ObjectiveFunctionTest.cpp

Go to the documentation of this file.

/* *****************************************************************
    MESQUITE -- The Mesh Quality Improvement Toolkit

    Copyright 2004 Sandia Corporation and Argonne National
    Laboratory.  Under the terms of Contract DE-AC04-94AL85000
    with Sandia Corporation, the U.S. Government retains certain
    rights in this software.

    This library is free software; you can redistribute it and/or
    modify it under the terms of the GNU Lesser General Public
    License as published by the Free Software Foundation; either
    version 2.1 of the License, or (at your option) any later version.

    This library 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
    Lesser General Public License for more details.

    You should have received a copy of the GNU Lesser General Public License
    (lgpl.txt) along with this library; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

    [email protected], [email protected], [email protected],
    [email protected], [email protected], [email protected]

  ***************************************************************** */
// -*- Mode : c++; tab-width: 3; c-tab-always-indent: t; indent-tabs-mode: nil; c-basic-offset: 3
// -*-
//
//   SUMMARY:
//     USAGE:
//
//    AUTHOR: Thomas Leurent <[email protected]>
//       ORG: Argonne National Laboratory
//    E-MAIL: [email protected]
//
// ORIG-DATE: 13-Nov-02 at 18:05:56
//  LAST-MOD: 23-Jul-03 at 17:40:28 by Thomas Leurent
//
// DESCRIPTION:
// ============
/*! \file ObjectiveFunctionTest.cpp

Unit testing of various functions in the ObjectiveFunction class.
*/
// DESCRIP-END.
//

#include "Mesquite.hpp"
#include "LPtoPTemplate.hpp"
#include "MaxTemplate.hpp"
#include "LInfTemplate.hpp"
#include "IdealWeightInverseMeanRatio.hpp"
#include "MsqHessian.hpp"

#include "PatchDataInstances.hpp"
#include "ObjectiveFunctionTests.hpp"
#include "cppunit/extensions/HelperMacros.h"
#include "UnitUtil.hpp"
#include <iterator>
#include <algorithm>

using namespace MBMesquite;
using std::cerr;
using std::cout;
using std::endl;

class ObjectiveFunctionTest : public CppUnit::TestFixture, public ObjectiveFunctionTests
{
  private:
    CPPUNIT_TEST_SUITE( ObjectiveFunctionTest );

    CPPUNIT_TEST( test_eval_OF_value_LPtoP_L1 );
    CPPUNIT_TEST( test_grad_OF_value_LPtoP_L1 );
    CPPUNIT_TEST( test_diag_OF_value_LPtoP_L1 );
    CPPUNIT_TEST( test_hess_OF_value_LPtoP_L1 );
    CPPUNIT_TEST( test_eval_OF_value_LPtoP_L2 );
    CPPUNIT_TEST( test_grad_OF_value_LPtoP_L2 );
    CPPUNIT_TEST( test_diag_OF_value_LPtoP_L2 );
    CPPUNIT_TEST( test_hess_OF_value_LPtoP_L2 );
    CPPUNIT_TEST( test_eval_OF_value_LPtoP_scaled );
    CPPUNIT_TEST( test_grad_OF_value_LPtoP_scaled );
    CPPUNIT_TEST( test_diag_OF_value_LPtoP_scaled );
    CPPUNIT_TEST( test_hess_OF_value_LPtoP_scaled );
    CPPUNIT_TEST( test_eval_OF_value_LInf );
    CPPUNIT_TEST( test_eval_OF_value_max );

    CPPUNIT_TEST( test_compute_gradient_LPtoPTemplate_L1 );
    CPPUNIT_TEST( test_compute_gradient_LPtoPTemplate_L2 );
    CPPUNIT_TEST( test_compute_gradient_LPtoPTemplate_L2_scaled );
    CPPUNIT_TEST( test_hessian_gradient_LPtoPTemplate_L1 );
    CPPUNIT_TEST( test_hessian_gradient_LPtoPTemplate_L2 );
    CPPUNIT_TEST( test_hessian_gradient_LPtoPTemplate_L2_scaled );
    CPPUNIT_TEST( test_diagonal_gradient_LPtoPTemplate_L1 );
    CPPUNIT_TEST( test_diagonal_gradient_LPtoPTemplate_L2 );
    CPPUNIT_TEST( test_diagonal_gradient_LPtoPTemplate_L2_scaled );
    CPPUNIT_TEST( test_hessian_diagonal_LPtoPTemplate_L1 );
    CPPUNIT_TEST( test_hessian_diagonal_LPtoPTemplate_L2 );
    CPPUNIT_TEST( test_hessian_diagonal_LPtoPTemplate_L2_scaled );
    CPPUNIT_TEST( test_compute_ana_hessian_tet );
    CPPUNIT_TEST( test_compute_ana_hessian_tet_scaled );

