private void parseStringColumn(int column) {
// don't include column number in message because it causes boxing which is expensive here
checkArgument(!isPartitionColumn[column], "Column is a partition key");
loaded[column] = true;
Object fieldData = rowInspector.getStructFieldData(rowData, structFields[column]);
if (fieldData == null) {
nulls[column] = true;
} else if (hiveTypes[column] == HiveType.MAP
|| hiveTypes[column] == HiveType.LIST
|| hiveTypes[column] == HiveType.STRUCT) {
// temporarily special case MAP, LIST, and STRUCT types as strings
slices[column] =
Slices.wrappedBuffer(
SerDeUtils.getJsonBytes(sessionTimeZone, fieldData, fieldInspectors[column]));
nulls[column] = false;
} else {
Object fieldValue =
((PrimitiveObjectInspector) fieldInspectors[column]).getPrimitiveJavaObject(fieldData);
checkState(fieldValue != null, "fieldValue should not be null");
if (fieldValue instanceof String) {
slices[column] = Slices.utf8Slice((String) fieldValue);
} else if (fieldValue instanceof byte[]) {
slices[column] = Slices.wrappedBuffer((byte[]) fieldValue);
} else {
throw new IllegalStateException(
"unsupported string field type: " + fieldValue.getClass().getName());
}
nulls[column] = false;
}
}
private void parseStringColumn(int column) {
// don't include column number in message because it causes boxing which is expensive here
checkArgument(!isPartitionColumn[column], "Column is a partition key");
loaded[column] = true;
Object fieldData = rowInspector.getStructFieldData(rowData, structFields[column]);
if (fieldData == null) {
nulls[column] = true;
} else {
Object fieldValue =
((PrimitiveObjectInspector) fieldInspectors[column]).getPrimitiveJavaObject(fieldData);
checkState(fieldValue != null, "fieldValue should not be null");
Slice value;
if (fieldValue instanceof String) {
value = Slices.utf8Slice((String) fieldValue);
} else if (fieldValue instanceof byte[]) {
value = Slices.wrappedBuffer((byte[]) fieldValue);
} else if (fieldValue instanceof HiveVarchar) {
value = Slices.utf8Slice(((HiveVarchar) fieldValue).getValue());
} else {
throw new IllegalStateException(
"unsupported string field type: " + fieldValue.getClass().getName());
}
Type type = types[column];
if (isVarcharType(type)) {
value = truncateToLength(value, type);
}
slices[column] = value;
nulls[column] = false;
}
}
private void newBuffer() {
this.buffer = allocator.allocate(PageFormat.PAGE_HEADER_SIZE + fixedRecordSize);
this.bufferSlice = Slices.wrappedBuffer(buffer.array(), buffer.offset(), buffer.capacity());
this.count = 0;
this.position = PageFormat.PAGE_HEADER_SIZE;
this.stringReferences.clear();
this.stringReferenceSize = 0;
}
private static Slice formatDatetime(
ISOChronology chronology, Locale locale, long timestamp, Slice formatString) {
String pattern = formatString.toString(Charsets.UTF_8);
DateTimeFormatter formatter =
DateTimeFormat.forPattern(pattern).withChronology(chronology).withLocale(locale);
String datetimeString = formatter.print(timestamp);
return Slices.wrappedBuffer(datetimeString.getBytes(Charsets.UTF_8));
}
private Page generateZeroPage(List<Type> types, int rowsCount, int fieldLength) {
byte[] constantBytes = new byte[fieldLength];
Arrays.fill(constantBytes, (byte) 42);
Slice constantSlice = Slices.wrappedBuffer(constantBytes);
Block[] blocks = new Block[types.size()];
for (int i = 0; i < blocks.length; i++) {
blocks[i] = createZeroBlock(types.get(i), rowsCount, constantSlice);
}
return new Page(rowsCount, blocks);
}
public GenericHiveRecordCursor(
RecordReader<K, V> recordReader,
long totalBytes,
Properties splitSchema,
List<HivePartitionKey> partitionKeys,
List<HiveColumnHandle> columns,
DateTimeZone hiveStorageTimeZone,
DateTimeZone sessionTimeZone) {
checkNotNull(recordReader, "recordReader is null");
checkArgument(totalBytes >= 0, "totalBytes is negative");
checkNotNull(splitSchema, "splitSchema is null");
checkNotNull(partitionKeys, "partitionKeys is null");
checkNotNull(columns, "columns is null");
checkArgument(!columns.isEmpty(), "columns is empty");
checkNotNull(hiveStorageTimeZone, "hiveStorageTimeZone is null");
checkNotNull(sessionTimeZone, "sessionTimeZone is null");
this.recordReader = recordReader;
this.totalBytes = totalBytes;
