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

Example 1 with VoidObjectInspector

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

the class GenericUDFDecode method initialize.

@Override
public ObjectInspector initialize(ObjectInspector[] arguments) throws UDFArgumentException {
    if (arguments.length != 2) {
        throw new UDFArgumentLengthException("Decode() requires exactly two arguments");
    }
    if (arguments[0].getCategory() != Category.PRIMITIVE) {
        throw new UDFArgumentTypeException(0, "The first argument to Decode() must be primitive");
    }
    PrimitiveCategory category = ((PrimitiveObjectInspector) arguments[0]).getPrimitiveCategory();
    if (category == PrimitiveCategory.BINARY) {
        bytesOI = (BinaryObjectInspector) arguments[0];
    } else if (category == PrimitiveCategory.VOID) {
        bytesOI = (VoidObjectInspector) arguments[0];
    } else {
        throw new UDFArgumentTypeException(0, "The first argument to Decode() must be binary");
    }
    if (arguments[1].getCategory() != Category.PRIMITIVE) {
        throw new UDFArgumentTypeException(1, "The second argument to Decode() must be primitive");
    }
    charsetOI = (PrimitiveObjectInspector) arguments[1];
    if (PrimitiveGrouping.STRING_GROUP != PrimitiveObjectInspectorUtils.getPrimitiveGrouping(charsetOI.getPrimitiveCategory())) {
        throw new UDFArgumentTypeException(1, "The second argument to Decode() must be from string group");
    }
    // If the character set for decoding is constant, we can optimize that
    if (arguments[1] instanceof ConstantObjectInspector) {
        String charSetName = ((ConstantObjectInspector) arguments[1]).getWritableConstantValue().toString();
        decoder = Charset.forName(charSetName).newDecoder().onMalformedInput(CodingErrorAction.REPORT).onUnmappableCharacter(CodingErrorAction.REPORT);
    }
    return PrimitiveObjectInspectorFactory.javaStringObjectInspector;
}

Example 2 with VoidObjectInspector

use of org.apache.hadoop.hive.serde2.objectinspector.primitive.VoidObjectInspector in project flink by apache.

the class HiveInspectors method getConversion.

/**
 * Get conversion for converting Flink object to Hive object from an ObjectInspector and the
 * corresponding Flink DataType.
 */
public static HiveObjectConversion getConversion(ObjectInspector inspector, LogicalType dataType, HiveShim hiveShim) {
    if (inspector instanceof PrimitiveObjectInspector) {
        HiveObjectConversion conversion;
        if (inspector instanceof BooleanObjectInspector || inspector instanceof StringObjectInspector || inspector instanceof ByteObjectInspector || inspector instanceof ShortObjectInspector || inspector instanceof IntObjectInspector || inspector instanceof LongObjectInspector || inspector instanceof FloatObjectInspector || inspector instanceof DoubleObjectInspector || inspector instanceof BinaryObjectInspector || inspector instanceof VoidObjectInspector) {
            conversion = IdentityConversion.INSTANCE;
        } else if (inspector instanceof DateObjectInspector) {
            conversion = hiveShim::toHiveDate;
        } else if (inspector instanceof TimestampObjectInspector) {
            conversion = hiveShim::toHiveTimestamp;
        } else if (inspector instanceof HiveCharObjectInspector) {
            conversion = o -> o == null ? null : new HiveChar((String) o, ((CharType) dataType).getLength());
        } else if (inspector instanceof HiveVarcharObjectInspector) {
            conversion = o -> o == null ? null : new HiveVarchar((String) o, ((VarCharType) dataType).getLength());
        } else if (inspector instanceof HiveDecimalObjectInspector) {
            conversion = o -> o == null ? null : HiveDecimal.create((BigDecimal) o);
        } else {
            throw new FlinkHiveUDFException("Unsupported primitive object inspector " + inspector.getClass().getName());
        }
        // currently this happens for constant arguments for UDFs
        if (((PrimitiveObjectInspector) inspector).preferWritable()) {
            conversion = new WritableHiveObjectConversion(conversion, hiveShim);
        }
        return conversion;
    }
    if (inspector instanceof ListObjectInspector) {
        HiveObjectConversion eleConvert = getConversion(((ListObjectInspector) inspector).getListElementObjectInspector(), ((ArrayType) dataType).getElementType(), hiveShim);
        return o -> {
            if (o == null) {
                return null;
            }
            Object[] array = (Object[]) o;
            List<Object> result = new ArrayList<>();
            for (Object ele : array) {
                result.add(eleConvert.toHiveObject(ele));
            }
            return result;
        };
    }
    if (inspector instanceof MapObjectInspector) {
        MapObjectInspector mapInspector = (MapObjectInspector) inspector;
        MapType kvType = (MapType) dataType;
        HiveObjectConversion keyConversion = getConversion(mapInspector.getMapKeyObjectInspector(), kvType.getKeyType(), hiveShim);
        HiveObjectConversion valueConversion = getConversion(mapInspector.getMapValueObjectInspector(), kvType.getValueType(), hiveShim);
        return o -> {
            if (o == null) {
                return null;
            }
            Map<Object, Object> map = (Map) o;
            Map<Object, Object> result = new HashMap<>(map.size());
            for (Map.Entry<Object, Object> entry : map.entrySet()) {
                result.put(keyConversion.toHiveObject(entry.getKey()), valueConversion.toHiveObject(entry.getValue()));
            }
            return result;
        };
    }
    if (inspector instanceof StructObjectInspector) {
        StructObjectInspector structInspector = (StructObjectInspector) inspector;
        List<? extends StructField> structFields = structInspector.getAllStructFieldRefs();
        List<RowType.RowField> rowFields = ((RowType) dataType).getFields();
        HiveObjectConversion[] conversions = new HiveObjectConversion[structFields.size()];
        for (int i = 0; i < structFields.size(); i++) {
            conversions[i] = getConversion(structFields.get(i).getFieldObjectInspector(), rowFields.get(i).getType(), hiveShim);
        }
        return o -> {
            if (o == null) {
                return null;
            }
            Row row = (Row) o;
            List<Object> result = new ArrayList<>(row.getArity());
            for (int i = 0; i < row.getArity(); i++) {
                result.add(conversions[i].toHiveObject(row.getField(i)));
            }
            return result;
        };
    }
    throw new FlinkHiveUDFException(String.format("Flink doesn't support convert object conversion for %s yet", inspector));
}

