Changeset 15518 for trunk/sources/HeuristicLab.DataPreprocessing/3.4/Data

Timestamp:

12/12/17 16:32:35 (7 years ago)

Author:

pfleck

Message:

#2809: merged branch to trunk

Location:

trunk/sources/HeuristicLab.DataPreprocessing

Files:

• . (modified) (1 prop)
• 3.4 (modified) (1 prop)
• 3.4/Data/FilteredPreprocessingData.cs (modified) (6 diffs)
• 3.4/Data/IFilteredPreprocessingData.cs (modified) (2 diffs)
• 3.4/Data/IPreprocessingData.cs (modified) (4 diffs)
• 3.4/Data/ITransactionalPreprocessingData.cs (deleted)
• 3.4/Data/PreprocessingData.cs (modified) (6 diffs)
• 3.4/Data/TransactionalPreprocessingData.cs (deleted)

trunk/sources/HeuristicLab.DataPreprocessing/3.4/Data/FilteredPreprocessingData.cs

@@ -22 +22 @@
 using System;
 using System.Collections.Generic;
+using System.Linq;
 using HeuristicLab.Common;
 using HeuristicLab.Core;
 using HeuristicLab.Data;
+using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
 using HeuristicLab.Problems.DataAnalysis;
 
 namespace HeuristicLab.DataPreprocessing {
-  public class FilteredPreprocessingData : NamedItem, IFilteredPreprocessingData {
-    private readonly ITransactionalPreprocessingData originalData;
-    private ITransactionalPreprocessingData filteredData;
-
+  [Item("FilteredPreprocessingData", "Represents filtered data used for preprocessing.")]
+  [StorableClass]
+  public sealed class FilteredPreprocessingData : NamedItem, IFilteredPreprocessingData {
+
+    [Storable]
+    private readonly IPreprocessingData originalData;
+    [Storable]
+    private IPreprocessingData filteredData;
+
+    public IPreprocessingData ActiveData {
+      get { return IsFiltered ? filteredData : originalData; }
+    }
+
+    #region Constructor, Cloning & Persistence
+    public FilteredPreprocessingData(IPreprocessingData preprocessingData)
+      : base() {
+      originalData = preprocessingData;
+      filteredData = null;
+    }
+
+    private FilteredPreprocessingData(FilteredPreprocessingData original, Cloner cloner)
+      : base(original, cloner) {
+      originalData = original.originalData;
+      filteredData = original.filteredData;
+    }
+    public override IDeepCloneable Clone(Cloner cloner) {
+      return new FilteredPreprocessingData(this, cloner);
+    }
+
+    [StorableConstructor]
+    private FilteredPreprocessingData(bool deserializing)
+      : base(deserializing) { }
+    #endregion
+
+    #region Cells
+    public bool IsCellEmpty(int columnIndex, int rowIndex) {
+      return ActiveData.IsCellEmpty(columnIndex, rowIndex);
+    }
+
+    public T GetCell<T>(int columnIndex, int rowIndex) {
+      return ActiveData.GetCell<T>(columnIndex, rowIndex);
+    }
+
+    public void SetCell<T>(int columnIndex, int rowIndex, T value) {
+      if (IsFiltered)
+        throw new InvalidOperationException("SetValues not possible while data is filtered");
+      originalData.SetCell<T>(columnIndex, rowIndex, value);
+    }
+
+    public string GetCellAsString(int columnIndex, int rowIndex) {
+      return ActiveData.GetCellAsString(columnIndex, rowIndex);
+    }
+
+    public IList<T> GetValues<T>(int columnIndex, bool considerSelection) {
+      return ActiveData.GetValues<T>(columnIndex, considerSelection);
+    }
+
+    public void SetValues<T>(int columnIndex, IList<T> values) {
+      if (IsFiltered)
+        throw new InvalidOperationException("SetValues not possible while data is filtered");
+
+      originalData.SetValues<T>(columnIndex, values);
+    }
+
+    public bool SetValue(string value, int columnIndex, int rowIndex) {
+      if (IsFiltered)
+        throw new InvalidOperationException("SetValue not possible while data is filtered");
+      return originalData.SetValue(value, columnIndex, rowIndex);
+    }
+
+    public int Columns {
+      get { return ActiveData.Columns; }
+    }
+
+    public int Rows {
+      get { return ActiveData.Rows; }
+    }
+    #endregion
+
+    #region Rows
+    public void InsertRow(int rowIndex) {
+      if (IsFiltered)
+        throw new InvalidOperationException("InsertRow not possible while data is filtered");
+
+      originalData.InsertRow(rowIndex);
+    }
+
+    public void DeleteRow(int rowIndex) {
+      if (IsFiltered)
+        throw new InvalidOperationException("DeleteRow not possible while data is filtered");
+
+      originalData.DeleteRow(rowIndex);
+    }
+
+    public void DeleteRowsWithIndices(IEnumerable<int> rows) {
+      if (IsFiltered)
+        throw new InvalidOperationException("DeleteRowsWithIndices not possible while data is filtered");
+
+      originalData.DeleteRowsWithIndices(rows);
+    }
+
+    public void InsertColumn<T>(string variableName, int columnIndex) {
+      if (IsFiltered)
+        throw new InvalidOperationException("InsertColumn not possible while data is filtered");
+
+      originalData.InsertColumn<T>(variableName, columnIndex);
+    }
+
+    public void DeleteColumn(int columnIndex) {
+      if (IsFiltered)
+        throw new InvalidOperationException("DeleteColumn not possible while data is filtered");
+      originalData.DeleteColumn(columnIndex);
+    }
+
+    public void RenameColumn(int columnIndex, string name) {
+      if (IsFiltered)
+        throw new InvalidOperationException("RenameColumn not possible while data is filtered");
+      originalData.RenameColumn(columnIndex, name);
+    }
+
+    public void RenameColumns(IList<string> names) {
+      if (IsFiltered)
+        throw new InvalidOperationException("RenameColumns not possible while data is filtered");
+      originalData.RenameColumns(names);
+    }
+
+    public bool AreAllStringColumns(IEnumerable<int> columnIndices) {
+      return originalData.AreAllStringColumns(columnIndices);
+    }
+    #endregion
+
+    #region Variables
+    public IEnumerable<string> VariableNames {
+      get { return ActiveData.VariableNames; }
+    }
+    public IEnumerable<string> GetDoubleVariableNames() {
+      return originalData.GetDoubleVariableNames();
+    }
+    public string GetVariableName(int columnIndex) {
+      return ActiveData.GetVariableName(columnIndex);
+    }
+
+    public int GetColumnIndex(string variableName) {
+      return ActiveData.GetColumnIndex(variableName);
+    }
+
+    public bool VariableHasType<T>(int columnIndex) {
+      return originalData.VariableHasType<T>(columnIndex);
+    }
+
+    public Type GetVariableType(int columnIndex) {
+      return ActiveData.GetVariableType(columnIndex);
+    }
+
+    public IList<string> InputVariables {
+      get { return ActiveData.InputVariables; }
+    }
+
+    public string TargetVariable {
+      get { return ActiveData.TargetVariable; }
+    } // optional
+    #endregion
+
+    #region Partitions
     public IntRange TrainingPartition {
       get { return originalData.TrainingPartition; }
@@ -39 +201 @@
       get { return originalData.TestPartition; }
     }
-
+    #endregion
+
+    #region Transformations
     public IList<ITransformation> Transformations {
       get { return originalData.Transformations; }
     }
-
-    public IEnumerable<string> VariableNames {
-      get { return ActiveData.VariableNames; }
-    }
-
-    public IList<string> InputVariables { get { return ActiveData.InputVariables; } }
-    public string TargetVariable { get { return ActiveData.TargetVariable; } } // optional
-
+    #endregion
+
+    #region Validation
+    public bool Validate(string value, out string errorMessage, int columnIndex) {
+      return originalData.Validate(value, out errorMessage, columnIndex);
+    }
+    #endregion
+
+    #region Import & Export
+    public void Import(IDataAnalysisProblemData problemData) {
+      if (IsFiltered)
+        throw new InvalidOperationException("Import not possible while data is filtered");
+      originalData.Import(problemData);
+    }
+
+    public Dataset ExportToDataset() {
+      return originalData.ExportToDataset();
+    }
+    #endregion
+
+    #region Selection
     public IDictionary<int, IList<int>> Selection {
       get { return originalData.Selection; }
@@ -56 +233 @@
     }
 
