nullMarker, notNullMarker, r.get(size)));
                    size++;
                }
                // Remove additional elements if the list is reused
                while (r.size() > size) {
                    r.remove(r.size() - 1);
                }
                return r;
            }
        case MAP:
            {
                MapTypeInfo mtype = (MapTypeInfo) type;
                TypeInfo ktype = mtype.getMapKeyTypeInfo();
                TypeInfo vtype = mtype.getMapValueTypeInfo();
                // Create the map if needed
                Map<Object, Object> r;
                if (reuse == null || reuse.getClass() != LinkedHashMap.class) {
                    r = new LinkedHashMap<Object, Object>();
                } else {
                    r = (Map<Object, Object>) reuse;
                    r.clear();
                }
                while (true) {
                    int more = buffer.read(invert);
                    if (more == 0) {
                        // \0 to terminate
                        break;
                    }
                    // \1 followed by each key and then each value
                    assert (more == 1);
                    Object k = deserialize(buffer, ktype, invert, nullMarker, notNullMarker, null);
                    Object v = deserialize(buffer, vtype, invert, nullMarker, notNullMarker, null);
                    r.put(k, v);
                }
                return r;
            }
        case STRUCT:
            {
                StructTypeInfo stype = (StructTypeInfo) type;
                List<TypeInfo> fieldTypes = stype.getAllStructFieldTypeInfos();
                int size = fieldTypes.size();
                // Create the struct if needed
                ArrayList<Object> r = reuse == null ? new ArrayList<Object>(size) : (ArrayList<Object>) reuse;
                assert (r.size() <= size);
                // Set the size of the struct
                while (r.size() < size) {
                    r.add(null);
                }
                // Read one field by one field
                for (int eid = 0; eid < size; eid++) {
                    r.set(eid, deserialize(buffer, fieldTypes.get(eid), invert, nullMarker, notNullMarker, r.get(eid)));
                }
                return r;
            }
        case UNION:
            {
                UnionTypeInfo utype = (UnionTypeInfo) type;
                StandardUnion r = reuse == null ? new StandardUnion() : (StandardUnion) reuse;
                // Read the tag
                byte tag = buffer.read(invert);
                r.setTag(tag);
                r.setObject(deserialize(buffer, utype.getAllUnionObjectTypeInfos().get(tag), invert, nullMarker, notNullMarker, null));
                return r;
            }
        default:
            {
                throw new RuntimeException("Unrecognized type: " + type.getCategory());
            }
    }
}

Example 77 with StructTypeInfo

use of org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo in project hive by apache.

the class AvroObjectInspectorGenerator method createObjectInspectorWorker.

private ObjectInspector createObjectInspectorWorker(TypeInfo ti) throws SerDeException {
    // at deserialization and the object inspector will never see the actual union.
    if (!supportedCategories(ti)) {
        throw new AvroSerdeException("Don't yet support this type: " + ti);
    }
    ObjectInspector result;
    switch(ti.getCategory()) {
        case PRIMITIVE:
            PrimitiveTypeInfo pti = (PrimitiveTypeInfo) ti;
            result = PrimitiveObjectInspectorFactory.getPrimitiveJavaObjectInspector(pti);
            break;
        case STRUCT:
            StructTypeInfo sti = (StructTypeInfo) ti;
            ArrayList<ObjectInspector> ois = new ArrayList<ObjectInspector>(sti.getAllStructFieldTypeInfos().size());
            for (TypeInfo typeInfo : sti.getAllStructFieldTypeInfos()) {
                ois.add(createObjectInspectorWorker(typeInfo));
            }
            result = ObjectInspectorFactory.getStandardStructObjectInspector(sti.getAllStructFieldNames(), ois);
            break;
        case MAP:
            MapTypeInfo mti = (MapTypeInfo) ti;
            result = ObjectInspectorFactory.getStandardMapObjectInspector(PrimitiveObjectInspectorFactory.getPrimitiveJavaObjectInspector(PrimitiveObjectInspector.PrimitiveCategory.STRING), createObjectInspectorWorker(mti.getMapValueTypeInfo()));
            break;
        case LIST:
            ListTypeInfo ati = (ListTypeInfo) ti;
            result = ObjectInspectorFactory.getStandardListObjectInspector(createObjectInspectorWorker(ati.getListElementTypeInfo()));
            break;
        case UNION:
            UnionTypeInfo uti = (UnionTypeInfo) ti;
            List<TypeInfo> allUnionObjectTypeInfos = uti.getAllUnionObjectTypeInfos();
            List<ObjectInspector> unionObjectInspectors = new ArrayList<ObjectInspector>(allUnionObjectTypeInfos.size());
            for (TypeInfo typeInfo : allUnionObjectTypeInfos) {
                unionObjectInspectors.add(createObjectInspectorWorker(typeInfo));
            }
            result = ObjectInspectorFactory.getStandardUnionObjectInspector(unionObjectInspectors);
            break;
        default:
            throw new AvroSerdeException("No Hive categories matched: " + ti);
    }
    return result;
}

Example 78 with StructTypeInfo

use of org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo in project hive by apache.

the class AvroSerializer method serializeStruct.

