starobject.cpp

/*
    SPDX-FileCopyrightText: 2001 Jason Harris <[email protected]>
 
    SPDX-License-Identifier: GPL-2.0-or-later
*/
 
#include "starobject.h"
 
#include "deepstardata.h"
#include "ksnumbers.h"
#ifndef KSTARS_LITE
#include "kspopupmenu.h"
#endif
#include "kstarsdata.h"
#include "ksutils.h"
#include "Options.h"
#include "skymap.h"
#include "stardata.h"
 
#include <typeinfo>
 
#ifdef PROFILE_UPDATECOORDS
double StarObject::updateCoordsCpuTime = 0.;
unsigned int StarObject::starsUpdated  = 0;
#include <cstdlib>
#include <ctime>
#endif
 
// DEBUG EDIT. Uncomment for testing Proper Motion
//#include "skycomponents/skymesh.h"
// END DEBUG
 
#include "skycomponents/skylabeler.h"
 
// DEBUG EDIT. Uncomment for testing Proper Motion
// You will also need to uncomment all related blocks
// from this file, starobject.h and also the trixel-boundaries
// block from lines 253 - 257 of skymapcomposite.cpp
//QVector<SkyPoint *> StarObject::Trail;
// END DEBUG
 
#include <KLocalizedString>
 
//----- Static Methods -----
//
double StarObject::reindexInterval(double pm)
{
    if (pm < 1.0e-6)
        return 1.0e6;
 
    // arcminutes * sec/min * milliarcsec/sec centuries/year
    // / [milliarcsec/year] = centuries
 
    return 25.0 * 60.0 * 10.0 / pm;
}
 
StarObject::StarObject(dms r, dms d, float m, const QString &n, const QString &n2, const QString &sptype, double pmra,
                       double pmdec, double par, bool mult, bool var, int hd)
    : SkyObject(SkyObject::STAR, r, d, m, n, n2, QString()), PM_RA(pmra), PM_Dec(pmdec), Parallax(par),
      Multiplicity(mult), Variability(var), HD(hd)
{
    QByteArray spt = sptype.toLatin1();
    SpType[0]      = spt[0];
    SpType[1]      = spt[1];
 
    QString lname;
    if (hasName())
    {
        lname = n;
        if (hasName2())
            lname += " (" + gname() + ')';
    }
    else if (hasName2())
    {
        lname = gname();
        //If genetive name exists, but no primary name, set primary name = genetive name.
        setName(gname());
    }
    else if (HD > 0)
    {
        lname = QLatin1String("HD ") + QString::number(HD);
    }
    setLongName(lname);
    updateID = updateNumID = 0;
}
 
StarObject::StarObject(double r, double d, float m, const QString &n, const QString &n2, const QString &sptype,
                       double pmra, double pmdec, double par, bool mult, bool var, int hd)
    : SkyObject(SkyObject::STAR, r, d, m, n, n2, QString()), PM_RA(pmra), PM_Dec(pmdec), Parallax(par),
      Multiplicity(mult), Variability(var), HD(hd)
{
    QByteArray spt = sptype.toLatin1();
    SpType[0]      = spt[0];
    SpType[1]      = spt[1];
 
    QString lname;
    if (hasName())
    {
        lname = n;
        if (hasName2())
            lname += " (" + gname() + ')';
    }
    else if (hasName2())
    {
        lname = gname();
        //If genetive name exists, but no primary name, set primary name = genetive name.
        setName(gname());
    }
    else if (HD > 0)
    {
        lname = QLatin1String("HD ") + QString::number(HD);
    }
    setLongName(lname);
    updateID = updateNumID = 0;
}
 
StarObject::StarObject(const StarObject &o)
    : SkyObject(o), PM_RA(o.PM_RA), PM_Dec(o.PM_Dec), Parallax(o.Parallax), Multiplicity(o.Multiplicity),
      Variability(o.Variability), HD(o.HD)
{
    SpType[0] = o.SpType[0];
    SpType[1] = o.SpType[1];
    updateID = updateNumID = 0;
}
 