    CPPUNIT_TEST( test_LPtoP_invalid_qm_eval );
    CPPUNIT_TEST( test_LPtoP_invalid_qm_grad );
    CPPUNIT_TEST( test_LPtoP_invalid_qm_diag );
    CPPUNIT_TEST( test_LPtoP_invalid_qm_hess );
    CPPUNIT_TEST( test_LInf_invalid_qm_eval );
    CPPUNIT_TEST( test_max_invalid_qm_eval );

    CPPUNIT_TEST( test_LPtoP_qm_error_eval );
    CPPUNIT_TEST( test_LPtoP_qm_error_grad );
    CPPUNIT_TEST( test_LPtoP_qm_error_diag );
    CPPUNIT_TEST( test_LPtoP_qm_error_hess );
    CPPUNIT_TEST( test_LInf_qm_error_eval );
    CPPUNIT_TEST( test_max_qm_error_eval );

    CPPUNIT_TEST( test_LPtoP_eval_calc );
    CPPUNIT_TEST( test_LPtoP_eval_accum );
    CPPUNIT_TEST( test_LPtoP_eval_save );
    CPPUNIT_TEST( test_LPtoP_eval_update );
    CPPUNIT_TEST( test_LPtoP_eval_temp );

    CPPUNIT_TEST( test_LPtoP_grad_calc );
    CPPUNIT_TEST( test_LPtoP_grad_save );
    CPPUNIT_TEST( test_LPtoP_grad_update );
    CPPUNIT_TEST( test_LPtoP_grad_temp );

    CPPUNIT_TEST( test_LPtoP_diag_calc );
    CPPUNIT_TEST( test_LPtoP_diag_save );
    CPPUNIT_TEST( test_LPtoP_diag_update );
    CPPUNIT_TEST( test_LPtoP_diag_temp );

    CPPUNIT_TEST( test_LPtoP_hess_calc );
    CPPUNIT_TEST( test_LPtoP_hess_save );
    CPPUNIT_TEST( test_LPtoP_hess_update );
    CPPUNIT_TEST( test_LPtoP_hess_temp );

    CPPUNIT_TEST( test_LPtoP_clone_L1 );
    CPPUNIT_TEST( test_LPtoP_clone_L2 );
    CPPUNIT_TEST( test_LInf_clone );
    CPPUNIT_TEST( test_max_clone );

    CPPUNIT_TEST( test_LPtoP_negate_flag_eval );
    CPPUNIT_TEST( test_LPtoP_negate_flag_grad );
    CPPUNIT_TEST( test_LPtoP_negate_flag_diag );
    CPPUNIT_TEST( test_LPtoP_negate_flag_hess );
    CPPUNIT_TEST( test_LInf_negate_flag_eval );
    CPPUNIT_TEST( test_max_negate_flag_eval );

    CPPUNIT_TEST_SUITE_END();

    void test_LPtoP_value( short P, bool scale, const std::vector< double >& values, OFTestMode mode );

    void test_LInf_value( const std::vector< double >& values );

    void test_max_value( const std::vector< double >& values );

    void test_max_negate_flag( ObjectiveFunctionTemplate& OF );

    std::vector< double > values1, values2;

  public:
    ObjectiveFunctionTest() {}

    void setUp()
    {
        const double v1[] = { 0, 1, 2, 3, 4, 5 };
        const double v2[] = { -2.5, 0.5, 3.0, M_PI, -25.0, 1.0 };
        std::copy( v1, v1 + sizeof( v1 ) / sizeof( v1[0] ), std::back_inserter( values1 ) );
        std::copy( v2, v2 + sizeof( v2 ) / sizeof( v2[0] ), std::back_inserter( values2 ) );
    }