this.key = recordReader.createKey();
this.value = recordReader.createValue();
this.hiveStorageTimeZone = hiveStorageTimeZone;
this.sessionTimeZone = sessionTimeZone;
this.deserializer = getDeserializer(splitSchema);
this.rowInspector = getTableObjectInspector(deserializer);
int size = columns.size();
String[] names = new String[size];
this.types = new Type[size];
this.hiveTypes = new HiveType[size];
this.structFields = new StructField[size];
this.fieldInspectors = new ObjectInspector[size];
this.isPartitionColumn = new boolean[size];
this.loaded = new boolean[size];
this.booleans = new boolean[size];
this.longs = new long[size];
this.doubles = new double[size];
this.slices = new Slice[size];
this.nulls = new boolean[size];
// initialize data columns
for (int i = 0; i < columns.size(); i++) {
HiveColumnHandle column = columns.get(i);
names[i] = column.getName();
types[i] = column.getType();
hiveTypes[i] = column.getHiveType();
if (!column.isPartitionKey()) {
StructField field = rowInspector.getStructFieldRef(column.getName());
structFields[i] = field;
fieldInspectors[i] = field.getFieldObjectInspector();
}
isPartitionColumn[i] = column.isPartitionKey();
}
// parse requested partition columns
Map<String, HivePartitionKey> partitionKeysByName =
uniqueIndex(partitionKeys, HivePartitionKey.nameGetter());
for (int columnIndex = 0; columnIndex < columns.size(); columnIndex++) {
HiveColumnHandle column = columns.get(columnIndex);
if (column.isPartitionKey()) {
HivePartitionKey partitionKey = partitionKeysByName.get(column.getName());
checkArgument(partitionKey != null, "Unknown partition key %s", column.getName());
byte[] bytes = partitionKey.getValue().getBytes(Charsets.UTF_8);
Type type = types[columnIndex];
if (BOOLEAN.equals(type)) {
if (isTrue(bytes, 0, bytes.length)) {
booleans[columnIndex] = true;
} else if (isFalse(bytes, 0, bytes.length)) {
booleans[columnIndex] = false;
} else {
String valueString = new String(bytes, Charsets.UTF_8);
throw new IllegalArgumentException(
String.format(
"Invalid partition value '%s' for BOOLEAN partition key %s",
valueString, names[columnIndex]));
}
} else if (BIGINT.equals(type)) {
if (bytes.length == 0) {
throw new IllegalArgumentException(
String.format(
"Invalid partition value '' for BIGINT partition key %s", names[columnIndex]));
}
longs[columnIndex] = parseLong(bytes, 0, bytes.length);
} else if (DOUBLE.equals(type)) {
if (bytes.length == 0) {
throw new IllegalArgumentException(
String.format(
"Invalid partition value '' for DOUBLE partition key %s", names[columnIndex]));
}
doubles[columnIndex] = parseDouble(bytes, 0, bytes.length);
} else if (VARCHAR.equals(type)) {
slices[columnIndex] = Slices.wrappedBuffer(Arrays.copyOf(bytes, bytes.length));
} else {
throw new UnsupportedOperationException("Unsupported column type: " + type);
}
}
}
}
private static void serializePrimitive(
Type type, BlockBuilder builder, Object object, PrimitiveObjectInspector inspector) {
requireNonNull(builder, "parent builder is null");
if (object == null) {
builder.appendNull();
return;
}
switch (inspector.getPrimitiveCategory()) {
case BOOLEAN:
BooleanType.BOOLEAN.writeBoolean(builder, ((BooleanObjectInspector) inspector).get(object));
return;
case BYTE:
TinyintType.TINYINT.writeLong(builder, ((ByteObjectInspector) inspector).get(object));
return;
case SHORT:
SmallintType.SMALLINT.writeLong(builder, ((ShortObjectInspector) inspector).get(object));
return;
case INT:
IntegerType.INTEGER.writeLong(builder, ((IntObjectInspector) inspector).get(object));
return;
case LONG:
BigintType.BIGINT.writeLong(builder, ((LongObjectInspector) inspector).get(object));
return;
case FLOAT:
DoubleType.DOUBLE.writeDouble(builder, ((FloatObjectInspector) inspector).get(object));
return;
case DOUBLE:
DoubleType.DOUBLE.writeDouble(builder, ((DoubleObjectInspector) inspector).get(object));
return;
case STRING:
type.writeSlice(
builder,
Slices.utf8Slice(((StringObjectInspector) inspector).getPrimitiveJavaObject(object)));
return;
case VARCHAR:
type.writeSlice(
builder,
Slices.utf8Slice(
((HiveVarcharObjectInspector) inspector)
.getPrimitiveJavaObject(object)
.getValue()));
return;
case CHAR:
CharType charType = checkType(type, CharType.class, "type");
