tdf#137679 Use Kahan summation for interpr2.cxx

The changes in the text files are in the latest decimals.
Those should be more accurate now.

Change-Id: I3814e1939f71debd5ddde9408a025af7a0a2cac5
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/114737
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
diff --git a/sc/inc/kahan.hxx b/sc/inc/kahan.hxx
index 968da16..ffeeab0 100644
--- a/sc/inc/kahan.hxx
+++ b/sc/inc/kahan.hxx
@@ -79,6 +79,20 @@ public:

    inline void operator-=(double fSum) { add(-fSum); }

    inline KahanSum operator+(double fSum) const
    {
        KahanSum fNSum(*this);
        fNSum.add(fSum);
        return fNSum;
    }

    inline KahanSum operator-(double fSum) const
    {
        KahanSum fNSum(*this);
        fNSum.add(-fSum);
        return fNSum;
    }

    /**
      * In some parts of the code of interpr_.cxx this may be used for
      * product instead of sum. This operator shall be used for that task.
diff --git a/sc/qa/unit/data/functions/financial/fods/vdb.fods b/sc/qa/unit/data/functions/financial/fods/vdb.fods
index aee0099..38ace90 100644
--- a/sc/qa/unit/data/functions/financial/fods/vdb.fods
+++ b/sc/qa/unit/data/functions/financial/fods/vdb.fods
@@ -2535,11 +2535,11 @@
     <table:table-cell table:number-columns-repeated="6"/>
    </table:table-row>
    <table:table-row table:style-name="ro7">
     <table:table-cell table:style-name="ce11" table:formula="of:=VDB(100000;20000;10;7;8)" office:value-type="currency" office:currency="USD" office:value="971.519999999989" calcext:value-type="currency">
     <table:table-cell table:style-name="ce11" table:formula="of:=VDB(100000;20000;10;7;8)" office:value-type="currency" office:currency="USD" office:value="971.520000000004" calcext:value-type="currency">
      <text:p>$971.52</text:p>
     </table:table-cell>
     <table:table-cell office:value-type="float" office:value="971.519999999989" calcext:value-type="float">
      <text:p>971.519999999989</text:p>
     <table:table-cell office:value-type="float" office:value="971.520000000004" calcext:value-type="float">
      <text:p>971.520000000004</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce18" table:formula="of:=ROUND([.A11];12)=ROUND([.B11];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
      <text:p>TRUE</text:p>
@@ -2948,11 +2948,11 @@
     <table:table-cell table:number-columns-repeated="12"/>
    </table:table-row>
    <table:table-row table:style-name="ro6">
     <table:table-cell table:formula="of:=VDB(100000;20000;10;7.25;7.75)" office:value-type="float" office:value="485.759999999995" calcext:value-type="float">
      <text:p>485.759999999995</text:p>
     <table:table-cell table:formula="of:=VDB(100000;20000;10;7.25;7.75)" office:value-type="float" office:value="485.760000000002" calcext:value-type="float">
      <text:p>$485.76</text:p>
     </table:table-cell>
     <table:table-cell office:value-type="float" office:value="485.759999999995" calcext:value-type="float">
      <text:p>485.759999999995</text:p>
     <table:table-cell office:value-type="float" office:value="485.760000000002" calcext:value-type="float">
      <text:p>485.760000000002</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce19" table:formula="of:=ROUND([.A33];12)=ROUND([.B33];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
      <text:p>TRUE</text:p>
@@ -3118,11 +3118,11 @@
     <table:table-cell table:number-columns-repeated="2"/>
    </table:table-row>
    <table:table-row table:style-name="ro6">
     <table:table-cell table:formula="of:=VDB([.G30];[.G31];[.G32];[.$G41];[.$G42])" office:value-type="float" office:value="971.519999999989" calcext:value-type="float">
      <text:p>971.519999999989</text:p>
     <table:table-cell table:formula="of:=VDB([.G30];[.G31];[.G32];[.$G41];[.$G42])" office:value-type="float" office:value="971.520000000004" calcext:value-type="float">
      <text:p>$971.52</text:p>
     </table:table-cell>
     <table:table-cell office:value-type="float" office:value="971.519999999989" calcext:value-type="float">
      <text:p>971.519999999989</text:p>
     <table:table-cell office:value-type="float" office:value="971.520000000004" calcext:value-type="float">
      <text:p>971.520000000004</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce19" table:formula="of:=ROUND([.A41];12)=ROUND([.B41];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
      <text:p>TRUE</text:p>
@@ -3536,10 +3536,10 @@
     <table:table-cell table:number-columns-repeated="16"/>
    </table:table-row>
    <table:table-row table:style-name="ro6">
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G49];[.G50])" office:value-type="float" office:value="242.879999999997" calcext:value-type="float">
      <text:p>242.879999999997</text:p>
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G49];[.G50])" office:value-type="float" office:value="242.880000000001" calcext:value-type="float">
      <text:p>$242.88</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="242.879999999997" calcext:value-type="float">
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="242.880000000001" calcext:value-type="float">
      <text:p>242.880000</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce19" table:formula="of:=ROUND([.A64];12)=ROUND([.B64];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
@@ -3551,10 +3551,10 @@
     <table:table-cell table:number-columns-repeated="16"/>
    </table:table-row>
    <table:table-row table:style-name="ro6">
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G50];[.G51])" office:value-type="float" office:value="485.759999999995" calcext:value-type="float">
      <text:p>485.759999999995</text:p>
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G50];[.G51])" office:value-type="float" office:value="485.760000000002" calcext:value-type="float">
      <text:p>$485.76</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="485.759999999995" calcext:value-type="float">
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="485.760000000002" calcext:value-type="float">
      <text:p>485.760000</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce19" table:formula="of:=ROUND([.A65];12)=ROUND([.B65];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
@@ -3566,10 +3566,10 @@
     <table:table-cell table:number-columns-repeated="16"/>
    </table:table-row>
    <table:table-row table:style-name="ro6">
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G49];[.G51])" office:value-type="float" office:value="728.639999999992" calcext:value-type="float">
      <text:p>728.639999999992</text:p>
     <table:table-cell table:formula="of:=VDB(100000;20000;10;[.G49];[.G51])" office:value-type="float" office:value="728.640000000003" calcext:value-type="float">
      <text:p>$728.64</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="728.639999999992" calcext:value-type="float">
     <table:table-cell table:style-name="ce15" office:value-type="float" office:value="728.640000000003" calcext:value-type="float">
      <text:p>728.640000</text:p>
     </table:table-cell>
     <table:table-cell table:style-name="ce19" table:formula="of:=ROUND([.A66];12)=ROUND([.B66];12)" office:value-type="boolean" office:boolean-value="true" calcext:value-type="boolean">
diff --git a/sc/source/core/tool/interpr2.cxx b/sc/source/core/tool/interpr2.cxx
index 62bdbc2..cac57fe 100644
--- a/sc/source/core/tool/interpr2.cxx
+++ b/sc/source/core/tool/interpr2.cxx
@@ -1308,7 +1308,7 @@ void ScInterpreter::ScNPV()
    if ( !MustHaveParamCountMin( nParamCount, 2) )
        return;

