use of org.apache.carbondata.core.metadata.datatype.DataType in project carbondata by apache.
the class CarbondataSplitManager method parseFilterExpression.
/**
* Convert presto-TupleDomain predication into Carbon scan express condition
* @param originalConstraint presto-TupleDomain
* @param carbonTable
* @return
*/
public Expression parseFilterExpression(TupleDomain<ColumnHandle> originalConstraint, CarbonTable carbonTable) {
ImmutableList.Builder<Expression> filters = ImmutableList.builder();
Domain domain = null;
for (ColumnHandle c : originalConstraint.getDomains().get().keySet()) {
CarbondataColumnHandle cdch = (CarbondataColumnHandle) c;
Type type = cdch.getColumnType();
List<CarbonColumn> ccols = carbonTable.getCreateOrderColumn(carbonTable.getFactTableName());
Optional<CarbonColumn> target = ccols.stream().filter(a -> a.getColName().equals(cdch.getColumnName())).findFirst();
if (target.get() == null)
return null;
DataType coltype = target.get().getDataType();
ColumnExpression colExpression = new ColumnExpression(cdch.getColumnName(), target.get().getDataType());
//colExpression.setColIndex(cs.getSchemaOrdinal());
colExpression.setDimension(target.get().isDimension());
colExpression.setDimension(carbonTable.getDimensionByName(carbonTable.getFactTableName(), cdch.getColumnName()));
colExpression.setCarbonColumn(target.get());
domain = originalConstraint.getDomains().get().get(c);
checkArgument(domain.getType().isOrderable(), "Domain type must be orderable");
if (domain.getValues().isNone()) {
}
if (domain.getValues().isAll()) {
}
List<Object> singleValues = new ArrayList<>();
List<Expression> rangeFilter = new ArrayList<>();
for (Range range : domain.getValues().getRanges().getOrderedRanges()) {
// Already checked
checkState(!range.isAll());
if (range.isSingleValue()) {
singleValues.add(range.getLow().getValue());
} else {
List<String> rangeConjuncts = new ArrayList<>();
if (!range.getLow().isLowerUnbounded()) {
Object value = ConvertDataByType(range.getLow().getValue(), type);
switch(range.getLow().getBound()) {
case ABOVE:
if (type == TimestampType.TIMESTAMP) {
//todo not now
} else {
GreaterThanExpression greater = new GreaterThanExpression(colExpression, new LiteralExpression(value, coltype));
rangeFilter.add(greater);
}
break;
case EXACTLY:
GreaterThanEqualToExpression greater = new GreaterThanEqualToExpression(colExpression, new LiteralExpression(value, coltype));
rangeFilter.add(greater);
break;
case BELOW:
throw new IllegalArgumentException("Low marker should never use BELOW bound");
default:
throw new AssertionError("Unhandled bound: " + range.getLow().getBound());
}
}
if (!range.getHigh().isUpperUnbounded()) {
Object value = ConvertDataByType(range.getHigh().getValue(), type);
switch(range.getHigh().getBound()) {
case ABOVE:
throw new IllegalArgumentException("High marker should never use ABOVE bound");
case EXACTLY:
LessThanEqualToExpression less = new LessThanEqualToExpression(colExpression, new LiteralExpression(value, coltype));
rangeFilter.add(less);
break;
case BELOW:
LessThanExpression less2 = new LessThanExpression(colExpression, new LiteralExpression(value, coltype));
rangeFilter.add(less2);
break;
default:
throw new AssertionError("Unhandled bound: " + range.getHigh().getBound());
}
}
}
}
if (singleValues.size() == 1) {
Expression ex = null;
if (coltype.equals(DataType.STRING)) {
ex = new EqualToExpression(colExpression, new LiteralExpression(((Slice) singleValues.get(0)).toStringUtf8(), coltype));
} else
ex = new EqualToExpression(colExpression, new LiteralExpression(singleValues.get(0), coltype));
filters.add(ex);
} else if (singleValues.size() > 1) {
ListExpression candidates = null;
List<Expression> exs = singleValues.stream().map((a) -> {
return new LiteralExpression(ConvertDataByType(a, type), coltype);
}).collect(Collectors.toList());
candidates = new ListExpression(exs);
if (candidates != null)
filters.add(new InExpression(colExpression, candidates));
} else if (rangeFilter.size() > 0) {
if (rangeFilter.size() > 1) {
Expression finalFilters = new OrExpression(rangeFilter.get(0), rangeFilter.get(1));
if (rangeFilter.size() > 2) {
for (int i = 2; i < rangeFilter.size(); i++) {
filters.add(new AndExpression(finalFilters, rangeFilter.get(i)));
}
}
} else if (//only have one value
rangeFilter.size() == 1)
filters.add(rangeFilter.get(0));
}
}
Expression finalFilters;
List<Expression> tmp = filters.build();
if (tmp.size() > 1) {
finalFilters = new AndExpression(tmp.get(0), tmp.get(1));
if (tmp.size() > 2) {
for (int i = 2; i < tmp.size(); i++) {
finalFilters = new AndExpression(finalFilters, tmp.get(i));
}
}
} else if (tmp.size() == 1)
finalFilters = tmp.get(0);
else
//no filter
return null;
return finalFilters;
}
use of org.apache.carbondata.core.metadata.datatype.DataType in project carbondata by apache.
the class ValueCompressionUtil method getDoubleCompressorFinder.
