/// <summary>
/// Write out dataset to the output stream
/// </summary>
/// <param name="dataSet">Dataset to write</param>
/// <param name="compression">Compression method</param>
/// <param name="append">When true, appends to the file, otherwise creates a new file.</param>
public void Write(DataSet dataSet, CompressionMethod compression = CompressionMethod.Gzip, bool append = false)
{
PrepareFile(dataSet, append);
var stats = new DataSetStats(dataSet);
int offset = 0;
int count;
do
{
count = Math.Min(_writerOptions.RowGroupsSize, dataSet.Count - offset);
Thrift.RowGroup rg = _meta.AddRowGroup();
long rgStartPos = _output.Position;
rg.Columns = dataSet.Schema.Elements
.Select(c =>
Write(c, dataSet.GetColumn(c.Name, offset, count), compression, stats.GetColumnStats(c)))
.ToList();
//row group's size is a sum of _uncompressed_ sizes of all columns in it, including the headers
//luckily ColumnChunk already contains sizes of page+header in it's meta
rg.Total_byte_size = rg.Columns.Sum(c => c.Meta_data.Total_compressed_size);
rg.Num_rows = count;
offset += _writerOptions.RowGroupsSize;
}while (offset < dataSet.Count);
_dataWritten = true;
}
/// <summary>
/// Write out dataset to the output stream
/// </summary>
/// <param name="dataSet">Dataset to write</param>
/// <param name="compression">Compression method</param>
/// <param name="append">When true, appends to the file, otherwise creates a new file.</param>
public void Write(DataSet dataSet, CompressionMethod compression = CompressionMethod.Gzip, bool append = false)
{
PrepareFile(dataSet, append);
int offset = 0;
int count;
do
{
count = Math.Min(_writerOptions.RowGroupsSize, dataSet.Count - offset);
Thrift.RowGroup rg = _meta.AddRowGroup();
long rgStartPos = Stream.Position;
rg.Columns = new List <Thrift.ColumnChunk>();
foreach (SchemaElement se in dataSet.Schema.Flatten())
{
var cw = new ColumnWriter(Stream, ThriftStream, _meta, se, compression, _formatOptions, _writerOptions);
IList values = dataSet.GetColumn(se, offset, count);
Thrift.ColumnChunk chunk = cw.Write(offset, count, values);
rg.Columns.Add(chunk);
}
//row group's size is a sum of _uncompressed_ sizes of all columns in it, including the headers
//luckily ColumnChunk already contains sizes of page+header in it's meta
rg.Total_byte_size = rg.Columns.Sum(c => c.Meta_data.Total_compressed_size);
rg.Num_rows = count;
offset += _writerOptions.RowGroupsSize;
}while (offset < dataSet.Count);
_dataWritten = true;
}
internal ParquetRowGroupReader(
Thrift.RowGroup rowGroup,
ThriftFooter footer,
Stream stream, ThriftStream thriftStream,
ParquetOptions parquetOptions)
{
_rowGroup = rowGroup ?? throw new ArgumentNullException(nameof(rowGroup));
_footer = footer ?? throw new ArgumentNullException(nameof(footer));
_stream = stream ?? throw new ArgumentNullException(nameof(stream));
_thriftStream = thriftStream ?? throw new ArgumentNullException(nameof(thriftStream));
_parquetOptions = parquetOptions ?? throw new ArgumentNullException(nameof(parquetOptions));
//cache chunks
foreach (Thrift.ColumnChunk thriftChunk in _rowGroup.Columns)
{
string path = thriftChunk.GetPath();
_pathToChunk[path] = thriftChunk;
}
}
internal ParquetRowGroupWriter(Schema schema,
Stream stream,
ThriftStream thriftStream,
ThriftFooter footer,
CompressionMethod compressionMethod,
ParquetOptions formatOptions)
{
_schema = schema ?? throw new ArgumentNullException(nameof(schema));
_stream = stream ?? throw new ArgumentNullException(nameof(stream));
_thriftStream = thriftStream ?? throw new ArgumentNullException(nameof(thriftStream));
_footer = footer ?? throw new ArgumentNullException(nameof(footer));