    double fVal = 0.0;
    KahanSum fVal = 0.0;
    // We turn the stack upside down!
    ReverseStack( nParamCount);
    if (nGlobalError == FormulaError::NONE)
@@ -1324,7 +1324,7 @@ void ScInterpreter::ScNPV()
            {
                case svDouble :
                {
                    fVal += (GetDouble() / pow(1.0 + fRate, fCount));
                    fVal += GetDouble() / pow(1.0 + fRate, fCount);
                    fCount++;
                }
                break;
@@ -1336,7 +1336,7 @@ void ScInterpreter::ScNPV()
                    if (!aCell.hasEmptyValue() && aCell.hasNumeric())
                    {
                        double fCellVal = GetCellValue(aAdr, aCell);
                        fVal += (fCellVal / pow(1.0 + fRate, fCount));
                        fVal += fCellVal / pow(1.0 + fRate, fCount);
                        fCount++;
                    }
                }
@@ -1350,7 +1350,7 @@ void ScInterpreter::ScNPV()
                    ScHorizontalValueIterator aValIter( mrDoc, aRange );
                    while ((nErr == FormulaError::NONE) && aValIter.GetNext(fCellVal, nErr))
                    {
                        fVal += (fCellVal / pow(1.0 + fRate, fCount));
                        fVal += fCellVal / pow(1.0 + fRate, fCount);
                        fCount++;
                    }
                    if ( nErr != FormulaError::NONE )
@@ -1384,7 +1384,7 @@ void ScInterpreter::ScNPV()
                                        return;
                                    }
                                    fx = pMat->GetDouble(j,k);
                                    fVal += (fx / pow(1.0 + fRate, fCount));
                                    fVal += fx / pow(1.0 + fRate, fCount);
                                    fCount++;
                                }
                            }
@@ -1396,7 +1396,7 @@ void ScInterpreter::ScNPV()
            }
        }
    }
    PushDouble(fVal);
    PushDouble(fVal.get());
}