-    public int Columns {
-      get { return ActiveData.Columns; }
-    }
-
-    public int Rows {
-      get { return ActiveData.Rows; }
-    }
-
-    public ITransactionalPreprocessingData ActiveData {
-      get { return IsFiltered ? filteredData : originalData; }
+    public void ClearSelection() {
+      originalData.ClearSelection();
+    }
+
+    public event EventHandler SelectionChanged {
+      add { originalData.SelectionChanged += value; }
+      remove { originalData.SelectionChanged -= value; }
+    }
+    #endregion
+
+    #region Transactions
+    public event DataPreprocessingChangedEventHandler Changed {
+      add { originalData.Changed += value; }
+      remove { originalData.Changed -= value; }
     }
 
@@ -72 +253 @@
     }
 
-    public bool IsFiltered {
-      get { return filteredData != null; }
-    }
-
-
-    public FilteredPreprocessingData(ITransactionalPreprocessingData preporcessingData)
-      : base() {
-      originalData = preporcessingData;
-      filteredData = null;
-    }
-
-    protected FilteredPreprocessingData(FilteredPreprocessingData original, Cloner cloner)
-      : base(original, cloner) {
-      originalData = original.originalData;
-      filteredData = original.filteredData;
-    }
-    public override IDeepCloneable Clone(Cloner cloner) {
-      return new FilteredPreprocessingData(this, cloner);
-    }
-
-    public T GetCell<T>(int columnIndex, int rowIndex) {
-      return ActiveData.GetCell<T>(columnIndex, rowIndex);
-    }
-
-    public void SetCell<T>(int columnIndex, int rowIndex, T value) {
-      if (IsFiltered)
-        throw new InvalidOperationException("SetValues not possible while data is filtered");
-      originalData.SetCell<T>(columnIndex, rowIndex, value);
-    }
-
-    public string GetCellAsString(int columnIndex, int rowIndex) {
-      return ActiveData.GetCellAsString(columnIndex, rowIndex);
-    }
-
-    public IList<T> GetValues<T>(int columnIndex, bool considerSelection) {
-      return ActiveData.GetValues<T>(columnIndex, considerSelection);
-    }
-
-    public void SetValues<T>(int columnIndex, IList<T> values) {
-      if (IsFiltered)
-        throw new InvalidOperationException("SetValues not possible while data is filtered");
-
-      originalData.SetValues<T>(columnIndex, values);
-    }
-
-    public void InsertRow(int rowIndex) {
-      if (IsFiltered)
-        throw new InvalidOperationException("InsertRow not possible while data is filtered");
-
-      originalData.InsertRow(rowIndex);
-    }
-
-    public void DeleteRow(int rowIndex) {
-      if (IsFiltered)
-        throw new InvalidOperationException("DeleteRow not possible while data is filtered");
-
-      originalData.DeleteRow(rowIndex);
-    }
-
-    public void InsertColumn<T>(string variableName, int columnIndex) {
-      if (IsFiltered)
-        throw new InvalidOperationException("InsertColumn not possible while data is filtered");
-
-      originalData.InsertColumn<T>(variableName, columnIndex);
-    }
-
-    public void DeleteColumn(int columnIndex) {
-      if (IsFiltered)
-        throw new InvalidOperationException("DeleteColumn not possible while data is filtered");
-      originalData.DeleteColumn(columnIndex);
-    }
-
-    public void RenameColumn(int columnIndex, string name) {
-      if (IsFiltered)
-        throw new InvalidOperationException("RenameColumn not possible while data is filtered");
-      originalData.RenameColumn(columnIndex, name);
-    }
-
-    public void RenameColumns(IList<string> names) {
-      if (IsFiltered)
-        throw new InvalidOperationException("RenameColumns not possible while data is filtered");
-      originalData.RenameColumns(names);
-    }
-
-    public string GetVariableName(int columnIndex) {
-      return ActiveData.GetVariableName(columnIndex);
-    }
-
-    public int GetColumnIndex(string variableName) {
-      return ActiveData.GetColumnIndex(variableName);
-    }
-
-    public bool VariableHasType<T>(int columnIndex) {
-      return originalData.VariableHasType<T>(columnIndex);
-    }
-
-    public Dataset ExportToDataset() {
-      return originalData.ExportToDataset();
-    }
-
-    public void SetFilter(bool[] rowFilters) {
-      filteredData = (ITransactionalPreprocessingData)originalData.Clone();
+    public void Undo() {
+      if (IsFiltered)
+        throw new InvalidOperationException("Undo not possible while data is filtered");
+
+      originalData.Undo();
+    }
+
+    public void InTransaction(Action action, DataPreprocessingChangedEventType type = DataPreprocessingChangedEventType.Any) {
+      if (IsFiltered)
+        throw new InvalidOperationException("Transaction not possible while data is filtered");
+      originalData.InTransaction(action, type);
+    }
+
+    public void BeginTransaction(DataPreprocessingChangedEventType type) {
+      if (IsFiltered)
+        throw new InvalidOperationException("Transaction not possible while data is filtered");
+      originalData.BeginTransaction(type);
+    }
+
+    public void EndTransaction() {
+      originalData.EndTransaction();
+    }
+    #endregion
+
+    #region Statistics
+    public T GetMin<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      return ActiveData.GetMin<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetMax<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      return ActiveData.GetMax<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetMean<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      return ActiveData.GetMean<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetMedian<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T> {
+      return ActiveData.GetMean<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetMode<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IEquatable<T> {
+      return ActiveData.GetMode<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetStandardDeviation<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      return ActiveData.GetStandardDeviation<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetVariance<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      return ActiveData.GetVariance<T>(columnIndex, considerSelection, emptyValue);
+    }
+    public T GetQuantile<T>(double alpha, int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T> {
+      return ActiveData.GetQuantile<T>(alpha, columnIndex, considerSelection, emptyValue);
+    }
+    public int GetDistinctValues<T>(int columnIndex, bool considerSelection = false) {
+      return ActiveData.GetDistinctValues<T>(columnIndex, considerSelection);
+    }
+
+    public int GetMissingValueCount() {
+      return ActiveData.GetMissingValueCount();
+    }
+    public int GetMissingValueCount(int columnIndex) {
+      return ActiveData.GetMissingValueCount(columnIndex);
+    }
+    public int GetRowMissingValueCount(int rowIndex) {
+      return ActiveData.GetRowMissingValueCount(rowIndex);
+    }
+    #endregion
+
+    #region Filters
+    public void SetFilter(bool[] remainingRows) {
+      filteredData = (IPreprocessingData)originalData.Clone();
       filteredData.InTransaction(() => {
-        for (int row = (rowFilters.Length - 1); row >= 0; --row) {
-          if (rowFilters[row]) {
-            filteredData.DeleteRow(row);
+        var remainingIndices = Enumerable.Range(0, remainingRows.Length).Where(x => remainingRows[x]);
+
+        foreach (var v in filteredData.VariableNames) {
+          var ci = filteredData.GetColumnIndex(v);
+          if (filteredData.VariableHasType<double>(ci)) {
+            var values = filteredData.GetValues<double>(ci);
+            var filteredValues = remainingIndices.Select(x => values[x]).ToList();
+            filteredData.SetValues(ci, filteredValues);
+          } else if (filteredData.VariableHasType<DateTime>(ci)) {
+            var values = filteredData.GetValues<DateTime>(ci);
+            var filteredValues = remainingIndices.Select(x => values[x]).ToList();
+            filteredData.SetValues(ci, filteredValues);
+          } else if (filteredData.VariableHasType<string>(ci)) {
+            var values = filteredData.GetValues<string>(ci);
+            var filteredValues = remainingIndices.Select(x => values[x]).ToList();
+            filteredData.SetValues(ci, filteredValues);
           }
         }
@@ -206 +365 @@
     }
 
