/*=========================================================================

  Program:   Visualization Toolkit
  Module:    vtkStatisticsAlgorithm.cxx

  Copyright (c) Ken Martin, Will Schroeder, Bill Lorensen
  All rights reserved.
  See Copyright.txt or http://www.kitware.com/Copyright.htm for details.

     This software is distributed WITHOUT ANY WARRANTY; without even
     the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
     PURPOSE.  See the above copyright notice for more information.

=========================================================================*/
/*-------------------------------------------------------------------------
  Copyright 2011 Sandia Corporation.
  Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
  the U.S. Government retains certain rights in this software.
-------------------------------------------------------------------------*/

#include "vtkToolkits.h"

#include "vtkStatisticsAlgorithm.h"

#include "vtkDataObjectCollection.h"
#include "vtkDoubleArray.h"
#include "vtkInformation.h"
#include "vtkInformationVector.h"
#include "vtkObjectFactory.h"
#include "vtkMultiBlockDataSet.h"
#include "vtkSmartPointer.h"
#include "vtkStatisticsAlgorithmPrivate.h"
#include "vtkStringArray.h"
#include "vtkTable.h"
#include "vtkVariantArray.h"

#include <sstream>

vtkCxxSetObjectMacro(vtkStatisticsAlgorithm,AssessNames,vtkStringArray);

// ----------------------------------------------------------------------
vtkStatisticsAlgorithm::vtkStatisticsAlgorithm()
{
  this->SetNumberOfInputPorts( 3 );
  this->SetNumberOfOutputPorts( 3 );

  // If not told otherwise, only run Learn option
  this->LearnOption = true;
  this->DeriveOption = true;
  this->AssessOption = false;
  this->TestOption = false;
  // Most engines have only 1 primary table.
  this->NumberOfPrimaryTables = 1;
  this->AssessNames = vtkStringArray::New();
  this->Internals = new vtkStatisticsAlgorithmPrivate;
}

// ----------------------------------------------------------------------
vtkStatisticsAlgorithm::~vtkStatisticsAlgorithm()
{
  this->SetAssessNames( 0 );
  delete this->Internals;
}

// ----------------------------------------------------------------------
void vtkStatisticsAlgorithm::PrintSelf( ostream &os, vtkIndent indent )
{
  this->Superclass::PrintSelf( os, indent );
  os << indent << "Learn: " << this->LearnOption << endl;
  os << indent << "Derive: " << this->DeriveOption << endl;
  os << indent << "Assess: " << this->AssessOption << endl;
  os << indent << "Test: " << this->TestOption << endl;
  os << indent << "NumberOfPrimaryTables: " << this->NumberOfPrimaryTables << endl;
  if ( this->AssessNames )
  {
    this->AssessNames->PrintSelf( os, indent.GetNextIndent() );
  }
  os << indent << "Internals: " << this->Internals << endl;
}

// ----------------------------------------------------------------------
int vtkStatisticsAlgorithm::FillInputPortInformation( int port, vtkInformation* info )
{
  if ( port == INPUT_DATA )
  {
    info->Set( vtkAlgorithm::INPUT_IS_OPTIONAL(), 1 );
    info->Set( vtkAlgorithm::INPUT_REQUIRED_DATA_TYPE(), "vtkTable" );
    return 1;
  }
  else if ( port == INPUT_MODEL )
  {
    info->Set( vtkAlgorithm::INPUT_IS_OPTIONAL(), 1 );
    info->Set( vtkAlgorithm::INPUT_REQUIRED_DATA_TYPE(), "vtkMultiBlockDataSet" );
    return 1;
  }
  else if ( port == LEARN_PARAMETERS )
  {
    info->Set( vtkAlgorithm::INPUT_IS_OPTIONAL(), 1 );
    info->Set( vtkAlgorithm::INPUT_REQUIRED_DATA_TYPE(), "vtkTable" );
    return 1;
  }

  return 0;
}

// ----------------------------------------------------------------------
int vtkStatisticsAlgorithm::FillOutputPortInformation( int port, vtkInformation* info )
{
  if ( port == OUTPUT_DATA )
  {
    info->Set( vtkDataObject::DATA_TYPE_NAME(), "vtkTable" );
    return 1;
  }
  else if ( port == OUTPUT_MODEL )
  {
    info->Set( vtkDataObject::DATA_TYPE_NAME(), "vtkMultiBlockDataSet" );
    return 1;
  }
  else if ( port == OUTPUT_TEST )
  {
    info->Set( vtkDataObject::DATA_TYPE_NAME(), "vtkTable" );
    return 1;
  }

  return 0;
}

//---------------------------------------------------------------------------
void vtkStatisticsAlgorithm::SetColumnStatus( const char* namCol, int status )
{
  this->Internals->SetBufferColumnStatus( namCol, status );
}