void ScInterpreter::ScIRR()
@@ -1458,8 +1458,8 @@ void ScInterpreter::ScIRR()
    FormulaError nIterError = FormulaError::NONE;
    while (fEps > SCdEpsilon && nItCount < nIterationsMax && nGlobalError == FormulaError::NONE)
    {                                       // Newtons method:
        double fNom = 0.0;
        double fDenom = 0.0;
        KahanSum fNom = 0.0;
        KahanSum fDenom = 0.0;
        double fCount = 0.0;
        if (bIsMatrix)
        {
@@ -1493,7 +1493,7 @@ void ScInterpreter::ScIRR()
            }
            SetError(nIterError);
        }
        double xNew = x - fNom / fDenom;  // x(i+1) = x(i)-f(x(i))/f'(x(i))
        double xNew = x - fNom.get() / fDenom.get();  // x(i+1) = x(i)-f(x(i))/f'(x(i))
        nItCount++;
        fEps = fabs(xNew - x);
        x = xNew;
@@ -1550,9 +1550,9 @@ void ScInterpreter::ScMIRR()
        PushError( nGlobalError );
    else
    {
        double fNPV_reinvest = 0.0;
        KahanSum fNPV_reinvest = 0.0;
        double fPow_reinvest = 1.0;
        double fNPV_invest = 0.0;
        KahanSum fNPV_invest = 0.0;
        double fPow_invest = 1.0;
        sal_uLong nCount = 0;
        bool bHasPosValue = false;
@@ -1623,7 +1623,7 @@ void ScInterpreter::ScMIRR()
            PushError( nGlobalError );
        else
        {
            double fResult = -fNPV_reinvest / fNPV_invest;
            double fResult = -fNPV_reinvest.get() / fNPV_invest.get();
            fResult *= pow( fRate1_reinvest, static_cast<double>( nCount - 1 ) );
            fResult = pow( fResult, div( 1.0, (nCount - 1)) );
            PushDouble( fResult - 1.0 );
@@ -1779,17 +1779,17 @@ void ScInterpreter::ScDB()
        fDb = fFirstOffRate;
    else
    {
        double fSumOffRate = fFirstOffRate;
        KahanSum fSumOffRate = fFirstOffRate;
        double fMin = fLife;
        if (fMin > fPeriod) fMin = fPeriod;
        sal_uInt16 iMax = static_cast<sal_uInt16>(::rtl::math::approxFloor(fMin));
        for (sal_uInt16 i = 2; i <= iMax; i++)
        {
            fDb = (fCost - fSumOffRate) * fOffRate;
            fDb = -(fSumOffRate - fCost).get() * fOffRate;
            fSumOffRate += fDb;
        }
        if (fPeriod > fLife)
            fDb = ((fCost - fSumOffRate) * fOffRate * (12.0 - fMonths)) / 12.0;
            fDb = (-(fSumOffRate - fCost).get() * fOffRate * (12.0 - fMonths)) / 12.0;
    }
    PushDouble(fDb);
}
@@ -1797,7 +1797,7 @@ void ScInterpreter::ScDB()
double ScInterpreter::ScInterVDB(double fCost, double fSalvage, double fLife,
                             double fLife1, double fPeriod, double fFactor)
{
    double fVdb=0;
    KahanSum fVdb = 0.0;
    double fIntEnd   = ::rtl::math::approxCeil(fPeriod);
    sal_uLong nLoopEnd   = static_cast<sal_uLong>(fIntEnd);

