private void _Add(string line, DataTableWriter writer)
{
var convertedData = writer.Columns.Zip(_Parse(line), (c, str) => _Convert(ref c._type, str)).
ToList();
writer.AddRow(convertedData);
}
//public void Process(Func<IRow, int, bool> processor)
//{
// int index = 0;
// foreach (var item in _data) {
// if (!processor(item, index++))
// break;
// }
//}
/// <summary>
/// Creates a data table
/// </summary>
/// <param name="output">Optional stream to write the data table to</param>
public IDataTable Build(Stream output = null)
{
var writer = new DataTableWriter(Columns, output);
Process(writer);
return(writer.GetDataTable());
}
public IDataTable ConvertToNumeric(IDataTableVectoriser vectoriser = null, bool useTargetColumnIndex = true, Stream output = null)
{
var writer = new DataTableWriter(output);
vectoriser = vectoriser ?? GetVectoriser(useTargetColumnIndex);
// add the numeric columns
foreach (var name in vectoriser.ColumnNames)
{
writer.AddColumn(name, ColumnType.Float);
}
// add the classification label column
var classColumnIndex = TargetColumnIndex;
if (useTargetColumnIndex)
{
var classColumn = _column[classColumnIndex];
writer.AddColumn(classColumn.Name, ColumnType.String, true);
}
// vectorise each row
_Iterate((row, i) => {
var rowData = vectoriser.GetInput(row).Data.AsEnumerable().Cast <object>();
if (useTargetColumnIndex)
{
rowData = rowData.Concat(new object[] { row.GetField <string>(classColumnIndex) });
}
writer.AddRow(new DataTableRow(this, rowData.ToArray(), _rowConverter));
return(true);
});
return(writer.GetDataTable());
}
public (IDataTable Training, IDataTable Test) Split(int?randomSeed = null, double trainPercentage = 0.8, bool shuffle = true, Stream output1 = null, Stream output2 = null)
{
var input = Enumerable.Range(0, RowCount);
if (shuffle)
{
input = input.Shuffle(randomSeed);
}
var final = input.ToList();
int trainingCount = Convert.ToInt32(RowCount * trainPercentage);
var writer1 = new DataTableWriter(Columns, output1);
foreach (var row in GetRows(final.Take(trainingCount)))
{
writer1.Process(row);
}
var writer2 = new DataTableWriter(Columns, output2);
foreach (var row in GetRows(final.Skip(trainingCount)))
{
writer2.Process(row);
}
return(writer1.GetDataTable(), writer2.GetDataTable());
}
public IEnumerable <(IDataTable Training, IDataTable Validation)> Fold(int k, int?randomSeed = null, bool shuffle = true)
{
var input = Enumerable.Range(0, RowCount);
if (shuffle)
{
input = input.Shuffle(randomSeed);
}
var final = input.ToList();
var foldSize = final.Count / k;
for (var i = 0; i < k; i++)
{
var trainingRows = final.Take(i * foldSize).Concat(final.Skip((i + 1) * foldSize));
var validationRows = final.Skip(i * foldSize).Take(foldSize);
var writer1 = new DataTableWriter(Columns, null);
foreach (var row in GetRows(trainingRows))
{
writer1.Process(row);
}
var writer2 = new DataTableWriter(Columns, null);
foreach (var row in GetRows(validationRows))
{
writer2.Process(row);
}
yield return(writer1.GetDataTable(), writer2.GetDataTable());
}
}
public void WriteTo(Stream stream)
{
var writer = new DataTableWriter(Columns, stream);
Process(writer);
writer.Flush();
}
private void _Preview(string line, DataTableWriter writer)
{
int index = 0;
var fields = _Parse(line).ToList();
foreach (var column in writer.Columns)
{
_Convert(ref column._type, fields[index++]);
}
}
public IDataTable CopyWithRows(IEnumerable <int> rowIndex, Stream output = null)
{
var writer = new DataTableWriter(_column, output);
foreach (var row in GetRows(rowIndex))
{
writer.AddRow(row);
}
return(writer.GetDataTable());
}
public IDataTable Bag(int?count = null, Stream output = null, int?randomSeed = null)
{
var input = Enumerable.Range(0, RowCount).ToList().Bag(count ?? RowCount, randomSeed);
var writer = new DataTableWriter(Columns, output);
foreach (var row in GetRows(input))
{
writer.Process(row);
}
return(writer.GetDataTable());
}
private void _Preview(string line, DataTableWriter writer)
{
int index = 0;
var fields = _Parse(line).ToList();
// append empty strings to fill any blanks
for (int i = fields.Count, len = writer.Columns.Count; i < len; i++)
{
fields.Add(string.Empty);
}
foreach (var column in writer.Columns)
{
_Convert(ref column._type, fields[index++]);
}
}
public IDataTable Zip(IDataTable dataTable, Stream output = null)
{
var writer = new DataTableWriter(_column.Concat(dataTable.Columns), output);
_Iterate((row, i) =>
{
if (i >= dataTable.RowCount)
{
return(false);
}
writer.AddRow(row.Data.Concat(dataTable.GetRow(i).Data).ToList());
return(true);
});
return(writer.GetDataTable());
}