+    public bool IsFiltered {
+      get { return filteredData != null; }
+    }
+
+    public event EventHandler FilterChanged;
+
     private void OnFilterChanged() {
       if (FilterChanged != null) {
@@ -211 +376 @@
       }
     }
-
-    public event DataPreprocessingChangedEventHandler Changed {
-      add { originalData.Changed += value; }
-      remove { originalData.Changed -= value; }
-    }
-
-    public bool SetValue(string value, int columnIndex, int rowIndex) {
-      if (IsFiltered)
-        throw new InvalidOperationException("SetValue not possible while data is filtered");
-      return originalData.SetValue(value, columnIndex, rowIndex);
-    }
-
-    public bool AreAllStringColumns(IEnumerable<int> columnIndices) {
-      return originalData.AreAllStringColumns(columnIndices);
-    }
-
-    public void DeleteRowsWithIndices(IEnumerable<int> rows) {
-      if (IsFiltered)
-        throw new InvalidOperationException("DeleteRowsWithIndices not possible while data is filtered");
-
-      originalData.DeleteRowsWithIndices(rows);
-    }
-
-    public void Undo() {
-      if (IsFiltered)
-        throw new InvalidOperationException("Undo not possible while data is filtered");
-
-      originalData.Undo();
-    }
-
-    public void InTransaction(Action action, DataPreprocessingChangedEventType type = DataPreprocessingChangedEventType.Any) {
-      if (IsFiltered)
-        throw new InvalidOperationException("Transaction not possible while data is filtered");
-      originalData.InTransaction(action, type);
-    }
-
-    public void BeginTransaction(DataPreprocessingChangedEventType type) {
-      if (IsFiltered)
-        throw new InvalidOperationException("Transaction not possible while data is filtered");
-      originalData.BeginTransaction(type);
-    }
-
-    public void EndTransaction() {
-      originalData.EndTransaction();
-    }
-
-    public IEnumerable<string> GetDoubleVariableNames() {
-      return originalData.GetDoubleVariableNames();
-    }
-
-    public void ClearSelection() {
-      originalData.ClearSelection();
-    }
-
-    public event EventHandler SelectionChanged {
-      add { originalData.SelectionChanged += value; }
-      remove { originalData.SelectionChanged -= value; }
-    }
-
-    #region IPreprocessingData Members
-    public bool Validate(string value, out string errorMessage, int columnIndex) {
-      return originalData.Validate(value, out errorMessage, columnIndex);
-    }
-
-    public event EventHandler FilterChanged;
     #endregion
   }

trunk/sources/HeuristicLab.DataPreprocessing/3.4/Data/IFilteredPreprocessingData.cs

@@ -23 +23 @@
 
 namespace HeuristicLab.DataPreprocessing {
-  public interface IFilteredPreprocessingData : ITransactionalPreprocessingData {
-    void SetFilter(bool[] rowFilters);
+  public interface IFilteredPreprocessingData : IPreprocessingData {
+    #region Filters
+    void SetFilter(bool[] remainingRows);
     void PersistFilter();
     void ResetFilter();
@@ -30 +31 @@
 
     event EventHandler FilterChanged;
+    #endregion
   }
 }

trunk/sources/HeuristicLab.DataPreprocessing/3.4/Data/IPreprocessingData.cs

@@ -28 +28 @@
 namespace HeuristicLab.DataPreprocessing {
   public interface IPreprocessingData : INamedItem {
+    #region Cells
+    bool IsCellEmpty(int columnIndex, int rowIndex);
     T GetCell<T>(int columnIndex, int rowIndex);
 
@@ -39 +41 @@
     bool SetValue(string value, int columnIndex, int rowIndex);
 
