private void testRoundTrip(HiveColumnHandle expected) {
String json = codec.toJson(expected);
HiveColumnHandle actual = codec.fromJson(json);
assertEquals(actual.getClientId(), expected.getClientId());
assertEquals(actual.getName(), expected.getName());
assertEquals(actual.getHiveType(), expected.getHiveType());
assertEquals(actual.getHiveColumnIndex(), expected.getHiveColumnIndex());
assertEquals(actual.isPartitionKey(), expected.isPartitionKey());
}
public PrestoReadSupport(
boolean useParquetColumnNames, List<HiveColumnHandle> columns, MessageType messageType) {
this.columns = columns;
this.useParquetColumnNames = useParquetColumnNames;
ImmutableList.Builder<Converter> converters = ImmutableList.builder();
for (int i = 0; i < columns.size(); i++) {
HiveColumnHandle column = columns.get(i);
if (!column.isPartitionKey()) {
parquet.schema.Type parquetType = getParquetType(column, messageType);
if (parquetType == null) {
continue;
}
if (parquetType.isPrimitive()) {
converters.add(new ParquetPrimitiveColumnConverter(i));
} else {
GroupType groupType = parquetType.asGroupType();
switch (column.getTypeSignature().getBase()) {
case ARRAY:
ParquetColumnConverter listConverter =
new ParquetColumnConverter(
new ParquetListConverter(types[i], groupType.getName(), groupType), i);
converters.add(listConverter);
break;
case StandardTypes.MAP:
ParquetColumnConverter mapConverter =
new ParquetColumnConverter(
new ParquetMapConverter(types[i], groupType.getName(), groupType), i);
converters.add(mapConverter);
break;
case ROW:
ParquetColumnConverter rowConverter =
new ParquetColumnConverter(
new ParquetStructConverter(types[i], groupType.getName(), groupType), i);
converters.add(rowConverter);
break;
default:
throw new IllegalArgumentException(
"Group column "
+ groupType.getName()
+ " type "
+ groupType.getOriginalType()
+ " not supported");
}
}
}
}
this.converters = converters.build();
}
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);
}
}
}
}
public GenericHiveRecordCursor(
RecordReader<K, V> recordReader,
long totalBytes,
Properties splitSchema,
List<HivePartitionKey> partitionKeys,
List<HiveColumnHandle> columns,
DateTimeZone hiveStorageTimeZone,
TypeManager typeManager) {
requireNonNull(recordReader, "recordReader is null");
checkArgument(totalBytes >= 0, "totalBytes is negative");
requireNonNull(splitSchema, "splitSchema is null");
requireNonNull(partitionKeys, "partitionKeys is null");
requireNonNull(columns, "columns is null");
requireNonNull(hiveStorageTimeZone, "hiveStorageTimeZone is null");
this.recordReader = recordReader;
this.totalBytes = totalBytes;
this.key = recordReader.createKey();
this.value = recordReader.createValue();
this.hiveStorageTimeZone = hiveStorageTimeZone;
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.objects = new Object[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] = typeManager.getType(column.getTypeSignature());
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::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)) {
nulls[columnIndex] = true;
} else if (BOOLEAN.equals(type)) {
booleans[columnIndex] = booleanPartitionKey(partitionKey.getValue(), name);
} else if (BIGINT.equals(type)) {
longs[columnIndex] = bigintPartitionKey(partitionKey.getValue(), name);
} else if (INTEGER.equals(type)) {
longs[columnIndex] = integerPartitionKey(partitionKey.getValue(), name);
} else if (SMALLINT.equals(type)) {
longs[columnIndex] = smallintPartitionKey(partitionKey.getValue(), name);
} else if (TINYINT.equals(type)) {
longs[columnIndex] = tinyintPartitionKey(partitionKey.getValue(), name);
} else if (DOUBLE.equals(type)) {
doubles[columnIndex] = doublePartitionKey(partitionKey.getValue(), name);
} else if (isVarcharType(type)) {
slices[columnIndex] = varcharPartitionKey(partitionKey.getValue(), name, type);
} else if (DATE.equals(type)) {
longs[columnIndex] = datePartitionKey(partitionKey.getValue(), name);
} else if (TIMESTAMP.equals(type)) {
longs[columnIndex] =
timestampPartitionKey(partitionKey.getValue(), hiveStorageTimeZone, name);
} else if (isShortDecimal(type)) {
longs[columnIndex] =
shortDecimalPartitionKey(partitionKey.getValue(), (DecimalType) type, name);
} else if (isLongDecimal(type)) {
slices[columnIndex] =
longDecimalPartitionKey(partitionKey.getValue(), (DecimalType) type, name);
} else {
throw new PrestoException(
NOT_SUPPORTED,
format(
"Unsupported column type %s for partition key: %s", type.getDisplayName(), name));
}
}
}
}
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));
}
}
}
}