private Object serializeStruct(StructTypeInfo typeInfo, StructObjectInspector ssoi, Object o, Schema schema) throws AvroSerdeException {
    int size = schema.getFields().size();
    List<? extends StructField> allStructFieldRefs = ssoi.getAllStructFieldRefs();
    List<Object> structFieldsDataAsList = ssoi.getStructFieldsDataAsList(o);
    GenericData.Record record = new GenericData.Record(schema);
    List<TypeInfo> allStructFieldTypeInfos = typeInfo.getAllStructFieldTypeInfos();
    for (int i = 0; i < size; i++) {
        Field field = schema.getFields().get(i);
        TypeInfo colTypeInfo = allStructFieldTypeInfos.get(i);
        StructField structFieldRef = allStructFieldRefs.get(i);
        Object structFieldData = structFieldsDataAsList.get(i);
        ObjectInspector fieldOI = structFieldRef.getFieldObjectInspector();
        Object val = serialize(colTypeInfo, fieldOI, structFieldData, field.schema());
        record.put(field.name(), val);
    }
    return record;
}

Example 79 with StructTypeInfo

use of org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo in project hive by apache.

the class TypeInfoToSchema method createAvroRecord.

private Schema createAvroRecord(TypeInfo typeInfo) {
    List<Schema.Field> childFields = new ArrayList<Schema.Field>();
    final List<String> allStructFieldNames = ((StructTypeInfo) typeInfo).getAllStructFieldNames();
    final List<TypeInfo> allStructFieldTypeInfos = ((StructTypeInfo) typeInfo).getAllStructFieldTypeInfos();
    if (allStructFieldNames.size() != allStructFieldTypeInfos.size()) {
        throw new IllegalArgumentException("Failed to generate avro schema from hive schema. " + "name and column type differs. names = " + allStructFieldNames + ", types = " + allStructFieldTypeInfos);
    }
    for (int i = 0; i < allStructFieldNames.size(); ++i) {
        final TypeInfo childTypeInfo = allStructFieldTypeInfos.get(i);
        final Schema.Field grandChildSchemaField = createAvroField(allStructFieldNames.get(i), childTypeInfo, childTypeInfo.toString());
        final List<Schema.Field> grandChildFields = getFields(grandChildSchemaField);
        childFields.addAll(grandChildFields);
    }
    Schema recordSchema = Schema.createRecord("record_" + recordCounter, typeInfo.toString(), null, false);
    ++recordCounter;
    recordSchema.setFields(childFields);
    return recordSchema;
}

Example 80 with StructTypeInfo

use of org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo in project hive by apache.

the class VectorSerializeRow method createField.

private Field createField(TypeInfo typeInfo) {
    final Field field = new Field();
    final Category category = typeInfo.getCategory();
    field.category = category;
    field.typeInfo = typeInfo;
    if (category == Category.PRIMITIVE) {
        field.isPrimitive = true;
        field.primitiveCategory = ((PrimitiveTypeInfo) typeInfo).getPrimitiveCategory();
        switch(field.primitiveCategory) {
            case BOOLEAN:
                field.writer = new VectorSerializeBooleanWriter();
                break;
            case BYTE:
                field.writer = new VectorSerializeByteWriter();
                break;
            case SHORT:
                field.writer = new VectorSerializeShortWriter();
                break;
            case INT:
                field.writer = new VectorSerializeIntWriter();
                break;
            case LONG:
                field.writer = new VectorSerializeLongWriter();
                break;
            case DATE:
                field.writer = new VectorSerializeDateWriter();
                break;
            case TIMESTAMP:
                field.writer = new VectorSerializeTimestampWriter();
                break;
            case FLOAT:
                field.writer = new VectorSerializeFloatWriter();
                break;
            case DOUBLE:
                field.writer = new VectorSerializeDoubleWriter();
                break;
            case STRING:
            case CHAR:
            case VARCHAR:
                field.writer = new VectorSerializeStringWriter();
                break;
            case BINARY:
                field.writer = new VectorSerializeBinaryWriter();
                break;
            case DECIMAL:
                field.writer = new VectorSerializeDecimalWriter();
                break;
            case INTERVAL_YEAR_MONTH:
                field.writer = new VectorSerializeHiveIntervalYearMonthWriter();
                break;
            case INTERVAL_DAY_TIME:
                field.writer = new VectorSerializeHiveIntervalDayTimeWriter();
                break;
            default:
                throw new RuntimeException("Unexpected primitive category " + field.primitiveCategory);
        }
    } else {
        field.isPrimitive = false;
        field.objectInspector = TypeInfoUtils.getStandardJavaObjectInspectorFromTypeInfo(typeInfo);
        switch(category) {
            case LIST:
                field.children = new Field[1];
                field.children[0] = createField(((ListTypeInfo) typeInfo).getListElementTypeInfo());
                field.writer = new VectorSerializeListWriter();
                break;
            case MAP:
                field.children = new Field[2];
                field.children[0] = createField(((MapTypeInfo) typeInfo).getMapKeyTypeInfo());
                field.children[1] = createField(((MapTypeInfo) typeInfo).getMapValueTypeInfo());
                field.writer = new VectorSerializeMapWriter();
                break;
            case STRUCT:
                StructTypeInfo structTypeInfo = (StructTypeInfo) typeInfo;
                List<TypeInfo> fieldTypeInfos = structTypeInfo.getAllStructFieldTypeInfos();
                field.children = createFields(fieldTypeInfos.toArray(new TypeInfo[fieldTypeInfos.size()]));
                field.writer = new VectorSerializeStructWriter();
                break;
            case UNION:
                UnionTypeInfo unionTypeInfo = (UnionTypeInfo) typeInfo;
                List<TypeInfo> objectTypeInfos = unionTypeInfo.getAllUnionObjectTypeInfos();
                field.children = createFields(objectTypeInfos.toArray(new TypeInfo[objectTypeInfos.size()]));
                field.writer = new VectorSerializeUnionWriter();
                break;
            default:
                throw new RuntimeException();
        }
        field.count = field.children.length;
    }
    return field;
}