//---------------------------------------------------------------------------
void vtkStatisticsAlgorithm::ResetAllColumnStates()
{
  this->Internals->ResetBuffer();
}

//---------------------------------------------------------------------------
int vtkStatisticsAlgorithm::RequestSelectedColumns()
{
  return this->Internals->AddBufferToRequests();
}

//---------------------------------------------------------------------------
void vtkStatisticsAlgorithm::ResetRequests()
{
  this->Internals->ResetRequests();
}

//---------------------------------------------------------------------------
vtkIdType vtkStatisticsAlgorithm::GetNumberOfRequests()
{
  return this->Internals->GetNumberOfRequests();
}

//---------------------------------------------------------------------------
vtkIdType vtkStatisticsAlgorithm::GetNumberOfColumnsForRequest( vtkIdType request )
{
  return this->Internals->GetNumberOfColumnsForRequest( request );
}

//---------------------------------------------------------------------------
const char* vtkStatisticsAlgorithm::GetColumnForRequest( vtkIdType r, vtkIdType c )
{
  static vtkStdString columnName;
  if ( this->Internals->GetColumnForRequest( r, c, columnName ) )
  {
    return columnName.c_str();
  }
  return 0;
}

//---------------------------------------------------------------------------
int vtkStatisticsAlgorithm::GetColumnForRequest( vtkIdType r, vtkIdType c, vtkStdString& columnName )
{
  return this->Internals->GetColumnForRequest( r, c, columnName ) ? 1 : 0;
}

// ----------------------------------------------------------------------
void vtkStatisticsAlgorithm::AddColumn( const char* namCol )
{
  if ( this->Internals->AddColumnToRequests( namCol ) )
  {
    this->Modified();
  }
}

// ----------------------------------------------------------------------
void vtkStatisticsAlgorithm::AddColumnPair( const char* namColX, const char* namColY )
{
  if ( this->Internals->AddColumnPairToRequests( namColX, namColY ) )
  {
    this->Modified();
  }
}

// ----------------------------------------------------------------------
bool vtkStatisticsAlgorithm::SetParameter( const char* vtkNotUsed(parameter),
                                           int vtkNotUsed(index),
                                           vtkVariant vtkNotUsed(value) )
{
  return false;
}

// ----------------------------------------------------------------------
int vtkStatisticsAlgorithm::RequestData( vtkInformation*,
                                         vtkInformationVector** inputVector,
                                         vtkInformationVector* outputVector )
{
  // Extract inputs
  vtkTable*             inData       = vtkTable::GetData( inputVector[INPUT_DATA], 0 );
  vtkMultiBlockDataSet* inModel      = vtkMultiBlockDataSet::GetData( inputVector[INPUT_MODEL], 0 );
  vtkTable*             inParameters = vtkTable::GetData( inputVector[LEARN_PARAMETERS], 0 );

  // Extract outputs
  vtkTable*             outData  = vtkTable::GetData( outputVector, OUTPUT_DATA );
  vtkMultiBlockDataSet* outModel = vtkMultiBlockDataSet::GetData( outputVector, OUTPUT_MODEL );
  vtkTable*             outTest  = vtkTable::GetData( outputVector, OUTPUT_TEST );

  // If input data table is not null then shallow copy it to output
  if ( inData )
  {
    outData->ShallowCopy( inData );
  }

  // If there are any columns selected in the buffer which have not been
  // turned into a request by RequestSelectedColumns(), add them now.
  // There should be no effect if vtkStatisticsAlgorithmPrivate::Buffer is empty.
  // This is here to accommodate the simpler user interfaces in OverView for
  // univariate and bivariate algorithms which will not call RequestSelectedColumns()
  // on their own.
  this->RequestSelectedColumns();

  // Calculate primary statistics if requested
  if ( this->LearnOption )
  {
    // First, learn primary statistics from data; otherwise, only use input model as output model
    this->Learn( inData, inParameters, outModel );

    // Second, aggregate learned models with input model if one is present
    if ( inModel )
    {
      vtkDataObjectCollection* models = vtkDataObjectCollection::New();
      models->AddItem( inModel );
      models->AddItem( outModel );
      this->Aggregate( models, outModel );
      models->Delete();
    }
  }
  else
  {
    // No input data and no input model result in an error condition
    if ( ! inModel )
    {
      vtkErrorMacro( "No model available AND no Learn phase requested. Cannot proceed with statistics algorithm." );
      return 1;
    }