StarObject *StarObject::clone() const
{
    Q_ASSERT(typeid(this) ==
             typeid(static_cast<const StarObject *>(this))); // Ensure we are not slicing a derived class
    return new StarObject(*this);
}
 
void StarObject::init(const StarData *stardata)
{
    double ra, dec;
    ra  = stardata->RA / 1000000.0;
    dec = stardata->Dec / 100000.0;
    setType(SkyObject::STAR);
    setMag(stardata->mag / 100.0);
    setRA0(ra);
    setDec0(dec);
    setRA(ra0());
    setDec(dec0());
    SpType[0]    = stardata->spec_type[0];
    SpType[1]    = stardata->spec_type[1];
    PM_RA        = stardata->dRA / 10.0;
    PM_Dec       = stardata->dDec / 10.0;
    Parallax     = stardata->parallax / 10.0;
    Multiplicity = stardata->flags & 0x02;
    Variability  = stardata->flags & 0x04;
    updateID = updateNumID = 0;
    HD                     = stardata->HD;
    if (HD > 0)
        setNames(QString(QLatin1String("HD ") + QString::number(HD)), QString());
    B = V = 99.9;
 
    // DEBUG Edit. For testing proper motion. Uncomment all related blocks to test.
    // WARNING: You can debug only ONE STAR AT A TIME, because
    //          the StarObject::Trail is static. It has to be
    //          static, because otherwise, we can run into segfaults
    //          due to the memcpy() that we do to create stars
    /*
    testStar = false;
    if( stardata->HD == 103095 && Trail.size() == 0 ) {
      // Populate Trail with various positions
        qDebug() << Q_FUNC_INFO << "TEST STAR FOUND!";
        testStar = true;
        KSNumbers num( J2000 ); // Some estimate, doesn't matter.
        long double jy;
        for( jy = -10000.0; jy <= 10000.0; jy += 500.0 ) {
            num.updateValues( J2000 + jy * 365.238 );
            double ra, dec;
            getIndexCoords( &num, &ra, &dec );
            Trail.append( new SkyPoint( ra / 15.0, dec ) );
        }
        qDebug() << Q_FUNC_INFO << "Populated the star's trail with " << Trail.size() << " entries.";
    }
    */
    // END DEBUG.
 
    lastPrecessJD = J2000;
}
 
void StarObject::init(const DeepStarData *stardata)
{
    double ra, dec, BV_Index;
 
    ra  = stardata->RA / 1000000.0;
    dec = stardata->Dec / 100000.0;
    setType(SkyObject::STAR);
 
    if (stardata->V == 30000 && stardata->B != 30000)
        setMag((stardata->B - 1600) / 1000.0); // FIXME: Is it okay to make up stuff like this?
    else
        setMag(stardata->V / 1000.0);
 
    setRA0(ra);
    setDec0(dec);
    setRA(ra);
    setDec(dec);
 
    SpType[1] = '?';
    SpType[0] = 'B';
    if (stardata->B == 30000 || stardata->V == 30000)
    {
        // Unused value
//        BV_Index  = -100;
        SpType[0] = '?';
    }
    else
    {
        BV_Index = (stardata->B - stardata->V) / 1000.0;
        if (BV_Index > 0.0) SpType[0] = 'A';
        if (BV_Index > 0.325) SpType[0] = 'F';
        if (BV_Index > 0.575) SpType[0] = 'G';
        if (BV_Index > 0.975) SpType[0] = 'K';
        if (BV_Index > 1.6) SpType[0] = 'M';
    }
 
    PM_RA        = stardata->dRA / 100.0;
    PM_Dec       = stardata->dDec / 100.0;
    Parallax     = 0.0;
    Multiplicity = 0;
    Variability  = 0;
    updateID = updateNumID = 0;
    B                      = stardata->B / 1000.0;
    V                      = stardata->V / 1000.0;
    lastPrecessJD          = J2000;
}
 
void StarObject::setNames(const QString &name, const QString &name2)
{
    QString lname;
 
    setName(name);
 
    setName2(name2);
 
    if (hasName() && name.startsWith(QLatin1String("HD")) == false)
    {
        lname = name;
        if (hasName2())
            lname += " (" + gname() + ')';
    }
    else if (hasName2())
        lname = gname();
    setLongName(lname);
}
void StarObject::initPopupMenu(KSPopupMenu *pmenu)
{
#ifdef KSTARS_LITE
    Q_UNUSED(pmenu)
#else
    pmenu->createStarMenu(this);
#endif
}
 
void StarObject::updateCoords(const KSNumbers *num, bool, const CachingDms *, const CachingDms *, bool)
{
//Correct for proper motion of stars.  Determine RA and Dec offsets.
//Proper motion is given im milliarcsec per year by the pmRA() and pmDec() functions.
//That is numerically identical to the number of arcsec per millenium, so multiply by
//KSNumbers::julianMillenia() to find the offsets in arcsec.
 