Example 3 with VoidObjectInspector

use of org.apache.hadoop.hive.serde2.objectinspector.primitive.VoidObjectInspector in project flink by apache.

the class HiveInspectors method toFlinkObject.

/**
 * Converts a Hive object to Flink object with an ObjectInspector.
 */
public static Object toFlinkObject(ObjectInspector inspector, Object data, HiveShim hiveShim) {
    if (data == null || inspector instanceof VoidObjectInspector) {
        return null;
    }
    if (inspector instanceof PrimitiveObjectInspector) {
        if (inspector instanceof BooleanObjectInspector || inspector instanceof StringObjectInspector || inspector instanceof ByteObjectInspector || inspector instanceof ShortObjectInspector || inspector instanceof IntObjectInspector || inspector instanceof LongObjectInspector || inspector instanceof FloatObjectInspector || inspector instanceof DoubleObjectInspector || inspector instanceof BinaryObjectInspector) {
            PrimitiveObjectInspector poi = (PrimitiveObjectInspector) inspector;
            return poi.getPrimitiveJavaObject(data);
        } else if (inspector instanceof DateObjectInspector) {
            PrimitiveObjectInspector poi = (PrimitiveObjectInspector) inspector;
            return hiveShim.toFlinkDate(poi.getPrimitiveJavaObject(data));
        } else if (inspector instanceof TimestampObjectInspector) {
            PrimitiveObjectInspector poi = (PrimitiveObjectInspector) inspector;
            return hiveShim.toFlinkTimestamp(poi.getPrimitiveJavaObject(data));
        } else if (inspector instanceof HiveCharObjectInspector) {
            HiveCharObjectInspector oi = (HiveCharObjectInspector) inspector;
            return oi.getPrimitiveJavaObject(data).getValue();
        } else if (inspector instanceof HiveVarcharObjectInspector) {
            HiveVarcharObjectInspector oi = (HiveVarcharObjectInspector) inspector;
            return oi.getPrimitiveJavaObject(data).getValue();
        } else if (inspector instanceof HiveDecimalObjectInspector) {
            HiveDecimalObjectInspector oi = (HiveDecimalObjectInspector) inspector;
            return oi.getPrimitiveJavaObject(data).bigDecimalValue();
        }
    }
    if (inspector instanceof ListObjectInspector) {
        ListObjectInspector listInspector = (ListObjectInspector) inspector;
        List<?> list = listInspector.getList(data);
        if (list == null) {
            return null;
        }
        // flink expects a specific array type (e.g. Integer[] instead of Object[]), so we have
        // to get the element class
        ObjectInspector elementInspector = listInspector.getListElementObjectInspector();
        Object[] result = (Object[]) Array.newInstance(HiveTypeUtil.toFlinkType(elementInspector).getConversionClass(), list.size());
        for (int i = 0; i < list.size(); i++) {
            result[i] = toFlinkObject(elementInspector, list.get(i), hiveShim);
        }
        return result;
    }
    if (inspector instanceof MapObjectInspector) {
        MapObjectInspector mapInspector = (MapObjectInspector) inspector;
        Map<?, ?> map = mapInspector.getMap(data);
        if (map == null) {
            return null;
        }
        Map<Object, Object> result = new HashMap<>(map.size());
        for (Map.Entry<?, ?> entry : map.entrySet()) {
            result.put(toFlinkObject(mapInspector.getMapKeyObjectInspector(), entry.getKey(), hiveShim), toFlinkObject(mapInspector.getMapValueObjectInspector(), entry.getValue(), hiveShim));
        }
        return result;
    }
    if (inspector instanceof StructObjectInspector) {
        StructObjectInspector structInspector = (StructObjectInspector) inspector;
        List<? extends StructField> fields = structInspector.getAllStructFieldRefs();
        Row row = new Row(fields.size());
        // list as data
        if (!data.getClass().isArray() && !(data instanceof List) && (inspector instanceof StandardStructObjectInspector)) {
            data = new Object[] { data };
        }
        for (int i = 0; i < row.getArity(); i++) {
            row.setField(i, toFlinkObject(fields.get(i).getFieldObjectInspector(), structInspector.getStructFieldData(data, fields.get(i)), hiveShim));
        }
        return row;
    }
    throw new FlinkHiveUDFException(String.format("Unwrap does not support ObjectInspector '%s' yet", inspector));
}