    void test_eval_OF_value_LPtoP_L1()
    {
        test_LPtoP_value( 1, false, values1, EVAL );
        test_LPtoP_value( 1, false, values2, EVAL );
    }

    void test_grad_OF_value_LPtoP_L1()
    {
        test_LPtoP_value( 1, false, values1, GRAD );
        test_LPtoP_value( 1, false, values2, GRAD );
    }

    void test_diag_OF_value_LPtoP_L1()
    {
        test_LPtoP_value( 1, false, values1, DIAG );
        test_LPtoP_value( 1, false, values2, DIAG );
    }

    void test_hess_OF_value_LPtoP_L1()
    {
        test_LPtoP_value( 1, false, values1, HESS );
        test_LPtoP_value( 1, false, values2, HESS );
    }

    void test_eval_OF_value_LPtoP_L2()
    {
        test_LPtoP_value( 2, false, values1, EVAL );
        test_LPtoP_value( 2, false, values2, EVAL );
    }

    void test_grad_OF_value_LPtoP_L2()
    {
        test_LPtoP_value( 2, false, values1, GRAD );
        test_LPtoP_value( 2, false, values2, GRAD );
    }

    void test_diag_OF_value_LPtoP_L2()
    {
        test_LPtoP_value( 2, false, values1, DIAG );
        test_LPtoP_value( 2, false, values2, DIAG );
    }

    void test_hess_OF_value_LPtoP_L2()
    {
        test_LPtoP_value( 2, false, values1, HESS );
        test_LPtoP_value( 2, false, values2, HESS );
    }

    void test_eval_OF_value_LPtoP_scaled()
    {
        test_LPtoP_value( 2, true, values1, EVAL );
        test_LPtoP_value( 2, true, values2, EVAL );
    }

    void test_grad_OF_value_LPtoP_scaled()
    {
        test_LPtoP_value( 2, true, values1, GRAD );
        test_LPtoP_value( 2, true, values2, GRAD );
    }

    void test_diag_OF_value_LPtoP_scaled()
    {
        test_LPtoP_value( 2, true, values1, DIAG );
        test_LPtoP_value( 2, true, values2, DIAG );
    }

    void test_hess_OF_value_LPtoP_scaled()
    {
        test_LPtoP_value( 2, true, values1, EVAL );
        test_LPtoP_value( 2, true, values2, EVAL );
    }

    void test_eval_OF_value_LInf()
    {
        test_LInf_value( values1 );
        test_LInf_value( values2 );
    }

    void test_eval_OF_value_max()
    {
        test_max_value( values1 );
        test_max_value( values2 );
    }

    void test_compute_gradient_LPtoPTemplate_L1()
    {
        LPtoPTemplate LP1( 1, NULL );
        compare_numerical_gradient( &LP1 );
    }

    void test_compute_gradient_LPtoPTemplate_L2()
    {
        LPtoPTemplate LP1( 2, NULL );
        compare_numerical_gradient( &LP1 );
    }

    void test_compute_gradient_LPtoPTemplate_L2_scaled()
    {
        LPtoPTemplate LP1( 2, NULL );
        LP1.set_dividing_by_n( true );
        compare_numerical_gradient( &LP1 );
    }

    void test_hessian_gradient_LPtoPTemplate_L1()
    {
        LPtoPTemplate LP1( 1, NULL );
        compare_hessian_gradient( &LP1 );
    }

    void test_hessian_gradient_LPtoPTemplate_L2()
    {
        LPtoPTemplate LP1( 2, NULL );
        compare_hessian_gradient( &LP1 );
    }

    void test_hessian_gradient_LPtoPTemplate_L2_scaled()
    {
        LPtoPTemplate LP1( 2, NULL );
        LP1.set_dividing_by_n( true );
        compare_hessian_gradient( &LP1 );
    }

    void test_diagonal_gradient_LPtoPTemplate_L1()
    {
        LPtoPTemplate LP1( 1, NULL );
        compare_diagonal_gradient( &LP1 );
    }

