private (ICollection definitions, ICollection repetitions, List <int> indexes) ReadDataPage(Thrift.PageHeader ph, IList destination, long maxValues)
{
byte[] data = ReadRawBytes(ph, _inputStream);
using (var dataStream = new MemoryStream(data))
{
using (var reader = new BinaryReader(dataStream))
{
//todo: read repetition levels (only relevant for nested columns)
//check if there are definitions at all
bool hasDefinitions = _schemaElement.Thrift.Repetition_type == Thrift.FieldRepetitionType.OPTIONAL;
List <int> definitions = hasDefinitions
? ReadDefinitionLevels(reader, (int)maxValues)
: null;
// these are pointers back to the Values table - lookup on values
List <int> indexes = ReadColumnValues(reader, ph.Data_page_header.Encoding, destination, maxValues);
//trim output if it exceeds max number of values
int numValues = ph.Data_page_header.Num_values;
if (definitions != null)
{
ValueMerger.Trim(definitions, numValues);
}
if (indexes != null)
{
ValueMerger.Trim(indexes, numValues);
}
return(definitions, null, indexes);
}
}
}
public IList Read(long offset, long count)
{
long fileOffset = GetFileOffset();
long maxValues = _thriftColumnChunk.Meta_data.Num_values;
_inputStream.Seek(fileOffset, SeekOrigin.Begin);
IList dictionary = null;
List <int> indexes = null;
List <int> repetitions = null;
List <int> definitions = null;
IList values = null;
//there can be only one dictionary page in column
Thrift.PageHeader ph = _thriftStream.Read <Thrift.PageHeader>();
if (TryReadDictionaryPage(ph, out dictionary))
{
ph = _thriftStream.Read <Thrift.PageHeader>();
}
int pagesRead = 0;
while (true)
{
int valuesSoFar = Math.Max(indexes == null ? 0 : indexes.Count, values == null ? 0 : values.Count);
PageData pd = ReadDataPage(ph, maxValues - valuesSoFar);
repetitions = AssignOrAdd(repetitions, pd.repetitions);
definitions = AssignOrAdd(definitions, pd.definitions);
indexes = AssignOrAdd(indexes, pd.indexes);
values = AssignOrAdd(values, pd.values);
pagesRead++;
int totalCount = Math.Max(
(values == null ? 0 : values.Count) +
(indexes == null ? 0 : indexes.Count),
(definitions == null ? 0 : definitions.Count));
if (totalCount >= maxValues)
{
break; //limit reached
}
ph = _thriftStream.Read <Thrift.PageHeader>();
if (ph.Type != Thrift.PageType.DATA_PAGE)
{
break;
}
}
IList mergedValues = new ValueMerger(
_maxDefinitionLevel,
_maxRepetitionLevel,
() => _dataTypeHandler.CreateEmptyList(_thriftSchemaElement.IsNullable(), false, 0),
values ?? _dataTypeHandler.CreateEmptyList(_thriftSchemaElement.IsNullable(), false, 0))
.Apply(dictionary, definitions, repetitions, indexes, (int)maxValues);
mergedValues.Trim((int)offset, (int)count);
return(mergedValues);
}
private PageData ReadDataPage(Thrift.PageHeader ph, IList destination, long maxValues)
{
byte[] data = ReadRawBytes(ph, _inputStream);
int max = ph.Data_page_header.Num_values;
using (var dataStream = new MemoryStream(data))
{
using (var reader = new BinaryReader(dataStream))
{
List <int> repetitions = _schema.HasRepetitionLevelsPage
? ReadRepetitionLevels(reader)
: null;
List <int> definitions = _schema.HasDefinitionLevelsPage
? ReadDefinitionLevels(reader)
: null;
// these are pointers back to the Values table - lookup on values
List <int> indexes = ReadColumnValues(reader, ph.Data_page_header.Encoding, destination, max);
//trim output if it exceeds max number of values
int numValues = ph.Data_page_header.Num_values;
if (!_schema.IsRepeated)
{
if (repetitions != null)
{
ValueMerger.TrimTail(repetitions, numValues);
}
if (definitions != null)
{
ValueMerger.TrimTail(definitions, numValues);
}
if (indexes != null)
{
ValueMerger.TrimTail(indexes, numValues);
}
}
return(new PageData {
definitions = definitions, repetitions = repetitions, indexes = indexes
