/// <summary>
/// Test read, to be defined
/// </summary>
public DataSet Read()
{
_readerOptions.Validate();
_meta = ReadMetadata();
var footer = new ThriftFooter(_meta);
var pathToValues = new Dictionary <string, IList>();
long pos = 0;
long rowsRead = 0;
foreach (Thrift.RowGroup rg in _meta.Row_groups)
{
//check whether to skip RG completely
if ((_readerOptions.Count != -1 && rowsRead >= _readerOptions.Count) ||
(_readerOptions.Offset > pos + rg.Num_rows - 1))
{
pos += rg.Num_rows;
continue;
}
long offset = Math.Max(0, _readerOptions.Offset - pos);
long count = _readerOptions.Count == -1 ? rg.Num_rows : Math.Min(_readerOptions.Count - rowsRead, rg.Num_rows);
for (int icol = 0; icol < rg.Columns.Count; icol++)
{
Thrift.ColumnChunk cc = rg.Columns[icol];
string path = cc.GetPath();
if (_fieldPredicates != null && !_fieldPredicates.Any(p => p.IsMatch(cc, path)))
{
continue;
}
var columnarReader = new ColumnarReader(_input, cc, footer, _formatOptions);
try
{
IList chunkValues = columnarReader.Read(offset, count);
if (!pathToValues.TryGetValue(path, out IList allValues))
{
pathToValues[path] = chunkValues;
}
else
{
foreach (object v in chunkValues)
{
allValues.Add(v);
}
}
if (icol == 0)
{
//todo: this may not work
rowsRead += chunkValues.Count;
}
}
catch (Exception ex)
{
throw new ParquetException($"fatal error reading column '{path}'", ex);
}
}
pos += rg.Num_rows;
}
Schema schema = footer.CreateModelSchema(_formatOptions);
schema = schema.Filter(_fieldPredicates);
var ds = new DataSet(schema, pathToValues, _meta.Num_rows, _meta.Created_by);
Dictionary <string, string> customMetadata = footer.CustomMetadata;
if (customMetadata != null)
{
ds.Metadata.Custom.AddRange(customMetadata);
}
ds.Thrift = _meta;
return(ds);
}
/// <summary>
/// Test read, to be defined
/// </summary>
public DataSet Read()
{
_readerOptions.Validate();
_meta = ReadMetadata();
var footer = new ThriftFooter(_meta);
var metaParser = new FileMetadataParser(_meta);
Schema schema = metaParser.ParseSchema(_formatOptions);
SchemaElement schema2 = metaParser.ParseSchemaExperimental(_formatOptions);
var pathToValues = new Dictionary <string, IList>();
long pos = 0;
long rowsRead = 0;
foreach (Thrift.RowGroup rg in _meta.Row_groups)
{
//check whether to skip RG completely
if ((_readerOptions.Count != -1 && rowsRead >= _readerOptions.Count) ||
(_readerOptions.Offset > pos + rg.Num_rows - 1))
{
pos += rg.Num_rows;
continue;
}
long offset = Math.Max(0, _readerOptions.Offset - pos);
long count = _readerOptions.Count == -1 ? rg.Num_rows : Math.Min(_readerOptions.Count - rowsRead, rg.Num_rows);
for (int icol = 0; icol < rg.Columns.Count; icol++)
{
Thrift.ColumnChunk cc = rg.Columns[icol];
SchemaElement se = schema[cc];
var p = new ColumnReader(cc, se, _input, ThriftStream, _formatOptions);
var columnarReader = new ColumnarReader(_input, cc, footer, _formatOptions);
try
{
IList chunkValues = p.Read(offset, count);
//columnarReader.Read(offset, count);
if (!pathToValues.TryGetValue(se.Path, out IList allValues))
{
pathToValues[se.Path] = chunkValues;
}
else
{
foreach (object v in chunkValues)
{
allValues.Add(v);
}
}
if (icol == 0)
{
//todo: this may not work
rowsRead += chunkValues.Count;
}
}
catch (Exception ex)
{
throw new ParquetException($"fatal error reading column '{se}'", ex);
}
}
pos += rg.Num_rows;
}
var ds = new DataSet(schema, pathToValues, _meta.Num_rows, _meta.Created_by);
metaParser.AddMeta(ds);
ds.Thrift = _meta;
return(ds);
}