// Correction:  The method below computes the proper motion before the
// precession.  If we precessed first then the direction of the proper
// motion correction would depend on how far we've precessed.  -jbb
#ifdef PROFILE_UPDATECOORDS
    std::clock_t start, stop;
    start = std::clock();
#endif
    CachingDms saveRA = ra0(), saveDec = dec0();
    CachingDms newRA, newDec;
 
    getIndexCoords(num, newRA, newDec);
 
    setRA0(newRA);
    setDec0(newDec);
    SkyPoint::updateCoords(num);
    setRA0(saveRA);
    setDec0(saveDec);
 
#ifdef PROFILE_UPDATECOORDS
    stop = std::clock();
    updateCoordsCpuTime += double(stop - start) / double(CLOCKS_PER_SEC);
    ++starsUpdated;
#endif
}
 
bool StarObject::getIndexCoords(const KSNumbers *num, CachingDms &ra, CachingDms &dec)
{
    static double pmms;
 
    // =================== NOTE: CODE DUPLICATION ====================
    // If you modify this, please also modify the other getIndexCoords
    // ===============================================================
    //
    // Reason for code duplication is as follows:
    //
    // This method is designed to use CachingDms, i.e. we know we are
    // going to use the sine and cosine of the returned values.
    //
    // The other method is designed to avoid CachingDms and try to
    // compute as little trigonometry as possible when the ra/dec has
    // to be returned in double (used in SkyMesh::indexStar() for
    // example)
    //
    // Thus, the philosophy of writing code is different. Granted, we
    // don't need to optimize for the smaller star catalogs (which use
    // SkyMesh::indexStar()), but it is nevertheless a good idea,
    // given that getIndexCoords() shows up in callgrind as one of the
    // slightly more expensive operations.
 
    // Old, Incorrect Proper motion Computation.  We retain this in a
    // comment because we might want to use it to come up with a
    // linear approximation that's faster.
    //    double dra = pmRA() * num->julianMillenia() / ( cos( dec0().radians() ) * 3600.0 );
    //    double ddec = pmDec() * num->julianMillenia() / 3600.0;
 
 
    pmms = pmMagnitudeSquared();
 
    if (std::isnan(pmms) || pmms * num->julianMillenia() * num->julianMillenia() < .01)
    {
        // Ignore corrections
        ra  = ra0();
        dec = dec0();
        return false;
    }
 
    /*
 
    // Proper Motion Correction should be implemented as motion along a great
    // circle passing through the given (ra0, dec0) in a direction of
    // atan2( pmRA(), pmDec() ) to an angular distance given by the Magnitude of
    // PM times the number of Julian millenia since J2000.0
 
    double pm = pmMagnitude() * num->julianMillenia(); // Proper Motion in arcseconds
 
    double dir0 = ((pm > 0) ? atan2(pmRA(), pmDec()) : atan2(-pmRA(), -pmDec())); // Bearing, in radian
 
    if (pm < 0)
        pm = -pm;
 
    double dst = (pm * M_PI / (180.0 * 3600.0));
    //    double phi = M_PI / 2.0 - dec0().radians();
 
    // Note: According to callgrind, dms::dms() + dms::setRadians()
    // takes ~ 40 CPU cycles, whereas, the advantage afforded by using
    // sincos() instead of sin() and cos() calls seems to be about 30
    // CPU cycles.
 
    // So it seems like it is not worth turning dir0 and dst into dms
    // objects and using SinCos(). However, caching the values of sin
    // and cos if we are going to reuse them avoids expensive (~120
    // CPU cycle) recomputation!
    CachingDms lat1, dtheta;
    double sinDst = sin(dst), cosDst = cos(dst);
    lat1.setUsing_asin(dec0().sin() * cosDst + dec0().cos() * sinDst * cos(dir0));
    dtheta.setUsing_atan2(sin(dir0) * sinDst * dec0().cos(), cosDst - dec0().sin() * lat1.sin());
 
    ra  = ra0() + dtheta; // Use operator + to avoid trigonometry
    dec = lat1;           // Need variable lat1 because dec may refer to dec0, so cannot construct result in-place
 