});
}
}
}
public IList Read(long offset, long count)
{
IList values = TypeFactory.Create(_schema, _options);
//get the minimum offset, we'll just read pages in sequence
long fileOffset = new[] { _thriftChunk.Meta_data.Dictionary_page_offset, _thriftChunk.Meta_data.Data_page_offset }.Where(e => e != 0).Min();
long maxValues = _thriftChunk.Meta_data.Num_values;
_inputStream.Seek(fileOffset, SeekOrigin.Begin);
Thrift.PageHeader ph = _thrift.Read <Thrift.PageHeader>();
IList dictionaryPage = null;
List <int> indexes = null;
List <int> definitions = null;
List <int> repetitions = null;
//there can be only one dictionary page in column
if (ph.Type == Thrift.PageType.DICTIONARY_PAGE)
{
dictionaryPage = ReadDictionaryPage(ph);
ph = _thrift.Read <Thrift.PageHeader>(); //get next page after dictionary
}
int dataPageCount = 0;
while (true)
{
int valuesSoFar = Math.Max(indexes == null ? 0 : indexes.Count, values.Count);
PageData page = ReadDataPage(ph, values, maxValues - valuesSoFar);
indexes = AssignOrAdd(indexes, page.indexes);
definitions = AssignOrAdd(definitions, page.definitions);
repetitions = AssignOrAdd(repetitions, page.repetitions);
dataPageCount++;
int totalCount = Math.Max(
values.Count +
(indexes == null
? 0
: indexes.Count),
definitions == null ? 0 : definitions.Count);
if (totalCount >= maxValues)
{
break; //limit reached
}
ph = ReadDataPageHeader(dataPageCount); //get next page
if (ph.Type != Thrift.PageType.DATA_PAGE)
{
break;
}
}
IList mergedValues = new ValueMerger(_schema, _options, values)
.Apply(dictionaryPage, definitions, repetitions, indexes, (int)maxValues);
//todo: this won't work for nested arrays
ValueMerger.Trim(mergedValues, (int)offset, (int)count);
return(mergedValues);
}
public IList Read(string columnName)
{
IList values = TypeFactory.Create(_schemaElement);
//get the minimum offset, we'll just read pages in sequence
long offset = new[] { _thriftChunk.Meta_data.Dictionary_page_offset, _thriftChunk.Meta_data.Data_page_offset }.Where(e => e != 0).Min();
long maxValues = _thriftChunk.Meta_data.Num_values;
_inputStream.Seek(offset, SeekOrigin.Begin);
Thrift.PageHeader ph = _thrift.Read <Thrift.PageHeader>();
IList dictionaryPage = null;
List <int> indexes = null;
List <int> definitions = null;
//there can be only one dictionary page in column
if (ph.Type == Thrift.PageType.DICTIONARY_PAGE)
{
dictionaryPage = ReadDictionaryPage(ph);
ph = _thrift.Read <Thrift.PageHeader>(); //get next page after dictionary
}
int dataPageCount = 0;
while (true)
{
int valuesSoFar = Math.Max(indexes == null ? 0 : indexes.Count, values.Count);
var page = ReadDataPage(ph, values, maxValues - valuesSoFar);
//merge indexes
if (page.indexes != null)
{
if (indexes == null)
{
indexes = page.indexes;
}
else
{
indexes.AddRange(page.indexes);
}
}
if (page.definitions != null)
{
if (definitions == null)
{
definitions = (List <int>)page.definitions;
}
else
{
definitions.AddRange((List <int>)page.definitions);
}
}
dataPageCount++;
if (page.repetitions != null)
{
throw new NotImplementedException();
}
if ((values.Count >= maxValues) || (indexes != null && indexes.Count >= maxValues) || (definitions != null && definitions.Count >= maxValues))
{
break; //limit reached
}
/*IList acc1 = new ValueMerger(_schemaElement, values).Apply(dictionaryPage, definitions, indexes, maxValues);
* dictionaryPage = null;
* definitions = null;
* indexes = null;
* values.Clear();
* foreach (var el in acc1) acc.Add(el);*/
ph = _thrift.Read <Thrift.PageHeader>(); //get next page
if (ph.Type != Thrift.PageType.DATA_PAGE)
{
break;
}
}
IList mergedValues = new ValueMerger(_schemaElement, values).Apply(dictionaryPage, definitions, indexes, maxValues);
return(mergedValues);
}