    void test_diagonal_gradient_LPtoPTemplate_L2()
    {
        LPtoPTemplate LP1( 2, NULL );
        compare_diagonal_gradient( &LP1 );
    }

    void test_diagonal_gradient_LPtoPTemplate_L2_scaled()
    {
        LPtoPTemplate LP1( 2, NULL );
        LP1.set_dividing_by_n( true );
        compare_diagonal_gradient( &LP1 );
    }

    void test_hessian_diagonal_LPtoPTemplate_L1()
    {
        LPtoPTemplate LP1( 1, NULL );
        compare_hessian_diagonal( &LP1 );
    }

    void test_hessian_diagonal_LPtoPTemplate_L2()
    {
        LPtoPTemplate LP1( 2, NULL );
        compare_hessian_diagonal( &LP1 );
    }

    void test_hessian_diagonal_LPtoPTemplate_L2_scaled()
    {
        LPtoPTemplate LP1( 2, NULL );
        LP1.set_dividing_by_n( true );
        compare_hessian_diagonal( &LP1 );
    }

    void test_compute_ana_hessian_tet();

    void test_compute_ana_hessian_tet_scaled();

    void test_LPtoP_invalid_qm_eval()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_invalid_qm( EVAL, &LP1 );
    }
    void test_LPtoP_invalid_qm_grad()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_invalid_qm( GRAD, &LP1 );
    }
    void test_LPtoP_invalid_qm_diag()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_invalid_qm( DIAG, &LP1 );
    }
    void test_LPtoP_invalid_qm_hess()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_invalid_qm( HESS, &LP1 );
    }
    void test_LInf_invalid_qm_eval()
    {
        LInfTemplate OF( NULL );
        test_handles_invalid_qm( EVAL, &OF );
    }
    void test_max_invalid_qm_eval()
    {
        MaxTemplate OF( NULL );
        test_handles_invalid_qm( EVAL, &OF );
    }

    void test_LPtoP_qm_error_eval()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_qm_error( EVAL, &LP1 );
    }
    void test_LPtoP_qm_error_grad()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_qm_error( GRAD, &LP1 );
    }
    void test_LPtoP_qm_error_diag()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_qm_error( DIAG, &LP1 );
    }
    void test_LPtoP_qm_error_hess()
    {
        LPtoPTemplate LP1( 1, NULL );
        test_handles_qm_error( HESS, &LP1 );
    }
    void test_LInf_qm_error_eval()
    {
        LInfTemplate OF( NULL );
        test_handles_qm_error( EVAL, &OF );
    }
    void test_max_qm_error_eval()
    {
        MaxTemplate OF( NULL );
        test_handles_qm_error( EVAL, &OF );
    }

    void test_LPtoP_eval_calc()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::CALCULATE, EVAL, &OF );
    }
    void test_LPtoP_eval_accum()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::ACCUMULATE, EVAL, &OF );
    }
    void test_LPtoP_eval_save()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::SAVE, EVAL, &OF );
    }
    void test_LPtoP_eval_update()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::UPDATE, EVAL, &OF );
    }
    void test_LPtoP_eval_temp()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::TEMPORARY, EVAL, &OF );
    }

    void test_LPtoP_grad_calc()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::CALCULATE, GRAD, &OF );
    }
    void test_LPtoP_grad_save()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::SAVE, GRAD, &OF );
    }
    void test_LPtoP_grad_update()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::UPDATE, GRAD, &OF );
    }
    void test_LPtoP_grad_temp()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::TEMPORARY, GRAD, &OF );
    }

    void test_LPtoP_diag_calc()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::CALCULATE, DIAG, &OF );
    }
    void test_LPtoP_diag_save()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::SAVE, DIAG, &OF );
    }
    void test_LPtoP_diag_update()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::UPDATE, DIAG, &OF );
    }
    void test_LPtoP_diag_temp()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::TEMPORARY, DIAG, &OF );
    }

    void test_LPtoP_hess_calc()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::CALCULATE, HESS, &OF );
    }
    void test_LPtoP_hess_save()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::SAVE, HESS, &OF );
    }
    void test_LPtoP_hess_update()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::UPDATE, HESS, &OF );
    }
    void test_LPtoP_hess_temp()
    {
        LPtoPTemplate OF( 1, NULL );
        test_eval_type( ObjectiveFunction::TEMPORARY, HESS, &OF );
    }