HiveChar hiveChar = ((HiveCharObjectInspector) inspector).getPrimitiveJavaObject(object);
type.writeSlice(
builder,
trimSpacesAndTruncateToLength(
Slices.utf8Slice(hiveChar.getValue()), charType.getLength()));
return;
case DATE:
DateType.DATE.writeLong(builder, formatDateAsLong(object, (DateObjectInspector) inspector));
return;
case TIMESTAMP:
TimestampType.TIMESTAMP.writeLong(
builder, formatTimestampAsLong(object, (TimestampObjectInspector) inspector));
return;
case BINARY:
VARBINARY.writeSlice(
builder,
Slices.wrappedBuffer(
((BinaryObjectInspector) inspector).getPrimitiveJavaObject(object)));
return;
case DECIMAL:
DecimalType decimalType = checkType(type, DecimalType.class, "type");
HiveDecimalWritable hiveDecimal =
((HiveDecimalObjectInspector) inspector).getPrimitiveWritableObject(object);
if (decimalType.isShort()) {
decimalType.writeLong(
builder, DecimalUtils.getShortDecimalValue(hiveDecimal, decimalType.getScale()));
} else {
decimalType.writeSlice(
builder, DecimalUtils.getLongDecimalValue(hiveDecimal, decimalType.getScale()));
}
return;
}
throw new RuntimeException("Unknown primitive type: " + inspector.getPrimitiveCategory());
}
private StringLiteral(Optional<NodeLocation> location, String value) {
super(location);
requireNonNull(value, "value is null");
this.value = value;
this.slice = Slices.wrappedBuffer(value.getBytes(UTF_8));
}
@Override
public void addBinary(Binary value) {
nulls[fieldIndex] = false;
slices[fieldIndex] = Slices.wrappedBuffer(value.getBytes());
}
@Override
public void addBinary(Binary value) {
VARBINARY.writeSlice(builder, Slices.wrappedBuffer(value.getBytes()));
wroteValue = true;
}
public ParquetHiveRecordCursor(
Configuration configuration,
Path path,
long start,
long length,
Properties splitSchema,
List<HivePartitionKey> partitionKeys,
List<HiveColumnHandle> columns,
boolean useParquetColumnNames,
TypeManager typeManager) {
requireNonNull(path, "path is null");
checkArgument(length >= 0, "totalBytes is negative");
requireNonNull(splitSchema, "splitSchema is null");
requireNonNull(partitionKeys, "partitionKeys is null");
requireNonNull(columns, "columns is null");
this.totalBytes = length;
int size = columns.size();
this.names = new String[size];
this.types = new Type[size];
this.isPartitionColumn = new boolean[size];
this.booleans = new boolean[size];
this.longs = new long[size];
this.doubles = new double[size];
this.slices = new Slice[size];
this.objects = new Object[size];
this.nulls = new boolean[size];
this.nullsRowDefault = new boolean[size];
for (int i = 0; i < columns.size(); i++) {
HiveColumnHandle column = columns.get(i);
names[i] = column.getName();
types[i] = typeManager.getType(column.getTypeSignature());
isPartitionColumn[i] = column.isPartitionKey();
nullsRowDefault[i] = !column.isPartitionKey();
}
this.recordReader =
createParquetRecordReader(
configuration, path, start, length, columns, useParquetColumnNames);
// parse requested partition columns
Map<String, HivePartitionKey> partitionKeysByName =
uniqueIndex(partitionKeys, HivePartitionKey::getName);
for (int columnIndex = 0; columnIndex < columns.size(); columnIndex++) {
HiveColumnHandle column = columns.get(columnIndex);
if (column.isPartitionKey()) {
HivePartitionKey partitionKey = partitionKeysByName.get(column.getName());
checkArgument(partitionKey != null, "Unknown partition key %s", column.getName());
byte[] bytes = partitionKey.getValue().getBytes(UTF_8);
String name = names[columnIndex];
Type type = types[columnIndex];
if (HiveUtil.isHiveNull(bytes)) {
nullsRowDefault[columnIndex] = true;
} else if (type.equals(BOOLEAN)) {
booleans[columnIndex] = booleanPartitionKey(partitionKey.getValue(), name);
} else if (type.equals(BIGINT)) {
longs[columnIndex] = bigintPartitionKey(partitionKey.getValue(), name);
} else if (type.equals(DOUBLE)) {
doubles[columnIndex] = doublePartitionKey(partitionKey.getValue(), name);
} else if (type.equals(VARCHAR)) {
slices[columnIndex] = Slices.wrappedBuffer(bytes);
} else {
throw new PrestoException(
NOT_SUPPORTED,
format(
"Unsupported column type %s for partition key: %s", type.getDisplayName(), name));
}
}
}
}
public static Slice base64Decode(byte[] bytes) {
return Slices.wrappedBuffer(Base64.getDecoder().decode(bytes));
}