private DataTableWriter _DetermineHeaders(Stream stream, List <string> lines, bool checkForHeader, ref bool hasHeader)
{
// see if there is a header (all strings)
var firstLineTypes = _Parse(lines.First()).ToList();
if (checkForHeader)
{
hasHeader = firstLineTypes.All(str => _DetermineType(str) == ColumnType.String);
}
// get the list of header names
var headerNames = new List <string>();
int index = 0;
foreach (var item in firstLineTypes)
{
headerNames.Add(hasHeader ? item : "_col" + index++);
}
// get the list of column types
IReadOnlyList <ColumnType> columnTypes;
if (_parseAsText)
{
columnTypes = firstLineTypes.Select(c => ColumnType.String).ToList();
}
else
{
columnTypes = lines
.Skip(hasHeader ? 1 : 0)
.SelectMany(line => _Parse(line).Select((str, pos) => Tuple.Create(str, pos)))
.GroupBy(l => l.Item2, l => _DetermineType(l.Item1))
.OrderBy(g => g.Key)
.Select(_GetColumnType)
.ToList()
;
}
// add the columns
var ret = new DataTableWriter(stream);
foreach (var column in headerNames.Zip(columnTypes, (name, type) => Tuple.Create(name, type)))
{
ret.AddColumn(column.Item1, column.Item2);
}
return(ret);
}
public IDataTable ConvertToNumeric(IDataTableVectoriser vectoriser = null, Stream output = null)
{
var writer = new DataTableWriter(output);
vectoriser = vectoriser ?? GetVectoriser();
foreach (var name in vectoriser.ColumnNames)
{
writer.AddColumn(name, ColumnType.Float);
}
var classColumnIndex = TargetColumnIndex;
var classColumn = _column[classColumnIndex];
writer.AddColumn(classColumn.Name, ColumnType.String, true);
_Iterate(row => {
var data = vectoriser.GetInput(row).AsEnumerable().Cast <object>().Concat(new object[] { row.GetField <string>(classColumnIndex) });
writer.AddRow(data);
return(true);
});
return(writer.GetDataTable());
}
public IDataTable Project(Func <IRow, IReadOnlyList <object> > mutator, Stream output = null)
{
var isFirst = true;
DataTableWriter writer = new DataTableWriter(output);
_Iterate((row, i) => {
var mutatedRow = mutator(row);
if (mutatedRow != null)
{
if (isFirst)
{
int index = 0;
foreach (var item in mutatedRow)
{
var column = Columns[index];
if (item == null)
{
writer.AddColumn(column.Name, ColumnType.Null, column.IsTarget);
}
else
{
var type = item.GetType();
ColumnType columnType;
if (type == typeof(string))
{
columnType = ColumnType.String;
}
else if (type == typeof(double))
{
columnType = ColumnType.Double;
}
else if (type == typeof(float))
{
columnType = ColumnType.Float;
}
else if (type == typeof(long))
{
columnType = ColumnType.Long;
}
else if (type == typeof(int))
{
columnType = ColumnType.Int;
}
else if (type == typeof(byte))
{
columnType = ColumnType.Byte;
}
else if (type == typeof(DateTime))
{
columnType = ColumnType.Date;
}
else if (type == typeof(bool))
{
columnType = ColumnType.Boolean;
}
else if (type == typeof(FloatVector))
{
columnType = ColumnType.Vector;
}
else if (type == typeof(FloatMatrix))
{
columnType = ColumnType.Matrix;
}
else if (type == typeof(FloatTensor))
{
columnType = ColumnType.Tensor;
}
else if (type == typeof(WeightedIndexList))
{
columnType = ColumnType.WeightedIndexList;
}
else if (type == typeof(IndexList))
{
columnType = ColumnType.IndexList;
}
else
{
throw new FormatException();
}
writer.AddColumn(column.Name, columnType, column.IsTarget);
}
++index;
}
isFirst = false;
}
writer.AddRow(new DataTableRow(this, mutatedRow, _rowConverter));
}
return(true);
});
return(writer.GetDataTable());
}
public IDataTable Project(Func <IRow, IReadOnlyList <object> > mutator, Stream output = null)
{
var isFirst = true;
DataTableWriter writer = new DataTableWriter(output);
_Iterate(row => {
var row2 = mutator(row);
if (row2 != null)
{
if (isFirst)
{
int index = 0;
foreach (var item in row2)
{
var column = Columns[index];
if (item == null)
{
writer.AddColumn(column.Name, ColumnType.Null, column.IsTarget);
}
else
{
var type = item.GetType();
ColumnType columnType;
if (type == typeof(string))
{
columnType = ColumnType.String;
}
else if (type == typeof(double))
{
columnType = ColumnType.Double;
}
else if (type == typeof(float))
{
columnType = ColumnType.Float;
}
else if (type == typeof(long))
{
columnType = ColumnType.Long;
}
else if (type == typeof(int))
{
columnType = ColumnType.Int;
}
else if (type == typeof(byte))
{
columnType = ColumnType.Byte;
}
else if (type == typeof(DateTime))
{
columnType = ColumnType.Date;
}
else if (type == typeof(bool))
{
columnType = ColumnType.Boolean;
}
else
{
throw new FormatException();
}
writer.AddColumn(column.Name, columnType, column.IsTarget);
}
++index;
}
isFirst = false;
}
writer.AddRow(row2);
}
return(true);
});
return(writer.GetDataTable());
}