@@ -1836,7 +1836,7 @@ double ScInterpreter::ScInterVDB(double fCost, double fSalvage, double fLife,

        fVdb += fTerm;
    }
    return fVdb;
    return fVdb.get();
}

void ScInterpreter::ScVDB()
@@ -1846,21 +1846,14 @@ void ScInterpreter::ScVDB()
    if ( !MustHaveParamCount( nParamCount, 5, 7 ) )
        return;

    double fCost, fSalvage, fLife, fStart, fEnd, fFactor, fVdb = 0.0;
    bool bNoSwitch;
    if (nParamCount == 7)
        bNoSwitch = GetBool();
    else
        bNoSwitch = false;
    if (nParamCount >= 6)
        fFactor = GetDouble();
    else
        fFactor = 2.0;
    fEnd   = GetDouble();
    fStart = GetDouble();
    fLife  = GetDouble();
    fSalvage   = GetDouble();
    fCost   = GetDouble();
    KahanSum fVdb = 0.0;
    bool bNoSwitch = nParamCount == 7 && GetBool();
    double  fFactor = nParamCount >= 6 ? GetDouble() : 2.0;
    double fEnd   = GetDouble();
    double fStart = GetDouble();
    double fLife  = GetDouble();
    double fSalvage   = GetDouble();
    double fCost   = GetDouble();
    if (fStart < 0.0 || fEnd < fStart || fEnd > fLife || fCost < 0.0
                      || fSalvage > fCost || fFactor <= 0.0)
        PushIllegalArgument();
@@ -1871,7 +1864,6 @@ void ScInterpreter::ScVDB()
        sal_uLong nLoopStart = static_cast<sal_uLong>(fIntStart);
        sal_uLong nLoopEnd   = static_cast<sal_uLong>(fIntEnd);

        fVdb = 0.0;
        if (bNoSwitch)
        {
            for (sal_uLong i = nLoopStart + 1; i <= nLoopEnd; i++)
@@ -1922,7 +1914,7 @@ void ScInterpreter::ScVDB()
            fVdb -= fPart;
        }
    }
    PushDouble(fVdb);
    PushDouble(fVdb.get());
}

void ScInterpreter::ScPDuration()
@@ -2321,7 +2313,7 @@ void ScInterpreter::ScCumIpmt()
        sal_uLong nStart = static_cast<sal_uLong>(fStart);
        sal_uLong nEnd = static_cast<sal_uLong>(fEnd) ;
        double fPmt = ScGetPMT(fRate, fNper, fPv, 0.0, bPayInAdvance);
        double fIpmt = 0.0;
        KahanSum fIpmt = 0.0;
        if (nStart == 1)
        {
            if (!bPayInAdvance)
@@ -2336,7 +2328,7 @@ void ScInterpreter::ScCumIpmt()
                fIpmt += ScGetFV(fRate, static_cast<double>(i-1), fPmt, fPv, false);
        }
        fIpmt *= fRate;
        PushDouble(fIpmt);
        PushDouble(fIpmt.get());
    }
}

@@ -2361,7 +2353,7 @@ void ScInterpreter::ScCumPrinc()
    {
        bool bPayInAdvance = static_cast<bool>(fFlag);
        double fPmt = ScGetPMT(fRate, fNper, fPv, 0.0, bPayInAdvance);
        double fPpmt = 0.0;
        KahanSum fPpmt = 0.0;
        sal_uLong nStart = static_cast<sal_uLong>(fStart);
        sal_uLong nEnd = static_cast<sal_uLong>(fEnd);
        if (nStart == 1)
@@ -2379,7 +2371,7 @@ void ScInterpreter::ScCumPrinc()
            else
                fPpmt += fPmt - ScGetFV(fRate, static_cast<double>(i-1), fPmt, fPv, false) * fRate;
        }
        PushDouble(fPpmt);
        PushDouble(fPpmt.get());
    }
}