+    int Columns { get; }
+    int Rows { get; }
+    #endregion
+
+    #region Rows
     void InsertRow(int rowIndex);
     void DeleteRow(int rowIndex);
@@ -50 +57 @@
 
     bool AreAllStringColumns(IEnumerable<int> columnIndices);
-    bool Validate(string value, out string errorMessage, int columnIndex);
+    #endregion
 
-    IntRange TrainingPartition { get; }
-    IntRange TestPartition { get; }
-
-    IList<ITransformation> Transformations { get; }
-
+    #region Variables
     IEnumerable<string> VariableNames { get; }
     IEnumerable<string> GetDoubleVariableNames();
@@ -63 +66 @@
 
     bool VariableHasType<T>(int columnIndex);
+    Type GetVariableType(int columnIndex);
 
     IList<string> InputVariables { get; }
     string TargetVariable { get; } // optional
+    #endregion
 
-    int Columns { get; }
-    int Rows { get; }
+    #region Partitions
+    IntRange TrainingPartition { get; }
+    IntRange TestPartition { get; }
+    #endregion
 
+    #region Transformations
+    IList<ITransformation> Transformations { get; }
+    #endregion
+
+    #region Validation
+    bool Validate(string value, out string errorMessage, int columnIndex);
+    #endregion
+
+    #region Import & Export
+    void Import(IDataAnalysisProblemData problemData);
     Dataset ExportToDataset();
+    #endregion
 
+    #region Selection
     IDictionary<int, IList<int>> Selection { get; set; }
     void ClearSelection();
 
     event EventHandler SelectionChanged;
+    #endregion
+
+    #region Transactions
+    event DataPreprocessingChangedEventHandler Changed;
+
+    bool IsUndoAvailable { get; }
+    void Undo();
+    void InTransaction(Action action, DataPreprocessingChangedEventType type = DataPreprocessingChangedEventType.Any);
+    void BeginTransaction(DataPreprocessingChangedEventType type);
+    void EndTransaction();
+    #endregion
+
+    #region Statistics
+    T GetMin<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T));
+    T GetMax<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T));
+    T GetMean<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T));
+    T GetMedian<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T>;
+    T GetMode<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IEquatable<T>;
+    T GetStandardDeviation<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T));
+    T GetVariance<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T));
+    T GetQuantile<T>(double alpha, int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T>;
+    int GetDistinctValues<T>(int columnIndex, bool considerSelection = false);
+
+    int GetMissingValueCount();
+    int GetMissingValueCount(int columnIndex);
+    int GetRowMissingValueCount(int rowIndex);
+    #endregion
   }
 }

trunk/sources/HeuristicLab.DataPreprocessing/3.4/Data/PreprocessingData.cs

@@ -23 +23 @@
 using System.Collections;
 using System.Collections.Generic;
+using System.Globalization;
 using System.Linq;
 using HeuristicLab.Common;
 using HeuristicLab.Core;
 using HeuristicLab.Data;
+using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
 using HeuristicLab.Problems.DataAnalysis;
 
@@ -32 +34 @@
 
   [Item("PreprocessingData", "Represents data used for preprocessing.")]
-  public abstract class PreprocessingData : NamedItem, IPreprocessingData {
-    public IntRange TrainingPartition { get; set; }
-    public IntRange TestPartition { get; set; }
-
-    public IList<ITransformation> Transformations { get; protected set; }
-
+  [StorableClass]
+  public class PreprocessingData : NamedItem, IPreprocessingData {
+
+    [Storable]
     protected IList<IList> variableValues;
+    [Storable]
     protected IList<string> variableNames;
 
-    public IEnumerable<string> VariableNames {
-      get { return variableNames; }
-    }
-
-    public IEnumerable<string> GetDoubleVariableNames() {
-      var doubleVariableNames = new List<string>();
-      for (int i = 0; i < Columns; ++i) {
-        if (VariableHasType<double>(i)) {
-          doubleVariableNames.Add(variableNames[i]);
-        }
-      }
-      return doubleVariableNames;
-    }
-
-    public IList<string> InputVariables { get; private set; }
-    public string TargetVariable { get; private set; } // optional
-
-    public int Columns {
-      get { return variableNames.Count; }
-    }
-
-    public int Rows {
-      get { return variableValues.Count > 0 ? variableValues[0].Count : 0; }
-    }
-
-    protected IDictionary<int, IList<int>> selection;
-    public IDictionary<int, IList<int>> Selection {
-      get { return selection; }
-      set {
-        selection = value;
-        OnSelectionChanged();
-      }
+    #region Constructor, Cloning & Persistence
+    public PreprocessingData(IDataAnalysisProblemData problemData)
+      : base() {
+      Name = "Preprocessing Data";
+
+      Transformations = new List<ITransformation>();
+      selection = new Dictionary<int, IList<int>>();
+
+      Import(problemData);
+
+      RegisterEventHandler();
     }
 
@@ -88 +68 @@
       RegisterEventHandler();
     }
-
-    protected PreprocessingData(IDataAnalysisProblemData problemData)
-      : base() {
-      Name = "Preprocessing Data";
-
-      Transformations = new List<ITransformation>();
-      selection = new Dictionary<int, IList<int>>();
-
-      Import(problemData);
-
+    public override IDeepCloneable Clone(Cloner cloner) {
+      return new PreprocessingData(this, cloner);
+    }
+
+    [StorableConstructor]
+    protected PreprocessingData(bool deserializing)
+      : base(deserializing) { }
+    [StorableHook(HookType.AfterDeserialization)]
+    private void AfterDeserialization() {
       RegisterEventHandler();
     }
 