    void test_LPtoP_clone_L1()
    {
        LPtoPTemplate OF( 1, NULL );
        test_clone( &OF );
    }
    void test_LPtoP_clone_L2()
    {
        LPtoPTemplate OF( 2, NULL );
        test_clone( &OF );
    }
    void test_LInf_clone()
    {
        LInfTemplate OF( NULL );
        test_clone( &OF );
    }
    void test_max_clone()
    {
        MaxTemplate OF( NULL );
        test_clone( &OF );
    }

    // This function tests to make sure that LPtoP handles the negate
    // flag correctly for the analytical Hessian, gradient, and evaluate.
    // It does this by creating two objective functions that are
    // identical except that the metric in one has negate flag of -1
    // and the metric in the other has a negate flag of 1.  Thus,
    // the Hessians, gradients, and function values should be the
    // the same except for a negative sign.
    void test_LPtoP_negate_flag_eval()
    {
        LPtoPTemplate LP( 2, NULL );
        test_negate_flag( EVAL, &LP );
    }
    void test_LPtoP_negate_flag_grad()
    {
        LPtoPTemplate LP( 2, NULL );
        test_negate_flag( GRAD, &LP );
    }
    void test_LPtoP_negate_flag_diag()
    {
        LPtoPTemplate LP( 2, NULL );
        test_negate_flag( DIAG, &LP );
    }
    void test_LPtoP_negate_flag_hess()
    {
        LPtoPTemplate LP( 2, NULL );
        test_negate_flag( HESS, &LP );
    }
    void test_LInf_negate_flag_eval()
    {
        LInfTemplate OF( NULL );
        test_max_negate_flag( OF );
    }
    void test_max_negate_flag_eval()
    {
        MaxTemplate OF( NULL );
        test_max_negate_flag( OF );
    }
};

CPPUNIT_TEST_SUITE_NAMED_REGISTRATION( ObjectiveFunctionTest, "ObjectiveFunctionTest" );
CPPUNIT_TEST_SUITE_NAMED_REGISTRATION( ObjectiveFunctionTest, "Unit" );

void ObjectiveFunctionTest::test_compute_ana_hessian_tet()
{
    MsqPrintError err( cout );
    PatchData tetPatch;
    create_qm_two_tet_patch( tetPatch, err );
    ASSERT_NO_ERROR( err );

    // creates a mean ratio quality metric ...
    IdealWeightInverseMeanRatio* mean_ratio = new IdealWeightInverseMeanRatio( err );
    CPPUNIT_ASSERT( !err );
    mean_ratio->set_averaging_method( QualityMetric::SUM );

    // ... and builds an objective function with it
    LPtoPTemplate LP2( mean_ratio, 2, err );

    MsqHessian H;
    std::vector< Vector3D > g;
    double dummy;
    H.initialize( tetPatch, err );
    CPPUNIT_ASSERT( !err );
    LP2.evaluate_with_Hessian( ObjectiveFunction::CALCULATE, tetPatch, dummy, g, H, err );
    CPPUNIT_ASSERT( !err );
    CPPUNIT_ASSERT_EQUAL( tetPatch.num_free_vertices(), g.size() );

    Matrix3D mat00( " 2.44444  0.2566   0.181444 "
                    " 0.2566   2.14815  0.104757 "
                    " 0.181444 0.104757 2.07407 " );

    Matrix3D mat13( " 5.47514 3.16659    9.83479 "
                    " -1.11704 -5.29718 -3.67406 "
                    " 10.3635 -13.5358  -15.5638 " );

    CPPUNIT_ASSERT_MATRICES_EQUAL( mat00, *H.get_block( 0, 0 ), 1e-4 );
    CPPUNIT_ASSERT_MATRICES_EQUAL( mat13, *H.get_block( 1, 3 ), 1e-4 );

    delete mean_ratio;
}

void ObjectiveFunctionTest::test_compute_ana_hessian_tet_scaled()
{
    MsqPrintError err( cout );
    PatchData tetPatch;
    create_qm_two_tet_patch( tetPatch, err );
    ASSERT_NO_ERROR( err );