_compressionMethod = compressionMethod;
_formatOptions = formatOptions;
_thriftRowGroup = _footer.AddRowGroup();
_rgStartPos = _stream.Position;
_thriftRowGroup.Columns = new List <Thrift.ColumnChunk>();
_thschema = _footer.GetWriteableSchema();
}
/// <summary>
/// Write out dataset to the output stream
/// </summary>
/// <param name="dataSet">Dataset to write</param>
/// <param name="compression">Compression method</param>
public void Write(DataSet dataSet, CompressionMethod compression = CompressionMethod.Gzip)
{
_meta.AddSchema(dataSet);
var stats = new DataSetStats(dataSet);
long totalCount = dataSet.Count;
Thrift.RowGroup rg = _meta.AddRowGroup();
long rgStartPos = _output.Position;
rg.Columns = dataSet.Schema.Elements
.Select(c =>
Write(c, dataSet.GetColumn(c.Name), compression, stats.GetColumnStats(c)))
.ToList();
//row group's size is a sum of _uncompressed_ sizes of all columns in it
rg.Total_byte_size = rg.Columns.Sum(c => c.Meta_data.Total_uncompressed_size);
rg.Num_rows = dataSet.Count;
_dataWritten = true;
}
/// <summary>
/// Write out dataset to the output stream
/// </summary>
/// <param name="dataSet">Dataset to write</param>
/// <param name="compression">Compression method</param>
/// <param name="append">When true, appends to the file, otherwise creates a new file.</param>
public void Write(DataSet dataSet, CompressionMethod compression = CompressionMethod.Gzip, bool append = false)
{
PrepareFile(dataSet, append);
_footer.CustomMetadata = dataSet.Metadata.Custom;
int offset = 0;
int count;
List <Thrift.SchemaElement> writeableSchema = _footer.GetWriteableSchema().ToList();
do
{
count = Math.Min(_writerOptions.RowGroupsSize, dataSet.Count - offset);
Thrift.RowGroup rg = _footer.AddRowGroup();
long rgStartPos = Stream.Position;
rg.Columns = new List <Thrift.ColumnChunk>();
foreach (Thrift.SchemaElement tse in writeableSchema)
{
List <string> path = _footer.GetPath(tse);
string flatPath = string.Join(Schema.PathSeparator, path);
var cw = new ColumnarWriter(Stream, ThriftStream, _footer, tse, path, compression, _formatOptions, _writerOptions);
IList values = dataSet.GetColumn(flatPath, offset, count);
Thrift.ColumnChunk chunk = cw.Write(offset, count, values);
rg.Columns.Add(chunk);
}
//row group's size is a sum of _uncompressed_ sizes of all columns in it, including the headers
//luckily ColumnChunk already contains sizes of page+header in it's meta
rg.Total_byte_size = rg.Columns.Sum(c => c.Meta_data.Total_compressed_size);
rg.Num_rows = count;
offset += _writerOptions.RowGroupsSize;
}while (offset < dataSet.Count);
_dataWritten = true;
}
public void Read(TProtocol iprot)
{
iprot.IncrementRecursionDepth();
try
{
bool isset_version = false;
bool isset_schema = false;
bool isset_num_rows = false;
bool isset_row_groups = false;
TField field;
iprot.ReadStructBegin();
while (true)
{
field = iprot.ReadFieldBegin();
if (field.Type == TType.Stop)
{
break;
}
switch (field.ID)
{
case 1:
if (field.Type == TType.I32)
{
Version = iprot.ReadI32();
isset_version = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 2:
if (field.Type == TType.List)
{
{
Schema = new List <SchemaElement>();
TList _list24 = iprot.ReadListBegin();
for (int _i25 = 0; _i25 < _list24.Count; ++_i25)
{
SchemaElement _elem26;
_elem26 = new SchemaElement();
_elem26.Read(iprot);
Schema.Add(_elem26);
}
iprot.ReadListEnd();
}
isset_schema = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 3:
if (field.Type == TType.I64)
{
Num_rows = iprot.ReadI64();
isset_num_rows = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 4:
if (field.Type == TType.List)
{
{
Row_groups = new List <RowGroup>();