Example 4 with VoidObjectInspector

use of org.apache.hadoop.hive.serde2.objectinspector.primitive.VoidObjectInspector in project haivvreo by jghoman.

the class TestAvroDeserializer method canDeserializeVoidType.

@Test
public void canDeserializeVoidType() throws IOException, SerDeException {
    String schemaString = "{\n" + "  \"type\": \"record\", \n" + "  \"name\": \"nullTest\",\n" + "  \"fields\" : [\n" + "    {\"name\": \"isANull\", \"type\": \"null\"}\n" + "  ]\n" + "}";
    Schema s = Schema.parse(schemaString);
    GenericData.Record record = new GenericData.Record(s);
    record.put("isANull", null);
    assertTrue(GENERIC_DATA.validate(s, record));
    AvroGenericRecordWritable garw = Utils.serializeAndDeserializeRecord(record);
    AvroObjectInspectorGenerator aoig = new AvroObjectInspectorGenerator(s);
    AvroDeserializer de = new AvroDeserializer();
    ArrayList<Object> row = (ArrayList<Object>) de.deserialize(aoig.getColumnNames(), aoig.getColumnTypes(), garw, s);
    assertEquals(1, row.size());
    Object theVoidObject = row.get(0);
    assertNull(theVoidObject);
    StandardStructObjectInspector oi = (StandardStructObjectInspector) aoig.getObjectInspector();
    StructField fieldRef = oi.getStructFieldRef("isANull");
    Object shouldBeNull = oi.getStructFieldData(row, fieldRef);
    assertNull(shouldBeNull);
    assertTrue(fieldRef.getFieldObjectInspector() instanceof VoidObjectInspector);
}

Example 5 with VoidObjectInspector

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

the class TestAvroDeserializer method canDeserializeVoidType.

@Test
public void canDeserializeVoidType() throws IOException, SerDeException {
    String schemaString = "{\n" + "  \"type\": \"record\", \n" + "  \"name\": \"nullTest\",\n" + "  \"fields\" : [\n" + "    {\"name\": \"isANull\", \"type\": \"null\"}\n" + "  ]\n" + "}";
    Schema s = AvroSerdeUtils.getSchemaFor(schemaString);
    GenericData.Record record = new GenericData.Record(s);
    record.put("isANull", null);
    assertTrue(GENERIC_DATA.validate(s, record));
    AvroGenericRecordWritable garw = Utils.serializeAndDeserializeRecord(record);
    AvroObjectInspectorGenerator aoig = new AvroObjectInspectorGenerator(s);
    AvroDeserializer de = new AvroDeserializer();
    ArrayList<Object> row = (ArrayList<Object>) de.deserialize(aoig.getColumnNames(), aoig.getColumnTypes(), garw, s);
    assertEquals(1, row.size());
    Object theVoidObject = row.get(0);
    assertNull(theVoidObject);
    StandardStructObjectInspector oi = (StandardStructObjectInspector) aoig.getObjectInspector();
    StructField fieldRef = oi.getStructFieldRef("isANull");
    Object shouldBeNull = oi.getStructFieldData(row, fieldRef);
    assertNull(shouldBeNull);
    assertTrue(fieldRef.getFieldObjectInspector() instanceof VoidObjectInspector);
}