private static CompressionFinder getDoubleCompressorFinder(Object maxValue, Object minValue, int mantissa, byte dataTypeSelected, DataType measureStoreType) {
//Here we should use the Max abs as max to getDatatype, let's say -1 and -10000000, -1 is max,
//but we can't use -1 to getDatatype, we should use -10000000.
double absMaxValue = Math.abs((double) maxValue) >= Math.abs((double) minValue) ? (double) maxValue : (double) minValue;
DataType adaptiveDataType = getDataType(absMaxValue, mantissa, dataTypeSelected);
DataType deltaDataType = getDataType((double) maxValue - (double) minValue, mantissa, dataTypeSelected);
if (mantissa == 0) {
// short, int, long
int adaptiveSize = getSize(adaptiveDataType);
int deltaSize = getSize(deltaDataType);
if (adaptiveSize > deltaSize) {
return new CompressionFinder(COMPRESSION_TYPE.DELTA_DOUBLE, DataType.DOUBLE, deltaDataType, measureStoreType);
} else if (adaptiveSize < deltaSize) {
return new CompressionFinder(COMPRESSION_TYPE.ADAPTIVE, DataType.DOUBLE, deltaDataType, measureStoreType);
} else {
return new CompressionFinder(COMPRESSION_TYPE.ADAPTIVE, DataType.DOUBLE, adaptiveDataType, measureStoreType);
}
} else {
// double
DataType maxNonDecDataType = getDataType(Math.pow(10, mantissa) * absMaxValue, 0, dataTypeSelected);
DataType diffNonDecDataType = getDataType(Math.pow(10, mantissa) * ((double) maxValue - (double) minValue), 0, dataTypeSelected);
CompressionFinder[] finders = { new CompressionFinder(COMPRESSION_TYPE.ADAPTIVE, adaptiveDataType, adaptiveDataType, CompressionFinder.PRIORITY.ACTUAL, measureStoreType), new CompressionFinder(COMPRESSION_TYPE.DELTA_DOUBLE, adaptiveDataType, deltaDataType, CompressionFinder.PRIORITY.DIFFSIZE, measureStoreType), new CompressionFinder(COMPRESSION_TYPE.BIGINT, adaptiveDataType, maxNonDecDataType, CompressionFinder.PRIORITY.MAXNONDECIMAL, measureStoreType), new CompressionFinder(COMPRESSION_TYPE.DELTA_NON_DECIMAL, adaptiveDataType, diffNonDecDataType, CompressionFinder.PRIORITY.DIFFNONDECIMAL, measureStoreType) };
// sort the compressionFinder.The top have the highest priority
Arrays.sort(finders);
return finders[0];
}
}
use of org.apache.carbondata.core.metadata.datatype.DataType in project carbondata by apache.
the class RangeExpressionEvaluator method addFilterExpressionMap.
private void addFilterExpressionMap(Map<String, List<FilterModificationNode>> filterExpressionMap, Expression currentNode, Expression parentNode) {
String colName = getColumnName(currentNode);
DataType dataType = getLiteralDataType(currentNode);
Object literalVal = getLiteralValue(currentNode);
ExpressionType expType = getExpressionType(currentNode);
FilterModificationNode filterExpression = new FilterModificationNode(currentNode, parentNode, expType, dataType, literalVal, colName);
if (null == filterExpressionMap.get(colName)) {
filterExpressionMap.put(colName, new ArrayList<FilterModificationNode>());
}
filterExpressionMap.get(colName).add(filterExpression);
}
use of org.apache.carbondata.core.metadata.datatype.DataType in project carbondata by apache.
the class ValueCompressionUtilTest method testToGetValueCompressionModelForByteAndIntAndDataTypeSelected0.
@Test
public void testToGetValueCompressionModelForByteAndIntAndDataTypeSelected0() {
Object[] maxValues = { -32766.00 };
Object[] minValues = { 32744.0 };
int[] decimalLength = { 1 };
Object[] uniqueValues = { 5 };
DataType[] types = { DataType.DOUBLE };
byte[] dataTypeSelected = { 0 };
MeasureMetaDataModel measureMetaDataModel = new MeasureMetaDataModel(maxValues, minValues, decimalLength, 1, uniqueValues, types, dataTypeSelected);
WriterCompressModel writerCompressModel = ValueCompressionUtil.getWriterCompressModel(measureMetaDataModel);
assertEquals(ValueCompressionUtil.COMPRESSION_TYPE.BIGINT, writerCompressModel.getCompType(0));
}
use of org.apache.carbondata.core.metadata.datatype.DataType in project carbondata by apache.
the class ValueCompressionUtilTest method testToGetValueCompressionModelForDefaultAggregatorType.
@Test
public void testToGetValueCompressionModelForDefaultAggregatorType() {
Object[] maxValues = { 10.0 };
Object[] minValues = { 1.0 };
int[] decimalLength = { 0 };
Object[] uniqueValues = { 5 };
DataType[] types = { DataType.DOUBLE };
byte[] dataTypeSelected = { 1 };
MeasureMetaDataModel measureMetaDataModel = new MeasureMetaDataModel(maxValues, minValues, decimalLength, 1, uniqueValues, types, dataTypeSelected);
WriterCompressModel writerCompressModel = ValueCompressionUtil.getWriterCompressModel(measureMetaDataModel);
assertEquals(ValueCompressionUtil.COMPRESSION_TYPE.ADAPTIVE, writerCompressModel.getCompType(0));
}
Aggregations