    */
 
    // Use the formula given in Seidelmann (Explanatory Supplement to
    // the Astronomical Almanac) instead. The formulas used here are
    // the combination of (3.23-1), (3.23-3), (3.23-5). Not only does
    // it reduce trigonometry use, it is more likely to be correct
    // than the stuff we came up with on our own above
 
    // Although it is not explained in the above reference, a formula
    // that reduces to α' = α + μ_α * t must have μ_α be the rate of
    // change of the angle at the center of the declination circle,
    // and not the rate of change of arclength, i.e. μ_α must _not_
    // include the cos(δ) factor. I have checked that the formulas
    // used here do reduce to the above, and therefore expect μ_α =
    // pmRa / cos(δ)
 
    double cosDec, sinDec, cosRa, sinRa;
    double scale = num->julianMillenia() * (M_PI / (180.0 * 3600.0));
    dec0().SinCos(sinDec, cosDec);
    ra0().SinCos(sinRa, cosRa);
 
    // Note: Below assumes that pmRA is already pre-scaled by cos(delta), as it is for Hipparcos
    double net_pmRA = pmRA() * scale, net_pmDec = pmDec() * scale;
 
    double x0 = cosDec * cosRa, y0 = cosDec * sinRa, z0 = sinDec;
    double dX = - net_pmRA * sinRa - net_pmDec * sinDec * cosRa;
    double dY = net_pmRA * cosRa - net_pmDec * sinDec * sinRa;
    double dZ = net_pmDec * cosDec;
    double x = x0 + dX, y = y0 + dY, z = z0 + dZ;
 
    ra.setUsing_atan2(y, x);
 
    // Note: dec = asin(z) is a poor choice, because we aren't
    // guaranteed that (x, y, z) lies on the unit sphere due to the
    // first-order approximation. Therefore, we must "project" out any
    // change in the length of the vector to get our best estimate,
    // and this is achieved by using atan. In fact atan gives the
    // least-squares estimate for an angle given both its sin and
    // cosine components.
    dec.setUsing_atan2(z, sqrt(x * x + y * y));
 
    return true;
}
 
bool StarObject::getIndexCoords(const KSNumbers *num, double *ra, double *dec)
{
    static double pmms;
 
    // =================== NOTE: CODE DUPLICATION ====================
    // If you modify this, please also modify the other getIndexCoords
    // ===============================================================
    //
    // Reason for code duplication is as follows:
    //
    // This method is designed to avoid CachingDms and try to compute
    // as little trigonometry as possible when the ra/dec has to be
    // returned in double (used in SkyMesh::indexStar() for example)
    //
    // The other method is designed to use CachingDms, i.e. we know we
    // are going to use the sine and cosine of the returned values.
    //
    // Thus, the philosophy of writing code is different. Granted, we
    // don't need to optimize for the smaller star catalogs (which use
    // SkyMesh::indexStar()), but it is nevertheless a good idea,
    // given that getIndexCoords() shows up in callgrind as one of the
    // slightly more expensive operations.
 
    // Old, Incorrect Proper motion Computation.  We retain this in a
    // comment because we might want to use it to come up with a
    // linear approximation that's faster.
    //    double dra = pmRA() * num->julianMillenia() / ( cos( dec0().radians() ) * 3600.0 );
    //    double ddec = pmDec() * num->julianMillenia() / 3600.0;
 
    // Proper Motion Correction should be implemented as motion along a great
    // circle passing through the given (ra0, dec0) in a direction of
    // atan2( pmRA(), pmDec() ) to an angular distance given by the Magnitude of
    // PM times the number of Julian millenia since J2000.0
 
    pmms = pmMagnitudeSquared();
 
    if (std::isnan(pmms) || pmms * num->julianMillenia() * num->julianMillenia() < .01)
    {
        // Ignore corrections
        *ra  = ra0().Degrees();
        *dec = dec0().Degrees();
        return false;
    }
 
    /*
    double pm = pmMagnitude() * num->julianMillenia(); // Proper Motion in arcseconds
 
    double dir0 = ((pm > 0) ? atan2(pmRA(), pmDec()) : atan2(-pmRA(), -pmDec())); // Bearing, in radian
 
    (pm < 0) && (pm = -pm);
 
    double dst = (pm * M_PI / (180.0 * 3600.0));
    //    double phi = M_PI / 2.0 - dec0().radians();
 
    // Note: According to callgrind, dms::dms() + dms::setRadians()
    // takes ~ 40 CPU cycles, whereas, the advantage afforded by using
    // sincos() instead of sin() and cos() calls seems to be about 30
    // CPU cycles.
 
