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

Example 26 with PrimitiveCategory

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

the class RexNodeConverter method convert.

protected RexNode convert(ExprNodeConstantDesc literal) throws CalciteSemanticException {
    RexBuilder rexBuilder = cluster.getRexBuilder();
    RelDataTypeFactory dtFactory = rexBuilder.getTypeFactory();
    PrimitiveTypeInfo hiveType = (PrimitiveTypeInfo) literal.getTypeInfo();
    RelDataType calciteDataType = TypeConverter.convert(hiveType, dtFactory);
    PrimitiveCategory hiveTypeCategory = hiveType.getPrimitiveCategory();
    ConstantObjectInspector coi = literal.getWritableObjectInspector();
    Object value = ObjectInspectorUtils.copyToStandardJavaObject(coi.getWritableConstantValue(), coi);
    RexNode calciteLiteral = null;
    // If value is null, the type should also be VOID.
    if (value == null) {
        hiveTypeCategory = PrimitiveCategory.VOID;
    }
    // TODO: Verify if we need to use ConstantObjectInspector to unwrap data
    switch(hiveTypeCategory) {
        case BOOLEAN:
            calciteLiteral = rexBuilder.makeLiteral(((Boolean) value).booleanValue());
            break;
        case BYTE:
            calciteLiteral = rexBuilder.makeExactLiteral(new BigDecimal((Byte) value), calciteDataType);
            break;
        case SHORT:
            calciteLiteral = rexBuilder.makeExactLiteral(new BigDecimal((Short) value), calciteDataType);
            break;
        case INT:
            calciteLiteral = rexBuilder.makeExactLiteral(new BigDecimal((Integer) value));
            break;
        case LONG:
            calciteLiteral = rexBuilder.makeBigintLiteral(new BigDecimal((Long) value));
            break;
        // TODO: is Decimal an exact numeric or approximate numeric?
        case DECIMAL:
            if (value instanceof HiveDecimal) {
                value = ((HiveDecimal) value).bigDecimalValue();
            } else if (value instanceof Decimal128) {
                value = ((Decimal128) value).toBigDecimal();
            }
            if (value == null) {
                // literals.
                throw new CalciteSemanticException("Expression " + literal.getExprString() + " is not a valid decimal", UnsupportedFeature.Invalid_decimal);
            // TODO: return createNullLiteral(literal);
            }
            BigDecimal bd = (BigDecimal) value;
            BigInteger unscaled = bd.unscaledValue();
            if (unscaled.compareTo(MIN_LONG_BI) >= 0 && unscaled.compareTo(MAX_LONG_BI) <= 0) {
                calciteLiteral = rexBuilder.makeExactLiteral(bd);
            } else {
                // CBO doesn't support unlimited precision decimals. In practice, this
                // will work...
                // An alternative would be to throw CboSemanticException and fall back
                // to no CBO.
                RelDataType relType = cluster.getTypeFactory().createSqlType(SqlTypeName.DECIMAL, unscaled.toString().length(), bd.scale());
                calciteLiteral = rexBuilder.makeExactLiteral(bd, relType);
            }
            break;
        case FLOAT:
            calciteLiteral = rexBuilder.makeApproxLiteral(new BigDecimal(Float.toString((Float) value)), calciteDataType);
            break;
        case DOUBLE:
            // TODO: The best solution is to support NaN in expression reduction.
            if (Double.isNaN((Double) value)) {
                throw new CalciteSemanticException("NaN", UnsupportedFeature.Invalid_decimal);
            }
            calciteLiteral = rexBuilder.makeApproxLiteral(new BigDecimal(Double.toString((Double) value)), calciteDataType);
            break;
        case CHAR:
            if (value instanceof HiveChar) {
                value = ((HiveChar) value).getValue();
            }
            calciteLiteral = rexBuilder.makeCharLiteral(asUnicodeString((String) value));
            break;
        case VARCHAR:
            if (value instanceof HiveVarchar) {
                value = ((HiveVarchar) value).getValue();
            }
            calciteLiteral = rexBuilder.makeCharLiteral(asUnicodeString((String) value));
            break;
        case STRING:
            calciteLiteral = rexBuilder.makeCharLiteral(asUnicodeString((String) value));
            break;
        case DATE:
            Calendar cal = new GregorianCalendar();
            cal.setTime((Date) value);
            calciteLiteral = rexBuilder.makeDateLiteral(cal);
            break;
        case TIMESTAMP:
            Calendar c = null;
            if (value instanceof Calendar) {
                c = (Calendar) value;
            } else {
                c = Calendar.getInstance();
                c.setTimeInMillis(((Timestamp) value).getTime());
            }
            calciteLiteral = rexBuilder.makeTimestampLiteral(c, RelDataType.PRECISION_NOT_SPECIFIED);
            break;
        case INTERVAL_YEAR_MONTH:
            // Calcite year-month literal value is months as BigDecimal
            BigDecimal totalMonths = BigDecimal.valueOf(((HiveIntervalYearMonth) value).getTotalMonths());
            calciteLiteral = rexBuilder.makeIntervalLiteral(totalMonths, new SqlIntervalQualifier(TimeUnit.YEAR, TimeUnit.MONTH, new SqlParserPos(1, 1)));
            break;
        case INTERVAL_DAY_TIME:
            // Calcite day-time interval is millis value as BigDecimal
            // Seconds converted to millis
            BigDecimal secsValueBd = BigDecimal.valueOf(((HiveIntervalDayTime) value).getTotalSeconds() * 1000);
            // Nanos converted to millis
            BigDecimal nanosValueBd = BigDecimal.valueOf(((HiveIntervalDayTime) value).getNanos(), 6);
            calciteLiteral = rexBuilder.makeIntervalLiteral(secsValueBd.add(nanosValueBd), new SqlIntervalQualifier(TimeUnit.MILLISECOND, null, new SqlParserPos(1, 1)));
            break;
        case VOID:
            calciteLiteral = cluster.getRexBuilder().makeLiteral(null, cluster.getTypeFactory().createSqlType(SqlTypeName.NULL), true);
            break;
        case BINARY:
        case UNKNOWN:
        default:
            throw new RuntimeException("UnSupported Literal");
    }
    return calciteLiteral;
}