+    private void RegisterEventHandler() {
+      Changed += (s, e) => {
+        switch (e.Type) {
+          case DataPreprocessingChangedEventType.DeleteRow:
+          case DataPreprocessingChangedEventType.Any:
+          case DataPreprocessingChangedEventType.Transformation:
+            int maxRowIndex = Math.Max(0, Rows);
+            TrainingPartition.Start = Math.Min(TrainingPartition.Start, maxRowIndex);
+            TrainingPartition.End = Math.Min(TrainingPartition.End, maxRowIndex);
+            TestPartition.Start = Math.Min(TestPartition.Start, maxRowIndex);
+            TestPartition.End = Math.Min(TestPartition.End, maxRowIndex);
+            break;
+        }
+      };
+    }
+    #endregion
+
+    #region Cells
+    public bool IsCellEmpty(int columnIndex, int rowIndex) {
+      var value = variableValues[columnIndex][rowIndex];
+      return IsMissingValue(value);
+    }
+
+    public T GetCell<T>(int columnIndex, int rowIndex) {
+      return (T)variableValues[columnIndex][rowIndex];
+    }
+
+    public void SetCell<T>(int columnIndex, int rowIndex, T value) {
+      SaveSnapshot(DataPreprocessingChangedEventType.ChangeItem, columnIndex, rowIndex);
+
+      for (int i = Rows; i <= rowIndex; i++)
+        InsertRow(i);
+      for (int i = Columns; i <= columnIndex; i++)
+        InsertColumn<T>(i.ToString(), i);
+
+      variableValues[columnIndex][rowIndex] = value;
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.ChangeItem, columnIndex, rowIndex);
+    }
+
+    public string GetCellAsString(int columnIndex, int rowIndex) {
+      return variableValues[columnIndex][rowIndex].ToString();
+    }
+
+    public IList<T> GetValues<T>(int columnIndex, bool considerSelection) {
+      if (considerSelection) {
+        var list = new List<T>();
+        foreach (var rowIdx in selection[columnIndex]) {
+          list.Add((T)variableValues[columnIndex][rowIdx]);
+        }
+        return list;
+      } else {
+        return (IList<T>)variableValues[columnIndex];
+      }
+    }
+
+    public void SetValues<T>(int columnIndex, IList<T> values) {
+      SaveSnapshot(DataPreprocessingChangedEventType.ChangeColumn, columnIndex, -1);
+      if (VariableHasType<T>(columnIndex)) {
+        variableValues[columnIndex] = (IList)values;
+      } else {
+        throw new ArgumentException("The datatype of column " + columnIndex + " must be of type " + variableValues[columnIndex].GetType().Name + " but was " + typeof(T).Name);
+      }
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.ChangeColumn, columnIndex, -1);
+    }
+
+    public bool SetValue(string value, int columnIndex, int rowIndex) {
+      bool valid = false;
+      if (VariableHasType<double>(columnIndex)) {
+        double val;
+        if (string.IsNullOrWhiteSpace(value)) {
+          val = double.NaN;
+          valid = true;
+        } else {
+          valid = double.TryParse(value, out val);
+        }
+        if (valid)
+          SetCell(columnIndex, rowIndex, val);
+      } else if (VariableHasType<string>(columnIndex)) {
+        valid = value != null;
+        if (valid)
+          SetCell(columnIndex, rowIndex, value);
+      } else if (VariableHasType<DateTime>(columnIndex)) {
+        DateTime date;
+        valid = DateTime.TryParse(value, out date);
+        if (valid)
+          SetCell(columnIndex, rowIndex, date);
+      } else {
+        throw new ArgumentException("column " + columnIndex + " contains a non supported type.");
+      }
+
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.ChangeColumn, columnIndex, -1);
+
+      return valid;
+    }
+
+    public int Columns {
+      get { return variableNames.Count; }
+    }
+
+    public int Rows {
+      get { return variableValues.Count > 0 ? variableValues[0].Count : 0; }
+    }
+
+    public static bool IsMissingValue(object value) {
+      if (value is double) return double.IsNaN((double)value);
+      if (value is string) return string.IsNullOrEmpty((string)value);
+      if (value is DateTime) return ((DateTime)value).Equals(DateTime.MinValue);
+      throw new ArgumentException();
+    }
+    #endregion
+
+    #region Rows
+    public void InsertRow(int rowIndex) {
+      SaveSnapshot(DataPreprocessingChangedEventType.DeleteRow, -1, rowIndex);
+      foreach (IList column in variableValues) {
+        Type type = column.GetType().GetGenericArguments()[0];
+        column.Insert(rowIndex, type.IsValueType ? Activator.CreateInstance(type) : null);
+      }
+      if (TrainingPartition.Start <= rowIndex && rowIndex <= TrainingPartition.End) {
+        TrainingPartition.End++;
+        if (TrainingPartition.End <= TestPartition.Start) {
+          TestPartition.Start++;
+          TestPartition.End++;
+        }
+      } else if (TestPartition.Start <= rowIndex && rowIndex <= TestPartition.End) {
+        TestPartition.End++;
+        if (TestPartition.End <= TrainingPartition.Start) {
+          TestPartition.Start++;
+          TestPartition.End++;
+        }
+      }
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.AddRow, -1, rowIndex);
+    }
+    public void DeleteRow(int rowIndex) {
+      SaveSnapshot(DataPreprocessingChangedEventType.AddRow, -1, rowIndex);
+      foreach (IList column in variableValues) {
+        column.RemoveAt(rowIndex);
+      }
+      if (TrainingPartition.Start <= rowIndex && rowIndex <= TrainingPartition.End) {
+        TrainingPartition.End--;
+        if (TrainingPartition.End <= TestPartition.Start) {
+          TestPartition.Start--;
+          TestPartition.End--;
+        }
+      } else if (TestPartition.Start <= rowIndex && rowIndex <= TestPartition.End) {
+        TestPartition.End--;
+        if (TestPartition.End <= TrainingPartition.Start) {
+          TestPartition.Start--;
+          TestPartition.End--;
+        }
+      }
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.DeleteRow, -1, rowIndex);
+    }
+    public void DeleteRowsWithIndices(IEnumerable<int> rows) {