    // So it seems like it is not worth turning dir0 and dst into dms
    // objects and using SinCos(). However, caching the values of sin
    // and cos if we are going to reuse them avoids expensive (~120
    // CPU cycle) recomputation!
    dms lat1, dtheta;
    double sinDst = sin(dst), cosDst = cos(dst);
    double sinLat1 = dec0().sin() * cosDst + dec0().cos() * sinDst * cos(dir0);
    lat1.setRadians(asin(sinLat1));
    dtheta.setRadians(atan2(sin(dir0) * sinDst * dec0().cos(), cosDst - dec0().sin() * sinLat1));
 
    // Using dms instead, to ensure that the numbers are in the right range.
    dms finalRA(ra0().Degrees() + dtheta.Degrees());
 
    *ra  = finalRA.Degrees();
    *dec = lat1.Degrees();
    */
 
 
    // Although it is not explained in the above reference, a formula
    // that reduces to α' = α + μ_α * t must have μ_α be the rate of
    // change of the angle at the center of the declination circle,
    // and not the rate of change of arclength, i.e. μ_α must _not_
    // include the cos(δ) factor. I have checked that the formulas
    // used in Seidelmann do reduce to the above, and therefore expect
    // μ_α = pmRa / cos(δ). Therefore, I'm absorbing the factor in the
    // implementation here.
 
    double cosDec, sinDec, cosRa, sinRa;
    double scale = num->julianMillenia() * (M_PI / (180.0 * 3600.0));
    dec0().SinCos(sinDec, cosDec);
    ra0().SinCos(sinRa, cosRa);
 
    // Note: Below assumes that pmRA is already pre-scaled by cos(delta), as it is for Hipparcos
    double net_pmRA = pmRA() * scale, net_pmDec = pmDec() * scale;
 
    double x0 = cosDec * cosRa, y0 = cosDec * sinRa, z0 = sinDec;
    double dX = - net_pmRA * sinRa - net_pmDec * sinDec * cosRa;
    double dY = net_pmRA * cosRa - net_pmDec * sinDec * sinRa;
    double dZ = net_pmDec * cosDec;
    double x = x0 + dX, y = y0 + dY, z = z0 + dZ;
 
    dms alpha, delta;
    alpha.setRadians(atan2(y, x));
 
    // Note: dec = asin(z) is a poor choice, because we aren't
    // guaranteed that (x, y, z) lies on the unit sphere due to the
    // first-order approximation. Therefore, we must "project" out any
    // change in the length of the vector to get our best estimate,
    // and this is achieved by using atan. In fact atan gives the
    // least-squares estimate for an angle given both its sin and
    // cosine components.
    delta.setRadians(atan2(z, sqrt(x * x + y * y)));
    *ra = alpha.reduce().Degrees();
    *dec = delta.Degrees();
 
    return true;
}
 
void StarObject::JITupdate()
{
    static KStarsData *data = KStarsData::Instance();
 
    if (updateNumID != data->updateNumID())
    {
        // TODO: This can be optimized and reorganized further in a better manner.
        // Maybe we should do this only for stars, since this is really a slow step only for stars
        Q_ASSERT(std::isfinite(lastPrecessJD));
 
        if (Options::alwaysRecomputeCoordinates() || (Options::useRelativistic() && checkBendLight()) ||
            std::abs(lastPrecessJD - data->updateNum()->getJD()) >= 0.00069444) // Update is once per solar minute
        {
            // Short circuit right here, if recomputing coordinates is not required. NOTE: POTENTIALLY DANGEROUS
            updateCoords(data->updateNum());
        }
 
        updateNumID = data->updateNumID();
    }
    EquatorialToHorizontal(data->lst(), data->geo()->lat());
    updateID = data->updateID();
}
 
QString StarObject::sptype(void) const
{
    return QString(QByteArray(SpType, 2));
}
 
char StarObject::spchar() const
{
    return SpType[0];
}
 
QString StarObject::gname(bool useGreekChars) const
{
    if (!name2().isEmpty())
        return greekLetter(useGreekChars) + ' ' + constell();
    else
        return QString();
}
 