TList _list27 = iprot.ReadListBegin();
for (int _i28 = 0; _i28 < _list27.Count; ++_i28)
{
RowGroup _elem29;
_elem29 = new RowGroup();
_elem29.Read(iprot);
Row_groups.Add(_elem29);
}
iprot.ReadListEnd();
}
isset_row_groups = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 5:
if (field.Type == TType.List)
{
{
Key_value_metadata = new List <KeyValue>();
TList _list30 = iprot.ReadListBegin();
for (int _i31 = 0; _i31 < _list30.Count; ++_i31)
{
KeyValue _elem32;
_elem32 = new KeyValue();
_elem32.Read(iprot);
Key_value_metadata.Add(_elem32);
}
iprot.ReadListEnd();
}
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 6:
if (field.Type == TType.String)
{
Created_by = iprot.ReadString();
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
default:
TProtocolUtil.Skip(iprot, field.Type);
break;
}
iprot.ReadFieldEnd();
}
iprot.ReadStructEnd();
if (!isset_version)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_schema)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_num_rows)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_row_groups)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
}
finally
{
iprot.DecrementRecursionDepth();
}
}
public void Read(TProtocol iprot)
{
iprot.IncrementRecursionDepth();
try
{
bool isset_version = false;
bool isset_schema = false;
bool isset_num_rows = false;
bool isset_row_groups = false;
TField field;
iprot.ReadStructBegin();
while (true)
{
field = iprot.ReadFieldBegin();
if (field.Type == TType.Stop)
{
break;
}
switch (field.ID)
{
case 1:
if (field.Type == TType.I32)
{
Version = iprot.ReadI32();
isset_version = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 2:
if (field.Type == TType.List)
{
{
Schema = new List <SchemaElement>();
TList _list48 = iprot.ReadListBegin();
for (int _i49 = 0; _i49 < _list48.Count; ++_i49)
{
SchemaElement _elem50;
_elem50 = new SchemaElement();
_elem50.Read(iprot);
Schema.Add(_elem50);
}
iprot.ReadListEnd();
}
isset_schema = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 3:
if (field.Type == TType.I64)
{
Num_rows = iprot.ReadI64();
isset_num_rows = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 4:
if (field.Type == TType.List)
{
{
Row_groups = new List <RowGroup>();
TList _list51 = iprot.ReadListBegin();
for (int _i52 = 0; _i52 < _list51.Count; ++_i52)
{
RowGroup _elem53;
_elem53 = new RowGroup();
_elem53.Read(iprot);
Row_groups.Add(_elem53);
}
iprot.ReadListEnd();
}
isset_row_groups = true;
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 5:
if (field.Type == TType.List)
{
{
Key_value_metadata = new List <KeyValue>();
TList _list54 = iprot.ReadListBegin();
for (int _i55 = 0; _i55 < _list54.Count; ++_i55)
{
KeyValue _elem56;
_elem56 = new KeyValue();
_elem56.Read(iprot);
Key_value_metadata.Add(_elem56);
}
iprot.ReadListEnd();
}
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 6:
if (field.Type == TType.String)
{
Created_by = iprot.ReadString();
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 7:
if (field.Type == TType.List)
{
{
Column_orders = new List <ColumnOrder>();
TList _list57 = iprot.ReadListBegin();
for (int _i58 = 0; _i58 < _list57.Count; ++_i58)
{
ColumnOrder _elem59;
_elem59 = new ColumnOrder();
_elem59.Read(iprot);
Column_orders.Add(_elem59);
}
iprot.ReadListEnd();
}
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 8:
if (field.Type == TType.Struct)
{
Encryption_algorithm = new EncryptionAlgorithm();
Encryption_algorithm.Read(iprot);
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
case 9:
if (field.Type == TType.String)
{
Footer_signing_key_metadata = iprot.ReadBinary();
}
else
{
TProtocolUtil.Skip(iprot, field.Type);
}
break;
default:
TProtocolUtil.Skip(iprot, field.Type);
break;
}
iprot.ReadFieldEnd();
}
iprot.ReadStructEnd();
if (!isset_version)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_schema)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_num_rows)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_row_groups)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
}
finally
{
iprot.DecrementRecursionDepth();
}
}
public async Task ReadAsync(TProtocol iprot, CancellationToken cancellationToken)
{
iprot.IncrementRecursionDepth();
try
{
bool isset_version = false;
bool isset_schema = false;
bool isset_num_rows = false;
bool isset_row_groups = false;
TField field;
await iprot.ReadStructBeginAsync(cancellationToken);
while (true)
{
field = await iprot.ReadFieldBeginAsync(cancellationToken);
if (field.Type == TType.Stop)
{
break;
}
switch (field.ID)
{
case 1:
if (field.Type == TType.I32)
{
Version = await iprot.ReadI32Async(cancellationToken);
isset_version = true;
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 2:
if (field.Type == TType.List)
{
{
Schema = new List <SchemaElement>();
TList _list44 = await iprot.ReadListBeginAsync(cancellationToken);
for (int _i45 = 0; _i45 < _list44.Count; ++_i45)
{
SchemaElement _elem46;
_elem46 = new SchemaElement();
await _elem46.ReadAsync(iprot, cancellationToken);
Schema.Add(_elem46);
}
await iprot.ReadListEndAsync(cancellationToken);
}
isset_schema = true;
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 3:
if (field.Type == TType.I64)
{
Num_rows = await iprot.ReadI64Async(cancellationToken);
isset_num_rows = true;
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 4:
if (field.Type == TType.List)
{
{
Row_groups = new List <RowGroup>();
TList _list47 = await iprot.ReadListBeginAsync(cancellationToken);
for (int _i48 = 0; _i48 < _list47.Count; ++_i48)
{
RowGroup _elem49;
_elem49 = new RowGroup();
await _elem49.ReadAsync(iprot, cancellationToken);
Row_groups.Add(_elem49);
}
await iprot.ReadListEndAsync(cancellationToken);
}
isset_row_groups = true;
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 5:
if (field.Type == TType.List)
{
{
Key_value_metadata = new List <KeyValue>();
TList _list50 = await iprot.ReadListBeginAsync(cancellationToken);
for (int _i51 = 0; _i51 < _list50.Count; ++_i51)
{
KeyValue _elem52;
_elem52 = new KeyValue();
await _elem52.ReadAsync(iprot, cancellationToken);
Key_value_metadata.Add(_elem52);
}
await iprot.ReadListEndAsync(cancellationToken);
}
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 6:
if (field.Type == TType.String)
{
Created_by = await iprot.ReadStringAsync(cancellationToken);
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
case 7:
if (field.Type == TType.List)
{
{
Column_orders = new List <ColumnOrder>();
TList _list53 = await iprot.ReadListBeginAsync(cancellationToken);
for (int _i54 = 0; _i54 < _list53.Count; ++_i54)
{
ColumnOrder _elem55;
_elem55 = new ColumnOrder();
await _elem55.ReadAsync(iprot, cancellationToken);
Column_orders.Add(_elem55);
}
await iprot.ReadListEndAsync(cancellationToken);
}
}
else
{
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
}
break;
default:
await TProtocolUtil.SkipAsync(iprot, field.Type, cancellationToken);
break;
}
await iprot.ReadFieldEndAsync(cancellationToken);
}
await iprot.ReadStructEndAsync(cancellationToken);
if (!isset_version)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_schema)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_num_rows)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
if (!isset_row_groups)
{
throw new TProtocolException(TProtocolException.INVALID_DATA);
}
}
finally
{
iprot.DecrementRecursionDepth();
}
}