+      SaveSnapshot(DataPreprocessingChangedEventType.AddRow, -1, -1);
+      foreach (int rowIndex in rows.OrderByDescending(x => x)) {
+        foreach (IList column in variableValues) {
+          column.RemoveAt(rowIndex);
+        }
+        if (TrainingPartition.Start <= rowIndex && rowIndex <= TrainingPartition.End) {
+          TrainingPartition.End--;
+          if (TrainingPartition.End <= TestPartition.Start) {
+            TestPartition.Start--;
+            TestPartition.End--;
+          }
+        } else if (TestPartition.Start <= rowIndex && rowIndex <= TestPartition.End) {
+          TestPartition.End--;
+          if (TestPartition.End <= TrainingPartition.Start) {
+            TestPartition.Start--;
+            TestPartition.End--;
+          }
+        }
+      }
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.DeleteRow, -1, -1);
+    }
+
+    public void InsertColumn<T>(string variableName, int columnIndex) {
+      SaveSnapshot(DataPreprocessingChangedEventType.DeleteColumn, columnIndex, -1);
+      variableValues.Insert(columnIndex, new List<T>(Enumerable.Repeat(default(T), Rows)));
+      variableNames.Insert(columnIndex, variableName);
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.AddColumn, columnIndex, -1);
+    }
+
+    public void DeleteColumn(int columnIndex) {
+      SaveSnapshot(DataPreprocessingChangedEventType.AddColumn, columnIndex, -1);
+      variableValues.RemoveAt(columnIndex);
+      variableNames.RemoveAt(columnIndex);
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.DeleteColumn, columnIndex, -1);
+    }
+
+    public void RenameColumn(int columnIndex, string name) {
+      SaveSnapshot(DataPreprocessingChangedEventType.ChangeColumn, columnIndex, -1);
+      if (columnIndex < 0 || columnIndex > variableNames.Count)
+        throw new ArgumentOutOfRangeException("columnIndex");
+      variableNames[columnIndex] = name;
+
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.ChangeColumn, -1, -1);
+    }
+
+    public void RenameColumns(IList<string> names) {
+      if (names == null) throw new ArgumentNullException("names");
+      if (names.Count != variableNames.Count) throw new ArgumentException("number of names must match the number of columns.", "names");
+
+      SaveSnapshot(DataPreprocessingChangedEventType.ChangeColumn, -1, -1);
+      for (int i = 0; i < names.Count; i++)
+        variableNames[i] = names[i];
+
+      if (!IsInTransaction)
+        OnChanged(DataPreprocessingChangedEventType.ChangeColumn, -1, -1);
+    }
+
+    public bool AreAllStringColumns(IEnumerable<int> columnIndices) {
+      return columnIndices.All(x => VariableHasType<string>(x));
+    }
+    #endregion
+
+    #region Variables
+    public IEnumerable<string> VariableNames {
+      get { return variableNames; }
+    }
+
+    public IEnumerable<string> GetDoubleVariableNames() {
+      var doubleVariableNames = new List<string>();
+      for (int i = 0; i < Columns; ++i) {
+        if (VariableHasType<double>(i)) {
+          doubleVariableNames.Add(variableNames[i]);
+        }
+      }
+      return doubleVariableNames;
+    }
+
+    public string GetVariableName(int columnIndex) {
+      return variableNames[columnIndex];
+    }
+
+    public int GetColumnIndex(string variableName) {
+      return variableNames.IndexOf(variableName);
+    }
+
+    public bool VariableHasType<T>(int columnIndex) {
+      return columnIndex >= variableValues.Count || variableValues[columnIndex] is List<T>;
+    }
+
+    public Type GetVariableType(int columnIndex) {
+      var listType = variableValues[columnIndex].GetType();
+      return listType.GenericTypeArguments.Single();
+    }
+
+    public IList<string> InputVariables { get; private set; }
+    public string TargetVariable { get; private set; } // optional
+    #endregion
+
+    #region Partitions
+    [Storable]
+    public IntRange TrainingPartition { get; set; }
+    [Storable]
+    public IntRange TestPartition { get; set; }
+    #endregion
+
+    #region Transformations
+    [Storable]
+    public IList<ITransformation> Transformations { get; protected set; }
+    #endregion
+
+    #region Validation
+    public bool Validate(string value, out string errorMessage, int columnIndex) {
+      if (columnIndex < 0 || columnIndex > VariableNames.Count()) {
+        throw new ArgumentOutOfRangeException("column index is out of range");
+      }
+
+      bool valid = false;
+      errorMessage = string.Empty;
+      if (VariableHasType<double>(columnIndex)) {
+        if (string.IsNullOrWhiteSpace(value)) {
+          valid = true;
+        } else {
+          double val;
+          valid = double.TryParse(value, out val);
+          if (!valid) {
+            errorMessage = "Invalid Value (Valid Value Format: \"" + FormatPatterns.GetDoubleFormatPattern() + "\")";
+          }
+        }
+      } else if (VariableHasType<string>(columnIndex)) {
+        valid = value != null;
+        if (!valid) {
+          errorMessage = "Invalid Value (string must not be null)";
+        }
+      } else if (VariableHasType<DateTime>(columnIndex)) {
+        DateTime date;
+        valid = DateTime.TryParse(value, out date);
+        if (!valid) {
+          errorMessage = "Invalid Value (Valid Value Format: \"" + CultureInfo.CurrentCulture.DateTimeFormat + "\"";
+        }
+      } else {
+        throw new ArgumentException("column " + columnIndex + " contains a non supported type.");
+      }
+
+      return valid;
+    }
+    #endregion
+
+    #region Import & Export
     public void Import(IDataAnalysisProblemData problemData) {
       Dataset dataset = (Dataset)problemData.Dataset;
@@ -107 +397 @@
       TargetVariable = (problemData is IRegressionProblemData) ? ((IRegressionProblemData)problemData).TargetVariable
         : (problemData is IClassificationProblemData) ? ((IClassificationProblemData)problemData).TargetVariable
-        : null;
+          : null;
 