    // creates a mean ratio quality metric ...
    IdealWeightInverseMeanRatio* mean_ratio = new IdealWeightInverseMeanRatio( err );
    CPPUNIT_ASSERT( !err );
    mean_ratio->set_averaging_method( QualityMetric::SUM );

    // ... and builds an objective function with it
    LPtoPTemplate LP2( mean_ratio, 2, err );
    LP2.set_dividing_by_n( true );

    MsqHessian H;
    std::vector< Vector3D > g;
    double dummy;
    H.initialize( tetPatch, err );
    CPPUNIT_ASSERT( !err );
    LP2.evaluate_with_Hessian( ObjectiveFunction::CALCULATE, tetPatch, dummy, g, H, err );
    CPPUNIT_ASSERT( !err );
    CPPUNIT_ASSERT_EQUAL( tetPatch.num_free_vertices(), g.size() );

    Matrix3D mat00( " 2.44444  0.2566   0.181444 "
                    " 0.2566   2.14815  0.104757 "
                    " 0.181444 0.104757 2.07407 " );

    mat00 *= .5;

    Matrix3D mat13( " 5.47514 3.16659    9.83479 "
                    " -1.11704 -5.29718 -3.67406 "
                    " 10.3635 -13.5358  -15.5638 " );

    mat13 *= .5;

    CPPUNIT_ASSERT_MATRICES_EQUAL( mat00, *H.get_block( 0, 0 ), 1e-4 );
    CPPUNIT_ASSERT_MATRICES_EQUAL( mat13, *H.get_block( 1, 3 ), 1e-4 );

    //    cout << H <<endl;
    delete mean_ratio;
}

void ObjectiveFunctionTest::test_LPtoP_value( short P,
                                              bool scale,
                                              const std::vector< double >& values,
                                              OFTestMode mode )
{
    CPPUNIT_ASSERT( !values.empty() );

    LPtoPTemplate OF( P, NULL );
    OF.set_dividing_by_n( scale );

    double expected = 0.0;
    for( std::vector< double >::const_iterator i = values.begin(); i != values.end(); ++i )
        expected += std::pow( fabs( *i ), P );
    if( scale ) expected /= values.size();

    test_value( arrptr( values ), values.size(), expected, mode, &OF );
}

void ObjectiveFunctionTest::test_LInf_value( const std::vector< double >& values )
{
    CPPUNIT_ASSERT( !values.empty() );

    LInfTemplate OF( NULL );

    double expected = fabs( values[0] );
    for( unsigned i = 1; i < values.size(); ++i )
        if( fabs( values[i] ) > expected ) expected = fabs( values[i] );

    test_value( arrptr( values ), values.size(), expected, EVAL, &OF );
}

void ObjectiveFunctionTest::test_max_value( const std::vector< double >& values )
{
    CPPUNIT_ASSERT( !values.empty() );

    MaxTemplate OF( NULL );

    double expected = fabs( values[0] );
    for( unsigned i = 1; i < values.size(); ++i )
        if( values[i] > expected ) expected = values[i];

    test_value( arrptr( values ), values.size(), expected, EVAL, &OF );
}

void ObjectiveFunctionTest::test_max_negate_flag( ObjectiveFunctionTemplate& of )
{
    // use only positive values so this test works for both
    // MaxTemplate and LInfTemplate
    const double some_vals[] = { 1, 2, 3, 4, 5, 6, 0.5 };
    const unsigned num_vals  = sizeof( some_vals ) / sizeof( some_vals[0] );
    OFTestQM metric( some_vals, num_vals );
    metric.set_negate_flag( -1 );
    of.set_quality_metric( &metric );

    // get OF value
    MsqPrintError err( cout );
    bool rval;
    double value;
    rval = of.evaluate( ObjectiveFunction::CALCULATE, patch(), value, false, err );
    ASSERT_NO_ERROR( err );
    CPPUNIT_ASSERT( rval );

    // find min value
    double expected = *std::min_element( some_vals, some_vals + num_vals );
    CPPUNIT_ASSERT_DOUBLES_EQUAL( -expected, value, 1e-6 );
}