#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2010 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
 *
 * This file is part of HeuristicLab.
 *
 * HeuristicLab is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * HeuristicLab is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
 */
#endregion

using System.Collections.Generic;
using System.Linq;
using System.Windows.Forms;
using HeuristicLab.Common;
using HeuristicLab.Data;
using HeuristicLab.MainForm;
using HeuristicLab.MainForm.WindowsForms;
using HeuristicLab.Optimization;
using System;
using HeuristicLab.Problems.DataAnalysis.Regression.Symbolic;
using HeuristicLab.Problems.DataAnalysis.Symbolic.Symbols;
using HeuristicLab.Problems.DataAnalysis.Evaluators;

namespace HeuristicLab.Problems.DataAnalysis.Views {
  [Content(typeof(RunCollection), false)]
  [View("RunCollection Winkler Variable Impact View")]
  public partial class RunCollectionWinklerVariableImpactView : AsynchronousContentView {
    private const string validationBestModelResultName = "Best solution (on validation set)";
    public RunCollectionWinklerVariableImpactView() {
      InitializeComponent();
    }

    public new RunCollection Content {
      get { return (RunCollection)base.Content; }
      set { base.Content = value; }
    }

    protected override void RegisterContentEvents() {
      base.RegisterContentEvents();
      this.Content.ItemsAdded += new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_ItemsAdded);
      this.Content.ItemsRemoved += new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_ItemsRemoved);
      this.Content.CollectionReset += new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_CollectionReset);
    }
    protected override void DeregisterContentEvents() {
      base.RegisterContentEvents();
      this.Content.ItemsAdded -= new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_ItemsAdded);
      this.Content.ItemsRemoved -= new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_ItemsRemoved);
      this.Content.CollectionReset -= new HeuristicLab.Collections.CollectionItemsChangedEventHandler<IRun>(Content_CollectionReset);
    }

    protected override void OnContentChanged() {
      base.OnContentChanged();
      this.UpdateData();
    }
    private void Content_ItemsAdded(object sender, HeuristicLab.Collections.CollectionItemsChangedEventArgs<IRun> e) {
      this.UpdateData();
    }
    private void Content_ItemsRemoved(object sender, HeuristicLab.Collections.CollectionItemsChangedEventArgs<IRun> e) {
      this.UpdateData();
    }
    private void Content_CollectionReset(object sender, HeuristicLab.Collections.CollectionItemsChangedEventArgs<IRun> e) {
      this.UpdateData();
    }

    private void UpdateData() {
      matrixView.Content = CalculateVariableImpactMatrix();
    }

    public DoubleMatrix CalculateVariableImpactMatrix() {
      DoubleMatrix matrix = null;
      if (Content != null) {
        List<IRun> runsWithSolutions = (from run in Content
                                        where run.Results.ContainsKey(validationBestModelResultName)
                                        select run)
                                              .ToList();
        IEnumerable<SymbolicRegressionSolution> allSolutions = (from run in Content
                                                                where run.Results.ContainsKey(validationBestModelResultName)
                                                                select run.Results[validationBestModelResultName]).Cast<SymbolicRegressionSolution>();

        Dictionary<SymbolicRegressionSolution, IEnumerable<string>> variableReferences = new Dictionary<SymbolicRegressionSolution, IEnumerable<string>>();
        foreach (var solution in allSolutions) {
          variableReferences[solution] = GetVariableReferences(solution).Distinct();
        }

        List<string> variableNames = (from modelVarRefs in variableReferences.Values
                                      from variableName in modelVarRefs
                                      select variableName)
                                     .Distinct()
                                     .ToList();

        List<string> statictics = new List<string> { "Median Impact", "Mean Impact", "StdDev", "pValue" };
        List<string> columnNames = (from run in runsWithSolutions
                                    select run.Name).ToList();
        columnNames.AddRange(statictics);

        matrix = new DoubleMatrix(variableNames.Count, columnNames.Count);
        matrix.SortableView = true;
        matrix.RowNames = variableNames;
        matrix.ColumnNames = columnNames;

        int columnIndex = 0;
        foreach (SymbolicRegressionSolution solution in variableReferences.Keys) {
          foreach (string variableName in variableReferences[solution]) {
            int rowIndex = variableNames.IndexOf(variableName);
            if (rowIndex > -1) {
              matrix[rowIndex, columnIndex] = CalculateMeanImpact(variableName, solution);
            }
          }
          columnIndex++;
        }
        List<List<double>> variableImpactValues = (from row in Enumerable.Range(0, variableNames.Count())
                                                   select GetRowValues(matrix, row).ToList())
                                                         .ToList();
        List<double> referenceValues = (from variableImpacts in variableImpactValues
                                        orderby variableImpacts.Average()
                                        select variableImpacts)
                                       .First();
        for (int row = 0; row < variableNames.Count; row++) {
          List<double> rowValues = variableImpactValues[row];
          matrix[row, columnIndex] = rowValues.Median();
          matrix[row, columnIndex + 1] = rowValues.Average();
          matrix[row, columnIndex + 2] = rowValues.StandardDeviation();

          double bothTails, leftTail, rightTail;
          bothTails = leftTail = rightTail = 0.0;
          double[] z = new double[rowValues.Count()];
          for (int i = 0; i < z.Length; i++) {
            z[i] = rowValues[i] - referenceValues[i];
          }
          alglib.wilcoxonsignedranktest(z, z.Length, 0.0, out bothTails, out leftTail, out rightTail);
          matrix[row, columnIndex + 3] = bothTails;
        }
      }
      return matrix;
    }

    private IEnumerable<double> GetRowValues(DoubleMatrix matrix, int row) {
      return from col in Enumerable.Range(0, matrix.Columns)
             select matrix[row, col];
    }

    private IEnumerable<string> GetVariableReferences(SymbolicRegressionSolution solution) {
      return from node in solution.Model.SymbolicExpressionTree.IterateNodesPostfix().OfType<VariableTreeNode>()
             select node.VariableName;
    }

    private double CalculateMeanImpact(string variableName, SymbolicRegressionSolution solution) {
      int variableIndex = solution.ProblemData.Dataset.GetVariableIndex(variableName);
      double meanVal = solution.ProblemData.Dataset.GetVariableValues(variableName).Average();
      List<double> originalOutput = new List<double>(solution.EstimatedValues);

      int rows = solution.ProblemData.Dataset.Rows;
      int columns = solution.ProblemData.Dataset.Columns;

      double[,] manipulatedData = new double[rows, columns];
      for (int row = 0; row < rows; row++) {
        for (int column = 0; column < columns; column++) {
          if (column != variableIndex) {
            manipulatedData[row, column] = solution.ProblemData.Dataset[row, column];
          } else {
            manipulatedData[row, column] = meanVal;
          }
        }
      }

      Dataset originalDataset = solution.ProblemData.Dataset;
      Dataset manipulatedDataset = new Dataset(solution.ProblemData.Dataset.VariableNames, manipulatedData);
      solution.ProblemData.Dataset = manipulatedDataset;
      List<double> newOuput = new List<double>(solution.EstimatedValues);
      solution.ProblemData.Dataset = originalDataset;

      return SimpleMSEEvaluator.Calculate(originalOutput, newOuput);
    }
  }
}