use of io.trino.spi.type.TimestampType.TIMESTAMP_NANOS in project trino by trinodb.
the class TupleDomainOrcPredicate method getDomain.
@VisibleForTesting
public static Domain getDomain(Type type, long rowCount, ColumnStatistics columnStatistics) {
if (rowCount == 0) {
return Domain.none(type);
}
if (columnStatistics == null) {
return Domain.all(type);
}
if (columnStatistics.hasNumberOfValues() && columnStatistics.getNumberOfValues() == 0) {
return Domain.onlyNull(type);
}
boolean hasNullValue = columnStatistics.getNumberOfValues() != rowCount;
if (type instanceof TimeType && columnStatistics.getIntegerStatistics() != null) {
// This is the representation of TIME used by Iceberg
return createDomain(type, hasNullValue, columnStatistics.getIntegerStatistics(), value -> ((long) value) * Timestamps.PICOSECONDS_PER_MICROSECOND);
}
if (type.getJavaType() == boolean.class && columnStatistics.getBooleanStatistics() != null) {
BooleanStatistics booleanStatistics = columnStatistics.getBooleanStatistics();
boolean hasTrueValues = (booleanStatistics.getTrueValueCount() != 0);
boolean hasFalseValues = (columnStatistics.getNumberOfValues() != booleanStatistics.getTrueValueCount());
if (hasTrueValues && hasFalseValues) {
return Domain.all(BOOLEAN);
}
if (hasTrueValues) {
return Domain.create(ValueSet.of(BOOLEAN, true), hasNullValue);
}
if (hasFalseValues) {
return Domain.create(ValueSet.of(BOOLEAN, false), hasNullValue);
}
} else if (isShortDecimal(type) && columnStatistics.getDecimalStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getDecimalStatistics(), value -> rescale(value, (DecimalType) type).unscaledValue().longValue());
} else if (isLongDecimal(type) && columnStatistics.getDecimalStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getDecimalStatistics(), value -> Int128.valueOf(rescale(value, (DecimalType) type).unscaledValue()));
} else if (type instanceof CharType && columnStatistics.getStringStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getStringStatistics(), value -> truncateToLengthAndTrimSpaces(value, type));
} else if (type instanceof VarcharType && columnStatistics.getStringStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getStringStatistics());
} else if (type instanceof DateType && columnStatistics.getDateStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getDateStatistics(), value -> (long) value);
} else if ((type.equals(TIMESTAMP_MILLIS) || type.equals(TIMESTAMP_MICROS)) && columnStatistics.getTimestampStatistics() != null) {
// upper bound of the domain we create must be adjusted accordingly, to includes the rounded timestamp.
return createDomain(type, hasNullValue, columnStatistics.getTimestampStatistics(), min -> min * MICROSECONDS_PER_MILLISECOND, max -> (max + 1) * MICROSECONDS_PER_MILLISECOND);
} else if (type.equals(TIMESTAMP_NANOS) && columnStatistics.getTimestampStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getTimestampStatistics(), min -> new LongTimestamp(min * MICROSECONDS_PER_MILLISECOND, 0), max -> new LongTimestamp((max + 1) * MICROSECONDS_PER_MILLISECOND, 0));
} else if (type.equals(TIMESTAMP_TZ_MILLIS) && columnStatistics.getTimestampStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getTimestampStatistics(), value -> packDateTimeWithZone(value, UTC_KEY));
} else if (type.equals(TIMESTAMP_TZ_MICROS) && (columnStatistics.getTimestampStatistics() != null)) {
return createDomain(type, hasNullValue, columnStatistics.getTimestampStatistics(), min -> LongTimestampWithTimeZone.fromEpochMillisAndFraction(min, 0, UTC_KEY), max -> LongTimestampWithTimeZone.fromEpochMillisAndFraction(max, 999_000_000, UTC_KEY));
} else if (type.equals(TIMESTAMP_TZ_NANOS) && columnStatistics.getTimestampStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getTimestampStatistics(), min -> LongTimestampWithTimeZone.fromEpochMillisAndFraction(min, 0, UTC_KEY), max -> LongTimestampWithTimeZone.fromEpochMillisAndFraction(max, 999_999_000, UTC_KEY));
} else if (type.getJavaType() == long.class && columnStatistics.getIntegerStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getIntegerStatistics());
} else if (type.getJavaType() == double.class && columnStatistics.getDoubleStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getDoubleStatistics());
} else if (REAL.equals(type) && columnStatistics.getDoubleStatistics() != null) {
return createDomain(type, hasNullValue, columnStatistics.getDoubleStatistics(), value -> (long) floatToRawIntBits(value.floatValue()));
}
return Domain.create(ValueSet.all(type), hasNullValue);
}
use of io.trino.spi.type.TimestampType.TIMESTAMP_NANOS in project trino by trinodb.