QString StarObject::greekLetter(bool gchar) const
{
    QString code   = name2().left(3);
    QString letter = code; //in case genitive name is *not* a Greek letter
    int alpha      = 0x03B1;
 
    auto checkAndGreekify = [&code, gchar, alpha, &letter](const QString &abbrev, int unicodeOffset,
                                                           const QString &expansion) {
        if (code == abbrev)
            gchar ? letter = QString(QChar(alpha + unicodeOffset)) : letter = expansion;
    };
 
    checkAndGreekify("alp", 0, i18n("alpha"));
    checkAndGreekify("bet", 1, i18n("beta"));
    checkAndGreekify("gam", 2, i18n("gamma"));
    checkAndGreekify("del", 3, i18n("delta"));
    checkAndGreekify("eps", 4, i18n("epsilon"));
    checkAndGreekify("zet", 5, i18n("zeta"));
    checkAndGreekify("eta", 6, i18n("eta"));
    checkAndGreekify("the", 7, i18n("theta"));
    checkAndGreekify("iot", 8, i18n("iota"));
    checkAndGreekify("kap", 9, i18n("kappa"));
    checkAndGreekify("lam", 10, i18n("lambda"));
    checkAndGreekify("mu ", 11, i18n("mu"));
    checkAndGreekify("nu ", 12, i18n("nu"));
    checkAndGreekify("xi ", 13, i18n("xi"));
    checkAndGreekify("omi", 14, i18n("omicron"));
    checkAndGreekify("pi ", 15, i18n("pi"));
    checkAndGreekify("rho", 16, i18n("rho"));
    //there are two unicode symbols for sigma;
    //skip the first one, the second is more widely used
    checkAndGreekify("sig", 18, i18n("sigma"));
    checkAndGreekify("tau", 19, i18n("tau"));
    checkAndGreekify("ups", 20, i18n("upsilon"));
    checkAndGreekify("phi", 21, i18n("phi"));
    checkAndGreekify("chi", 22, i18n("chi"));
    checkAndGreekify("psi", 23, i18n("psi"));
    checkAndGreekify("ome", 24, i18n("omega"));
 
    if (name2().length() && name2().mid(3, 1) != " ")
        letter += '[' + name2().mid(3, 1) + ']';
 
    return letter;
}
 
// FIXME: Move this somewhere else, make this static, and give it a better name.
// Mostly for code cleanliness. Also, try to put it in a DB.
QString StarObject::constell() const
{
    QString code = name2().mid(4, 3);
 
    return KSUtils::constGenetiveFromAbbrev(code);
}
 
// The two routines below seem overly complicated but at least they are doing
// the right thing now.  Please resist the temptation to simplify them unless
// you are prepared to ensure there is no ugly label overlap for all 8 cases
// they deal with ( drawName x DrawMag x star-has-name).  -jbb
 
QString StarObject::nameLabel(bool drawName, bool drawMag) const
{
    QString sName;
 
    if (drawName)
    {
        QString translation = translatedName();
        if (translation != i18n("star") && !translation.startsWith("HD"))
            sName = translation;
        else if (!gname().trimmed().isEmpty())
            sName = gname(true);
        else
        {
            if (drawMag)
                return ('[' + QLocale().toString(mag(), 'f', 1) + "m]");
        }
        if (!drawMag)
            return sName;
        else
            return sName + " [" + QLocale().toString(mag(), 'f', 1) + "m]";
    }
    return ('[' + QLocale().toString(mag(), 'f', 1) + "m]");
}
 
//If this works, we can maybe get rid of customLabel() and nameLabel()??
QString StarObject::labelString() const
{
    return nameLabel(Options::showStarNames(), Options::showStarMagnitudes());
}
 
double StarObject::labelOffset() const
{
    return (6. + 0.5 * (5.0 - mag()) + 0.01 * (Options::zoomFactor() / 500.));
}
 
SkyObject::UID StarObject::getUID() const
{
    // mag takes 10 bit
    SkyObject::UID m = mag() * 10;
    if (m < 0)
        m = 0;
 
    // Both RA & dec fits in 24-bits
    SkyObject::UID ra  = ra0().Degrees() * 36000;
    SkyObject::UID dec = (ra0().Degrees() + 91) * 36000;
 
    Q_ASSERT("Magnitude is expected to fit into 10bits" && m >= 0 && m < (1 << 10));
    Q_ASSERT("RA should fit into 24bits" && ra >= 0 && ra < (1 << 24));
    Q_ASSERT("Dec should fit into 24bits" && dec >= 0 && dec < (1 << 24));
 
    return (SkyObject::UID_STAR << 60) | (m << 48) | (ra << 24) | dec;
}