Example 27 with PrimitiveCategory

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

the class Vectorizer method specializeReduceSinkOperator.

private Operator<? extends OperatorDesc> specializeReduceSinkOperator(Operator<? extends OperatorDesc> op, VectorizationContext vContext, ReduceSinkDesc desc, VectorReduceSinkInfo vectorReduceSinkInfo) throws HiveException {
    Operator<? extends OperatorDesc> vectorOp = null;
    Class<? extends Operator<?>> opClass = null;
    Type[] reduceSinkKeyColumnVectorTypes = vectorReduceSinkInfo.getReduceSinkKeyColumnVectorTypes();
    // By default, we can always use the multi-key class.
    VectorReduceSinkDesc.ReduceSinkKeyType reduceSinkKeyType = VectorReduceSinkDesc.ReduceSinkKeyType.MULTI_KEY;
    // Look for single column optimization.
    if (reduceSinkKeyColumnVectorTypes.length == 1) {
        LOG.info("Vectorizer vectorizeOperator groupby typeName " + vectorReduceSinkInfo.getReduceSinkKeyTypeInfos()[0]);
        Type columnVectorType = reduceSinkKeyColumnVectorTypes[0];
        switch(columnVectorType) {
            case LONG:
                {
                    PrimitiveCategory primitiveCategory = ((PrimitiveTypeInfo) vectorReduceSinkInfo.getReduceSinkKeyTypeInfos()[0]).getPrimitiveCategory();
                    switch(primitiveCategory) {
                        case BOOLEAN:
                        case BYTE:
                        case SHORT:
                        case INT:
                        case LONG:
                            reduceSinkKeyType = VectorReduceSinkDesc.ReduceSinkKeyType.LONG;
                            break;
                        default:
                            // Other integer types not supported yet.
                            break;
                    }
                }
                break;
            case BYTES:
                reduceSinkKeyType = VectorReduceSinkDesc.ReduceSinkKeyType.STRING;
            default:
                // Stay with multi-key.
                break;
        }
    }
    switch(reduceSinkKeyType) {
        case LONG:
            opClass = VectorReduceSinkLongOperator.class;
            break;
        case STRING:
            opClass = VectorReduceSinkStringOperator.class;
            break;
        case MULTI_KEY:
            opClass = VectorReduceSinkMultiKeyOperator.class;
            break;
        default:
            throw new HiveException("Unknown reduce sink key type " + reduceSinkKeyType);
    }
    VectorReduceSinkDesc vectorDesc = (VectorReduceSinkDesc) desc.getVectorDesc();
    vectorDesc.setReduceSinkKeyType(reduceSinkKeyType);
    vectorDesc.setVectorReduceSinkInfo(vectorReduceSinkInfo);
    vectorOp = OperatorFactory.getVectorOperator(opClass, op.getCompilationOpContext(), op.getConf(), vContext);
    LOG.info("Vectorizer vectorizeOperator reduce sink class " + vectorOp.getClass().getSimpleName());
    return vectorOp;
}

Example 28 with PrimitiveCategory

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

the class PTFTranslator method validateValueBoundaryExprType.

private static void validateValueBoundaryExprType(ObjectInspector OI) throws SemanticException {
    if (!OI.getCategory().equals(Category.PRIMITIVE)) {
        throw new SemanticException(String.format("Value Boundary expression must be of primitive type. Found: %s", OI.getTypeName()));
    }
    PrimitiveObjectInspector pOI = (PrimitiveObjectInspector) OI;
    PrimitiveCategory pC = pOI.getPrimitiveCategory();
    switch(pC) {
        case BOOLEAN:
        case BYTE:
        case DOUBLE:
        case FLOAT:
        case INT:
        case LONG:
        case SHORT:
        case DECIMAL:
        case TIMESTAMP:
        case DATE:
        case STRING:
        case VARCHAR:
        case CHAR:
            break;
        default:
            throw new SemanticException(String.format("Primitive type %s not supported in Value Boundary expression", pC));
    }
}

Example 29 with PrimitiveCategory

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

the class GenericUDFDateDiff method convertToDate.

private Date convertToDate(PrimitiveCategory inputType, Converter converter, DeferredObject argument) throws HiveException {
    assert (converter != null);
    assert (argument != null);
    if (argument.get() == null) {
        return null;
    }
    Date date = new Date(0);
    switch(inputType) {
        case STRING:
        case VARCHAR:
        case CHAR:
            String dateString = converter.convert(argument.get()).toString();
            try {
                date.setTime(formatter.parse(dateString).getTime());
            } catch (ParseException e) {
                return null;
            }
            break;
        case TIMESTAMP:
            Timestamp ts = ((TimestampWritable) converter.convert(argument.get())).getTimestamp();
            date.setTime(ts.getTime());
            break;
        case DATE:
            DateWritable dw = (DateWritable) converter.convert(argument.get());
            date = dw.get();
            break;
        default:
            throw new UDFArgumentException("TO_DATE() only takes STRING/TIMESTAMP/DATEWRITABLE types, got " + inputType);
    }
    return date;
}

Example 30 with PrimitiveCategory

use of org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector.PrimitiveCategory 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;
}