       int columnIndex = 0;
@@ -128 +418 @@
     }
 
-    private void RegisterEventHandler() {
-      Changed += (s, e) => {
-        switch (e.Type) {
-          case DataPreprocessingChangedEventType.DeleteRow:
-            CheckPartitionRanges();
-            break;
-          case DataPreprocessingChangedEventType.Any:
-            CheckPartitionRanges();
-            break;
-          case DataPreprocessingChangedEventType.Transformation:
-            CheckPartitionRanges();
-            break;
-        }
-      };
-    }
-
-    private void CheckPartitionRanges() {
-      int maxRowIndex = Math.Max(0, Rows);
-      TrainingPartition.Start = Math.Min(TrainingPartition.Start, maxRowIndex);
-      TrainingPartition.End = Math.Min(TrainingPartition.End, maxRowIndex);
-      TestPartition.Start = Math.Min(TestPartition.Start, maxRowIndex);
-      TestPartition.End = Math.Min(TestPartition.End, maxRowIndex);
-    }
-
-    protected IList<IList> CopyVariableValues(IList<IList> original) {
-      var copy = new List<IList>(original);
-      for (int i = 0; i < original.Count; ++i) {
-        copy[i] = (IList)Activator.CreateInstance(original[i].GetType(), original[i]);
-      }
-      return copy;
-    }
-
-
-    #region IPreprocessingData Members
-    public abstract T GetCell<T>(int columnIndex, int rowIndex);
-
-    public abstract void SetCell<T>(int columnIndex, int rowIndex, T value);
-
-    public abstract string GetCellAsString(int columnIndex, int rowIndex);
-
-    public abstract string GetVariableName(int columnIndex);
-
-    public abstract int GetColumnIndex(string variableName);
-
-    public abstract bool VariableHasType<T>(int columnIndex);
-
-    [Obsolete("use the index based variant, is faster")]
-    public abstract IList<T> GetValues<T>(string variableName, bool considerSelection);
-
-    public abstract IList<T> GetValues<T>(int columnIndex, bool considerSelection);
-
-    public abstract void SetValues<T>(int columnIndex, IList<T> values);
-
-    public abstract bool SetValue(string value, int columnIndex, int rowIndex);
-
-    public abstract bool Validate(string value, out string errorMessage, int columnIndex);
-
-    public abstract bool AreAllStringColumns(IEnumerable<int> columnIndices);
-
-    public abstract void DeleteRowsWithIndices(IEnumerable<int> rows);
-
-    public abstract void InsertRow(int rowIndex);
-
-    public abstract void DeleteRow(int rowIndex);
-
-    public abstract void InsertColumn<T>(string variableName, int columnIndex);
-
-    public abstract void DeleteColumn(int columnIndex);
-
-    public abstract void RenameColumn(int columnIndex, string name);
-    public abstract void RenameColumns(IList<string> list);
-
-    public abstract Dataset ExportToDataset();
-
-    public abstract void ClearSelection();
-
-    public abstract event EventHandler SelectionChanged;
-    protected abstract void OnSelectionChanged();
+    public Dataset ExportToDataset() {
+      IList<IList> values = new List<IList>();
+
+      for (int i = 0; i < Columns; ++i) {
+        values.Add(variableValues[i]);
+      }
+
+      var dataset = new Dataset(variableNames, values);
+      return dataset;
+    }
+    #endregion
+
+    #region Selection
+    [Storable]
+    protected IDictionary<int, IList<int>> selection;
+    public IDictionary<int, IList<int>> Selection {
+      get { return selection; }
+      set {
+        selection = value;
+        OnSelectionChanged();
+      }
+    }
+    public void ClearSelection() {
+      Selection = new Dictionary<int, IList<int>>();
+    }
+
+    public event EventHandler SelectionChanged;
+    protected void OnSelectionChanged() {
+      var listeners = SelectionChanged;
+      if (listeners != null) listeners(this, EventArgs.Empty);
+    }
+    #endregion
+
+    #region Transactions
+    // Stapshot/History are nost storable/cloneable on purpose
+    private class Snapshot {
+      public IList<IList> VariableValues { get; set; }
+      public IList<string> VariableNames { get; set; }
+
+      public IntRange TrainingPartition { get; set; }
+      public IntRange TestPartition { get; set; }
+      public IList<ITransformation> Transformations { get; set; }
+      public DataPreprocessingChangedEventType ChangedType { get; set; }
+
+      public int ChangedColumn { get; set; }
+      public int ChangedRow { get; set; }
+    }
 