the class OrcTester method preprocessWriteValueHive.
private static Object preprocessWriteValueHive(Type type, Object value) {
if (value == null) {
return null;
}
if (type.equals(BOOLEAN)) {
return value;
}
if (type.equals(TINYINT)) {
return ((Number) value).byteValue();
}
if (type.equals(SMALLINT)) {
return ((Number) value).shortValue();
}
if (type.equals(INTEGER)) {
return ((Number) value).intValue();
}
if (type.equals(BIGINT)) {
return ((Number) value).longValue();
}
if (type.equals(REAL)) {
return ((Number) value).floatValue();
}
if (type.equals(DOUBLE)) {
return ((Number) value).doubleValue();
}
if (type instanceof VarcharType) {
return value;
}
if (type instanceof CharType) {
return new HiveChar((String) value, ((CharType) type).getLength());
}
if (type.equals(VARBINARY)) {
return ((SqlVarbinary) value).getBytes();
}
if (type.equals(DATE)) {
return Date.ofEpochDay(((SqlDate) value).getDays());
}
if (type.equals(TIMESTAMP_MILLIS) || type.equals(TIMESTAMP_MICROS) || type.equals(TIMESTAMP_NANOS)) {
LocalDateTime dateTime = ((SqlTimestamp) value).toLocalDateTime();
return Timestamp.ofEpochSecond(dateTime.toEpochSecond(ZoneOffset.UTC), dateTime.getNano());
}
if (type.equals(TIMESTAMP_TZ_MILLIS) || type.equals(TIMESTAMP_TZ_MICROS) || type.equals(TIMESTAMP_TZ_NANOS)) {
SqlTimestampWithTimeZone timestamp = (SqlTimestampWithTimeZone) value;
int nanosOfMilli = roundDiv(timestamp.getPicosOfMilli(), PICOSECONDS_PER_NANOSECOND);
return Timestamp.ofEpochMilli(timestamp.getEpochMillis(), nanosOfMilli);
}
if (type instanceof DecimalType) {
return HiveDecimal.create(((SqlDecimal) value).toBigDecimal());
}
if (type instanceof ArrayType) {
Type elementType = type.getTypeParameters().get(0);
return ((List<?>) value).stream().map(element -> preprocessWriteValueHive(elementType, element)).collect(toList());
}
if (type instanceof MapType) {
Type keyType = type.getTypeParameters().get(0);
Type valueType = type.getTypeParameters().get(1);
Map<Object, Object> newMap = new HashMap<>();
for (Entry<?, ?> entry : ((Map<?, ?>) value).entrySet()) {
newMap.put(preprocessWriteValueHive(keyType, entry.getKey()), preprocessWriteValueHive(valueType, entry.getValue()));
}
return newMap;
}
if (type instanceof RowType) {
List<?> fieldValues = (List<?>) value;
List<Type> fieldTypes = type.getTypeParameters();
List<Object> newStruct = new ArrayList<>();
for (int fieldId = 0; fieldId < fieldValues.size(); fieldId++) {
newStruct.add(preprocessWriteValueHive(fieldTypes.get(fieldId), fieldValues.get(fieldId)));
}
return newStruct;
}
throw new IllegalArgumentException("unsupported type: " + type);
}
Aggregations