    // Since no learn phase was requested, the output model is equal to the input one
    outModel->ShallowCopy( inModel );
  }

  // Calculate derived statistics if requested
  if ( this->DeriveOption )
  {
    this->Derive( outModel );
  }

  // Assess data with respect to statistical model if requested
  if ( this->AssessOption )
  {
    this->Assess( inData, outModel, outData );
  }

  // Calculate test statistics if requested
  if ( this->TestOption )
  {
    this->Test( inData, outModel, outTest );
  }

  return 1;
}

// ----------------------------------------------------------------------
void vtkStatisticsAlgorithm::Assess( vtkTable* inData,
                                     vtkMultiBlockDataSet* inMeta,
                                     vtkTable* outData,
                                     int numVariables )
{
  if ( ! inData )
  {
    return;
  }

  if ( ! inMeta )
  {
    return;
  }

  // Loop over requests
  for ( std::set<std::set<vtkStdString> >::const_iterator rit = this->Internals->Requests.begin();
        rit != this->Internals->Requests.end(); ++ rit )
  {
    // Storage for variable names of the request (smart pointer because of several exit points)
    vtkSmartPointer<vtkStringArray> varNames = vtkSmartPointer<vtkStringArray>::New();
    varNames->SetNumberOfValues( numVariables );

    // Each request must contain numVariables columns of interest (additional columns are ignored)
    bool invalidRequest = false;
    int v = 0;
    for ( std::set<vtkStdString>::const_iterator it = rit->begin();
          v < numVariables && it != rit->end(); ++ v, ++ it )
    {
      // Try to retrieve column with corresponding name in input data
      vtkStdString varName = *it;

      // If requested column does not exist in input, ignore request
      if ( ! inData->GetColumnByName( varName ) )
      {
        vtkWarningMacro( "InData table does not have a column "
                         << varName.c_str()
                         << ". Ignoring request containing it." );

        invalidRequest = true;
        break;
      }

      // If column with corresponding name was found, store name
      varNames->SetValue( v, varName );
    }
    if ( invalidRequest )
    {
      continue;
    }

    // If request is too short, it must also be ignored
    if ( v < numVariables )
    {
      vtkWarningMacro( "Only "
                       << v
                       << " variables in the request while "
                       << numVariables
                       << "are needed. Ignoring request." );

      continue;
    }

    // Store names to be able to use SetValueByName, and create the outData columns
    vtkIdType nAssessments = this->AssessNames->GetNumberOfValues();
    vtkStdString* names = new vtkStdString[nAssessments];
    vtkIdType nRowData = inData->GetNumberOfRows();
    for ( vtkIdType a = 0; a < nAssessments; ++ a )
    {
      // Prepare string for numVariables-tuple of variable names
      std::ostringstream assessColName;
      assessColName << this->AssessNames->GetValue( a )
                    << "(";
      for ( int i = 0 ; i < numVariables ; ++ i )
      {
        // Insert comma before each variable name, save the first one
        if ( i > 0 )
        {
          assessColName << ",";
        }
        assessColName << varNames->GetValue( i );
      }
      assessColName << ")";

      names[a] = assessColName.str().c_str();

      // Create assessment columns with names <AssessmentName>(var1,...,varN)
      vtkDoubleArray* assessColumn = vtkDoubleArray::New();
      assessColumn->SetName( names[a] );
      assessColumn->SetNumberOfTuples( nRowData  );
      outData->AddColumn( assessColumn );
      assessColumn->Delete();
    }

    // Select assess functor
    AssessFunctor* dfunc;
    this->SelectAssessFunctor( outData,
                               inMeta,
                               varNames,
                               dfunc );

    if ( ! dfunc )
    {
      // Functor selection did not work. Do nothing.
      vtkWarningMacro( "AssessFunctors could not be allocated. Ignoring request." );
    }
    else
    {
      // Assess each entry of the column
      vtkDoubleArray* assessResult = vtkDoubleArray::New();
      for ( vtkIdType r = 0; r < nRowData; ++ r )
      {
        // Apply functor
        (*dfunc)( assessResult, r );
        for ( vtkIdType a = 0; a < nAssessments; ++ a )
        {
          // Store each assessment value in corresponding assessment column
          outData->SetValueByName( r,
                                   names[a],
                                   assessResult->GetValue( a ) );
        }
      }

      assessResult->Delete();
    }

    delete dfunc;
    delete [] names;
  }
}