     public event DataPreprocessingChangedEventHandler Changed;
@@ -212 +471 @@
       if (listeners != null) listeners(this, new DataPreprocessingChangedEventArgs(type, column, row));
     }
-    #endregion
+
+    private const int MAX_UNDO_DEPTH = 5;
+
+    private readonly IList<Snapshot> undoHistory = new List<Snapshot>();
+    private readonly Stack<DataPreprocessingChangedEventType> eventStack = new Stack<DataPreprocessingChangedEventType>();
+
+    public bool IsInTransaction { get { return eventStack.Count > 0; } }
+
+    private void SaveSnapshot(DataPreprocessingChangedEventType changedType, int column, int row) {
+      if (IsInTransaction) return;
+
+      var currentSnapshot = new Snapshot {
+        VariableValues = CopyVariableValues(variableValues),
+        VariableNames = new List<string>(variableNames),
+        TrainingPartition = new IntRange(TrainingPartition.Start, TrainingPartition.End),
+        TestPartition = new IntRange(TestPartition.Start, TestPartition.End),
+        Transformations = new List<ITransformation>(Transformations),
+        ChangedType = changedType,
+        ChangedColumn = column,
+        ChangedRow = row
+      };
+
+      if (undoHistory.Count >= MAX_UNDO_DEPTH)
+        undoHistory.RemoveAt(0);
+
+      undoHistory.Add(currentSnapshot);
+    }
+
+    public bool IsUndoAvailable {
+      get { return undoHistory.Count > 0; }
+    }
+
+    public void Undo() {
+      if (IsUndoAvailable) {
+        Snapshot previousSnapshot = undoHistory[undoHistory.Count - 1];
+        variableValues = previousSnapshot.VariableValues;
+        variableNames = previousSnapshot.VariableNames;
+        TrainingPartition = previousSnapshot.TrainingPartition;
+        TestPartition = previousSnapshot.TestPartition;
+        Transformations = previousSnapshot.Transformations;
+        undoHistory.Remove(previousSnapshot);
+        OnChanged(previousSnapshot.ChangedType,
+          previousSnapshot.ChangedColumn,
+          previousSnapshot.ChangedRow);
+      }
+    }
+
+    public void InTransaction(Action action, DataPreprocessingChangedEventType type = DataPreprocessingChangedEventType.Any) {
+      BeginTransaction(type);
+      action();
+      EndTransaction();
+    }
+
+    public void BeginTransaction(DataPreprocessingChangedEventType type) {
+      SaveSnapshot(type, -1, -1);
+      eventStack.Push(type);
+    }
+
+    public void EndTransaction() {
+      if (eventStack.Count == 0)
+        throw new InvalidOperationException("There is no open transaction that can be ended.");
+
+      var @event = eventStack.Pop();
+      OnChanged(@event, -1, -1);
+    }
+    #endregion
+
+    #region Statistics
+    public T GetMin<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.Any() ? values.Min() : emptyValue;
+    }
+
+    public T GetMax<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.Any() ? values.Max() : emptyValue;
+    }
+
+    public T GetMean<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      if (typeof(T) == typeof(double)) {
+        var values = GetValuesWithoutMissingValues<double>(columnIndex, considerSelection);
+        return values.Any() ? Convert<T>(values.Average()) : emptyValue;
+      }
+      if (typeof(T) == typeof(string)) {
+        return Convert<T>(string.Empty);
+      }
+      if (typeof(T) == typeof(DateTime)) {
+        var values = GetValuesWithoutMissingValues<DateTime>(columnIndex, considerSelection);
+        return values.Any() ? Convert<T>(AggregateAsDouble(values, Enumerable.Average)) : emptyValue;
+      }
+
+      throw new InvalidOperationException(typeof(T) + " not supported");
+    }
+
+    public T GetMedian<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T> {
+      if (typeof(T) == typeof(double)) {// IEnumerable<double> is faster  
+        var doubleValues = GetValuesWithoutMissingValues<double>(columnIndex, considerSelection);
+        return doubleValues.Any() ? Convert<T>(doubleValues.Median()) : emptyValue;
+      }
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.Any() ? values.Quantile(0.5) : emptyValue;
+    }
+
+    public T GetMode<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IEquatable<T> {
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.Any() ? values.GroupBy(x => x).OrderByDescending(g => g.Count()).Select(g => g.Key).First() : emptyValue;
+    }
+
+    public T GetStandardDeviation<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      if (typeof(T) == typeof(double)) {
+        var values = GetValuesWithoutMissingValues<double>(columnIndex, considerSelection);
+        return values.Any() ? Convert<T>(values.StandardDeviation()) : emptyValue;
+      }
+      // For DateTime, std.dev / variance would have to be TimeSpan
+      //if (typeof(T) == typeof(DateTime)) {
+      //  var values = GetValuesWithoutMissingValues<DateTime>(columnIndex, considerSelection);
+      //  return values.Any() ? Convert<T>(AggregateAsDouble(values, EnumerableStatisticExtensions.StandardDeviation)) : emptyValue;
+      //}
+      return default(T);
+    }
+
+    public T GetVariance<T>(int columnIndex, bool considerSelection = false, T emptyValue = default(T)) {
+      if (typeof(T) == typeof(double)) {
+        var values = GetValuesWithoutMissingValues<double>(columnIndex, considerSelection);
+        return values.Any() ? Convert<T>(values.Variance()) : emptyValue;
+      }
+      // DateTime variance often overflows long, thus the corresponding DateTime is invalid
+      //if (typeof(T) == typeof(DateTime)) {
+      //  var values = GetValuesWithoutMissingValues<DateTime>(columnIndex, considerSelection);
+      //  return values.Any() ? Convert<T>(AggregateAsDouble(values, EnumerableStatisticExtensions.Variance)) : emptyValue;
+      //}
+      return default(T);
+    }
+
+    public T GetQuantile<T>(double alpha, int columnIndex, bool considerSelection = false, T emptyValue = default(T)) where T : IComparable<T> {
+      if (typeof(T) == typeof(double)) {// IEnumerable<double> is faster  
+        var doubleValues = GetValuesWithoutMissingValues<double>(columnIndex, considerSelection);
+        return doubleValues.Any() ? Convert<T>(doubleValues.Quantile(alpha)) : emptyValue;
+      }
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.Any() ? values.Quantile(alpha) : emptyValue;
+    }
+
+    public int GetDistinctValues<T>(int columnIndex, bool considerSelection = false) {
+      var values = GetValuesWithoutMissingValues<T>(columnIndex, considerSelection);
+      return values.GroupBy(x => x).Count();
+    }
+
+    private IEnumerable<T> GetValuesWithoutMissingValues<T>(int columnIndex, bool considerSelection) {
+      return GetValues<T>(columnIndex, considerSelection).Where(x => !IsMissingValue(x));
+    }
+
+    private static DateTime AggregateAsDouble(IEnumerable<DateTime> values, Func<IEnumerable<double>, double> func) {
+      return new DateTime((long)(func(values.Select(x => (double)x.Ticks / TimeSpan.TicksPerSecond)) * TimeSpan.TicksPerSecond));
+    }
+    private static T Convert<T>(object obj) { return (T)obj; }
+
+    public int GetMissingValueCount() {
+      int count = 0;
+      for (int i = 0; i < Columns; ++i) {
+        count += GetMissingValueCount(i);
+      }
+      return count;
+    }
+    public int GetMissingValueCount(int columnIndex) {
+      int sum = 0;
+      for (int i = 0; i < Rows; i++) {
+        if (IsCellEmpty(columnIndex, i))
+          sum++;
+      }
+      return sum;
+    }
+    public int GetRowMissingValueCount(int rowIndex) {
+      int sum = 0;
+      for (int i = 0; i < Columns; i++) {
+        if (IsCellEmpty(i, rowIndex))
+          sum++;
+      }
+      return sum;
+    }
+    #endregion
+
+    #region Helpers
+    private static IList<IList> CopyVariableValues(IList<IList> original) {
+      var copy = new List<IList>(original);
+      for (int i = 0; i < original.Count; ++i) {
+        copy[i] = (IList)Activator.CreateInstance(original[i].GetType(), original[i]);
+      }
+      return copy;
+    }
+    #endregion
+  }
+
+  // Adapted from HeuristicLab.Common.EnumerableStatisticExtensions
+  internal static class EnumerableExtensions {
+    public static T Quantile<T>(this IEnumerable<T> values, double alpha) where T : IComparable<T> {
+      T[] valuesArr = values.ToArray();
+      int n = valuesArr.Length;
+      if (n == 0) throw new InvalidOperationException("Enumeration contains no elements.");
+
+      var pos = n * alpha;
+
+      return Select((int)Math.Ceiling(pos) - 1, valuesArr);
+
+    }
+
+    private static T Select<T>(int k, T[] arr) where T : IComparable<T> {
+      int i, ir, j, l, mid, n = arr.Length;
+      T a;
+      l = 0;
+      ir = n - 1;
+      for (;;) {
+        if (ir <= l + 1) {
+          // Active partition contains 1 or 2 elements.
+          if (ir == l + 1 && arr[ir].CompareTo(arr[l]) < 0) {
+            // Case of 2 elements.
+            Swap(arr, l, ir);
+          }
+          return arr[k];
+        } else {
+          mid = (l + ir) >> 1; // Choose median of left, center, and right elements
+          Swap(arr, mid, l + 1); // as partitioning element a. Also
+
+          if (arr[l].CompareTo(arr[ir]) > 0) {  // rearrange so that arr[l] arr[ir] <= arr[l+1],
+            Swap(arr, l, ir); // . arr[ir] >= arr[l+1]
+          }
+
+          if (arr[l + 1].CompareTo(arr[ir]) > 0) {
+            Swap(arr, l + 1, ir);
+          }
+          if (arr[l].CompareTo(arr[l + 1]) > 0) {
+            Swap(arr, l, l + 1);
+          }
+          i = l + 1; // Initialize pointers for partitioning.
+          j = ir;
+          a = arr[l + 1]; // Partitioning element.
+          for (;;) { // Beginning of innermost loop.
+            do i++; while (arr[i].CompareTo(a) < 0); // Scan up to find element > a.
+            do j--; while (arr[j].CompareTo(a) > 0); // Scan down to find element < a.
+            if (j < i) break; // Pointers crossed. Partitioning complete.
+            Swap(arr, i, j);
+          } // End of innermost loop.
+          arr[l + 1] = arr[j]; // Insert partitioning element.
+          arr[j] = a;
+          if (j >= k) ir = j - 1; // Keep active the partition that contains the
+          if (j <= k) l = i; // kth element.
+        }
+      }
+    }
+
+    private static void Swap<T>(T[] arr, int i, int j) {
+      T temp = arr[i];
+      arr[i] = arr[j];
+      arr[j] = temp;
+    }
   }
 }